Home

Sources of Variance in Bite Count

1. Have the participant sit at the eating table Put the Inertia cube on the dominant wrist with the cord pointing toward the elbow Put the Bite Counter above the Inertia cube on the same wrist ee I would like you to eat as you usually would You can take as much time as you like to complete the meal and I would like you to stop when you are full or when all of the food has been eaten While you eat I will be monitoring the sensor on the computer Do not start eating until I tell you to do so First I need to turn on the video camera and activate the sensor on the computer Do you have any questions before we begin Turn on the video camera Press the start stop button to begin recording The green circle should turn to red when you are recording Start scale recording by opening the scale xls file Open Summer2010 exe from the desktop Select Start Select RightHand or LeftHand 237 12 13 14 15 16 17 18 Please turn on your bite counter You may begin eating I will be sitting right here behind the divider Please let me know when you are done eating by saying I m done or I m finished Start stop watch timer When the participant says I m done a Stop the stop watch timer b Record meal time on the final meeting sheet c Stop the Intertia cube by selecting Stop d Stop the scale by pausing Winwedge Immediately save the excel fil
2. eene 71 Final review of foods drinks and details eene 71 Forgotten foods pEOITIDL 6 esoc eset oderat edt ie RU R IE VANS OH en etn Lee tap EUR 12 Bite counter data decision making process for error identification correction and removal sseeeeeeeeeeneneeeeeeeeneehenet eere nenne eee 83 ASA24 data decision making process for error identification correction and removal 0 0 ccc cece cece ce cecececececececececesececeseceeeeeeeeess 84 Example of a turning off bite counter data series esssss 85 Example of bite counter data with corrected duration and bite count sorted by meal CU ALOR e ie rt e reu EHE edes d iau t dut 86 Example of screening for a low bite count error with data sorted by Bite COMUNE m O 87 Example of a low kcal value that was removed from the data set sorted by kcal values cao edat de ied ea o ete ei c see le aca P 88 Example of an error in ASA24 that inflated the kcal value for a food 88 The Kilocalorie x Energy Density interaction demonstrating that the relationship between Kilocalories and Bites is strongest for low Energy Density meals iet tnn ee ape raa enasi GE PR VER 109 The Kilocalorie x Height interaction at the meal level demonstrating that the relationship between Kilocalories and Bites is strongest for Shorter PALU CHATS eb 119 The Kilocalorie x Energy Density interaction at the day le
3. Explain bite counter instructions by reading through the participant bite counter instructions and demoing each step Schedule 14 days of bite counter use with the earliest start date as tomorrow a Record dates on prescreening sheet and participant take home instructions Schedule 14 days of recalls a Record dates on prescreening sheet and participant take home instructions Explain ASA24 and daily meals questionnaire by reading through the participant instructions Demo both by having the participant recall two meals that they ate yesterday a Demo website http asa24demo westat com b Demo ID for survey BiteCD999 c Suggest using a small notebook provided or another immediate method e g typing into your phone to record times and important information that will help to improve recall accuracy This is not required but recommended Schedule reminders for preferred e mail address and preferred daily time a Record e mail address phone number and preferred recall time on pre screening sheet Schedule dates and time for 2 follow up meetings and record on appointment slip One date should be on the 6 7 or 8 day of data collection The other date should be the day of the last recall or the following day a Record dates and time on prescreening sheet and appointment sheet Add meetings to lab scheduler b Remind participant to bring the bite counter to both meetings and to not eat or drink anything othe
4. A copy of this consent form will be given to you 223 Appendix H Bite Counter Instructions How do I wear the Bite Counter The Bite Counter should be worn on your dominant wrist that you normally eat with It is worn like a watch The Velcro or leather strap should be adjusted so that it fits snugly When do I wear the Bite Counter Please wear the Bite Counter at all times except when exercising showering swimming or sleeping By wearing the Bite Counter during most of the day it will be easier for you to remember to turn the Bite Counter on when you are eating Warning This device is not waterproof or water resistant What is the Bite Counter default mode The default mode for the Bite Counter is Time mode The display will show the time with an arrow to the left of the screen to indicate PM when appropriate How do I use the Bite Counter to record bites during a meal Once you have prepared all of your food and you are ready to take your first bite press the left button once A beep will indicate that the device has turned on This action will turn on Bite Count mode and the device will now display the word on to indicate that it is in Bite Count mode This picture shows the Bite Counter in Time mode before the left button is pressed Press the left button to begin counting bites and to stop counting bites nd Continue to eat and drink normally 3 Once you have finished and hav
5. Note The factors described are often combined to create a multi component self monitoring intervention Future efforts to increase adherence to self monitoring could focus on improving self monitoring tools incorporating human counselor support feedback and reminders into self monitoring programs or accounting for individual differences when implementing these programs Our research group has developed a new food intake self monitoring tool the bite counter device Hoover Muth amp Dong 2009 It is possible that the bite counter will be able to simplify the food intake self monitoring process and increase adherence to self monitoring 19 However bites are a new construct in the weight loss literature In order for the bite counter to be an effective self monitoring tool the reasons why bite count may vary must be understood by both the individuals implementing a self monitoring intervention and by the people following the self monitoring intervention As a first step toward this understanding the sources of variance in bite count must be identified and studied In the next section the bite counter design and functionality is described and the foundation for predicted sources of variance in bite count is discussed The Bite Counter The bite counter is a newly invented device designed to help people self monitor their eating It is worn on the wrist like a watch and tracks a pattern of wrist roll motion in o
6. 17 What about the bite counter made it easy or difficult to use 18 In the past two weeks how much did you like or dislike using the bite counter 211 19 20 21 22 23 Extremely liked Liked very much Liked somewhat Neither liked nor disliked Disliked somewhat Disliked very much Extremely disliked What did you like or dislike about using the bite counter In the past two weeks did you have any problems wearing the bite counter due to physical discomfort or other reasons O Yes No What could be done to make it easier to wear the bite counter for longer periods of time In the past two weeks did you have any problems using the bite counter Yes O No Please describe ant problems you had with the bite counter 212 24 25 26 27 Did you feel that using the bite counter changed your eating behavior L Yes O No How did you feel the bite counter changed your eating behavior Which did you prefer using the 24 hour dietary recall or the bite counter O 24 hour dietary recall Bite counter Why did you choose the 24 hour dietary recall or the bite counter as your preferred tool 213 Appendix E Initial Participant Contact and Online Pre screening Protocol 1 Assign the interested participant the next available ID number in the Excel w
7. 225 Appendix I ASA24 Dietary Recall and Daily Meals Survey Instructions When do I complete the ASA24 dietary recall and daily meals survey Complete them every 24 hours for the previous day that you recorded your meals with the bite counter You can complete then anytime from midnight to midnight You cannot complete an ASA24 dietary recall after more than 24 hours have passed Your days of Bite Counter use Days to complete ASA24 dietary recall and daily meal survey How do I access the ASA24 dietary recall 1 In your web browser go to https asa24 westat com 2 Enter your unique participant ID 3 Enter your password How do I access the daily meals survey In your web browser go to https www surveymonkey com s dail ymeals How do I complete the ASA24 dietary recall and the daily meals survey e Start the ASA24 dietary recall first When you are on the final review page start the daily meals survey in another web browser window e Follow the instructions provided by the interviewer in the ASA24 dietary recall Report all meals foods and drinks you ate and drank during the previous day Remember to report all details of your meals including portion sizes and added foods Help buttons are available in ASA24 if you are unsure of how to complete a step in the recall e The daily meals survey will ask for additional details about each meal as well as your experience with the bite counter for each meal P
8. Additional Model with Outlier Participants Removed Further inspection of the within participant correlations between Bites and Kilocalories revealed 14 participants with correlations ranging from 0 01 to 0 3 as can be seen in Figure 3 4 The remaining 69 participants correlations were normally distributed within a range of 0 31 to 0 80 25 N Frequency H A H A eo un S Q8 A87 HM AS AS ARE D SUO S GU QU QU Ph Q9 Q9 QU Q9 9 o Qo Qo o Qo Within participant kilocalories bites correlations Figure 3 4 Within participant correlations between Kilocalories and Bites for the original 83 participants Descriptions of the quality of the data from each participant are provided in Appendix P and these 14 outlying participants are indicated by an asterisk next to the participant ID There were a number of reasons why these participants may have had poor data quality the bite counter turning off frequently during meals a broken bite counter speaker resulting in decreased turning off feedback low battery levels from not charging the bite counter difficulty remembering to turn the bite counter on and off 126 holding down the button on the bite counter to get past the calibration screen abnormal eating patterns indications that some meals may have been incorrectly reported in ASA24 feeling overwhelmed by the study requirements and a low sample size for matched meals With these justifications these 14 par
9. 0 04 0 02 8 Body weight 0 10 0 25 0 04 0 07 0 05 0 02 0 51 9 BMI 0 00 0 18 0 06 0 06 0 05 0 02 0 28 0 92 10 Height 0 05 0 29 0 02 0 06 0 03 0 01 0 71 0 61 027 11 Bite size 0 10 0 20 0 00 0 04 0 07 0 01 0 36 0 26 0 19 028 Note p 0 05 Location coded 0 Home 1 Not at Home Social coded 0 Alone 1 With Others Intake Day coded 0 Weekday 1 Weekend Gender coded 0 Male 1 Female Bite size calculated as kilocalories per bite during the lab meal 142 The final model identified in the outliers removed sample with Kilocalories Energy Density Kilocalories x Energy Density Location Social Height and Kilocalories x Height as fixed effects and Kilocalories as a random effect was run with the addition of Bite Size and Bite Size x Kilocalories as fixed effects All variables were centered at the grand mean for the data set with 60 participants When Bite Size and Bite Size x Kilocalories were added to the model the main effect of Height and the Height x Kilocalories interaction became non significant This indicated that when controlling for the effect of Bite Size Height no longer explained significant variance in Bites Location also became a non significant effect indicating that when controlling for the effect of Bite Size Location no longer explained significant variance in Bites in this sample Thus Height Height x Kilocalories and Loca
10. 214 Dear name Thank you for completing the eligibility survey for our research study being conducted by the Department of Psychology at Clemson University Your responses have indicated that you are eligible to participate in the study I would like to schedule a meeting with you to provide participation instructions and your wrist worn device This meeting will take approximately one hour Please let me know some times that you are available to meet within the next week insert dates here and I will select a time for this meeting Sincerely Jenna Scisco Department of Psychology Clemson University 864 656 1144 4 a When the participant responds send the following e mail to schedule the meeting Dear name Thank you for your response We will have your first meeting at insert time on insert day We will meet in Brackett Hall room 422 for approximately one hour Please bring your personal calendar to this meeting This will allow us to schedule two follow up meetings and your two weeks of participation Sincerely Jenna Scisco Department of Psychology Clemson University 864 656 1144 5 If the participant is eligible and there are not available bite counters send the following e mail for future participation Dear name Thank you for completing the eligibility survey for our research study being conducted by the Department of Psychology at Clemson University Your 215 responses have indicated
11. 4 inch round down e g 1 2 inch Record height and weight values on the prescreening sheet Measure the participant s body fat percentage using the handheld Omron device PRT Press blue On button Will flash Guest Press Set Will flash Normal Press Set Use Up and Down to enter height weight age and gender Press Set after each Will say Ready Have participant stand with feet shoulder width apart Ask them to grasp both sides of the analyzer firmly with their arms straight out in front of them at a 90 degree angle to the floor 219 10 11 13 13 f Press Start g Record BMI and body fat percentage on the pre screening sheet Measure waist and hips using the MyoTape a Waist is the smallest circumference typically just above the belly button b Hips are the largest circumference around the buttocks c Record measurements on the pre screening sheet Explain study instructions broadly For this study you will be wearing a device called the Bite Counter on your wrist during the day for two weeks This device can measure how much you are eating just like a pedometer can measure how much you are exercising Then each day after you use the bite counter you are going to use your computer to tell me about the foods that you ate some features of the meal and your experience with the bite counter First we will go over the bite counter how it is used and when you will use it
12. Different methods of dietary assessment have been thoroughly reviewed by Thompson and Subar 2008 who have identified a number of advantages of 24 hour dietary recalls The immediacy of the recall period helps participants to recall most of their intake Additionally in comparison to keeping food records participants find 24 hour recalls less burdensome This reduces selection bias and allows for a more representative sample Also dietary recalls occur after the food has been consumed which reduces the chance of the assessment method interfering with food and drink selection and consumption The main weakness of the 24 hour dietary recall is that participants may not report their intake accurately due to problems with knowledge or memory Thompson and Subar s 2008 review of the literature indicates that 12 underreporting of energy using 24 hour dietary recalls ranges from 3 to 26 with underreporting affecting up to 15 of all recalls However the interviewer prompts and multiple pass approach of the AMPM 24 hour recall are designed to reduce underreporting Thompson amp Subar 2008 Ina controlled study of adult men AMPM dietary recall accurately estimated energy intake regardless of BMI Conway Ingwersen amp Moshfegh 2004 In a controlled study of adult women AMPM dietary recall resulted in overestimation of energy intake by 8 10 and there were no energy recall differences between normal and obese women Conway Ingwers
13. 005 15 53 2 13 Note y00 grand mean of bites y10 kilocalories bites slope y20 energy density bites slope y30 location bites slope y40 social bites slope y50 intake day bites slope y120 kilocalories x energy density interaction y01 gender bites slope y04 height bites slope y14 kilocalories x height interaction p lt 05 123 The day level model had significant within participants variance and between participants variance as can be seen in Table 3 10 In Table 3 11 it can be seen that all of the significant relationships in the meal level model remain in the day level model The significant positive relationship between Kilocalories and Bites and the significant negative relationship between Energy Density and Bites were qualified by the significant Kilocalories x Energy Density interaction In order to examine the nature of the interaction simple slopes were calculated in accordance with Cohen et al 2003 using the fixed effects coefficients at high 1 SD and low 1 SD values of Kilocalories These slopes were significant at low B 0 034 SE 0 003 t 10 06 p lt 05 moderate B 0 03 SE 0 002 t 12 33 p lt 05 and high B 0 026 SE 0 003 t 9 57 p lt 05 values of Energy Density Figure 3 3 shows that the relationship between Kilocalories and Bites is strongest for days with overall lower Energy Density However when compared to Figure 3 1 which shows the relation
14. 19 The results from this initial study indicate that slowing bite rate with the bite counter may be most effective for reducing energy intake for individuals who consume larger amounts of food These first studies with the bite counter were conducted in laboratory settings and were limited to only one meal or one food consumed by an individual Ideally the bite counter will be used by an individual for months or years to self monitor their food intake in their daily life Therefore it is necessary to determine the variables that will explain variance in bite count in order to guide long term bite counter use in real life settings Bite Count Variance An assumption of the bite counter method is that bites will serve as a proxy for energy intake As number of bites taken during a meal increases for an individual we assume that this increase will equate to an increase in energy intake However there are a number of other reasons why bite count may vary We can parse these potential explanatory factors into within person variance and between person variance in bite count For example analyses of 24 hour dietary recalls have indicated that about half of the variation in daily energy consumption kcal day is due to differences within people with the other half being due to differences between people Beaton et al 1979 Although bite count variance and energy intake variance are not the same the present study is assuming that they are positi
15. 2005 The NWCR has examined a primarily female Caucasian and married sample Wing amp Hill 2001 Therefore it is possible that successful weight loss maintenance strategies may differ in other populations In a review of 42 randomized clinical trials of weight maintenance conducted from 1984 through 2007 a number of behaviors associated with successful weight loss maintenance were identified including medications e g orlistat consuming a lower fat diet adherence to physical activity continued contact with individuals problem solving therapy increased protein intake increased caffeine intake and acupressure Turk et al 2009 Some researchers have also addressed behavioral differences between individuals who have successfully maintained weight loss and those who have regained weight Kayman Bruvold and Stern 1990 interviewed and surveyed weight loss maintainers and relapsers and discovered that although both groups used similar strategies to lose weight maintainers more frequently adapted these weight loss strategies to their own lifestyle That is maintainers more often devised their own personal eating and exercise plan whereas relapsers were more likely to use a specific program like Weight Watchers Relapsers used more restrictive diets and negative life events caused them to relapse back to their old behaviors Maintainers also distinguished themselves by self monitoring their eating and weight In another study Krug
16. 775 784 Prentice A M 1998 Manipulation of dietary fat and energy density and subsequent effects on substrate flux and food intake American Journal of Clinical Nutrition 67 535S 541S Rhodes D G Cleveland L E Murayi T amp Moshfegh A J 2007 The effect of weekend eating on nutrient intakes and dietary patterns FASEB Journal 21 6 A1064 Rock C L Flatt S W Sherwood N E Karanja N Pakiz B amp Thomson C A 2010 Effect of a free prepared meal and incentive weight loss program on weight loss and weight loss maintenance in obese and overweight women A randomized control trial Journal of the American Medical Association 304 1803 1810 doi 10 1001 jama 2010 1503 Rolls B J 2007 The Volumetrics Eating Plan New York Harper Rolls B Ello Martin J Ledikwe J 2005 Portion size and food intake In D J Mela Ed Food Diet and Obesity Boca Raton FL CRC Press Rolls B J Fedoroff I C amp Guthrie J F 1991 Gender differences in eating behavior and body weight regulation Health Psychology 10 2 133 142 Rolls B J Morris E L amp Roe L S 2002 Portion size of food affects energy intake in normal weight and overweight men and women American Journal of Clinical Nutrition 76 1207 1213 R ssner S Hammarstrand M Hemmingsson E Neovius M amp Johansson K 2008 Long term weight loss and weight loss maintenance strategies Obesit
17. PA which imported real time data into Microsoft Excel The participant wore an InteriaCube3 InterSense Inc Bedford MD on their dominant wrist with a bite counter above their wrist on the lower part of the forearm The meal was video recorded Participants were instructed to eat normally and to stop eating when they felt full or when all of the food had been eaten Satiety before and after the meal was measured using the Satiety Labeled Intensity Magnitude SLIM scale Cardello Schutz Lesher amp Merrill 2005 Appendix M Liking or disliking the meal was measured after the meal using the Labeled Affective Magnitude LAM scale Schutz amp Cardello 2001 Appendix N At the conclusion of this laboratory session the participant was debriefed and received the 50 incentive for participation Statistical Analyses Data Merging and Error Screening Data was prepared for statistical analysis using Microsoft Excel Each participant s data was merged and screened for errors individually The steps for merging 81 the data from three sources bite counter data files ASA24 Individual Food and Nutrient INF data file and Survey Monkey daily meals questionnaire data files are outlined in Appendix O Date and time were the primary indicators used to merge the data sets After the data was merged it was screened for errors using the steps outlined in Appendix O Errors originated from the bite counter device failure or user error and t
18. Therefore self monitoring is an essential part of self regulation but self regulation will only be successful if an individual also has clear and reasonable comparison standards as well as a way to enact a behavioral change Kanfer 1971 has also described a model of self regulation with three sequential stages 1 self monitoring 2 self evaluation and 3 self reinforcement In this model an individual begins the self regulation process by self monitoring one s behavior and attending to response feedback which can be proprioceptive sensory or affective Then 12 an individual engages in self evaluation and compares the feedback to the performance criteria used to judge the feedback The performance criteria originates from the individual s history including task standards social norms prior reinforcements and motivation for success The outcome of this comparison is judged as less than the standard at the standard or greater than the standard and the individual self reinforces positively or negatively based on the outcome The individual may decide to engage in a new behavior continue with the current behavior or end the behavior based on their evaluation Once again it is clear that self monitoring is an important component of self regulation but it should be used in combination with self evaluation and self reinforcement to ensure that behavior change is successful Kanfer 1970 Kanfer amp Gaelick 1986 Bandura 1
19. and Exit Carver 1979 An example of a TOTE loop for weight loss is presented in Figure 1 2 First an individual compares their goal weight to their current weight In the first Test if there is a discrepancy between the two weights e g the individual weighs more than their goal weight an Operation takes place and the individual eats less and or exercises more Then the individual engages in another Test to determine if their current weight matches their goal weight If there is no longer a discrepancy the individual Exits the loop If there is a discrepancy the loop continues with another Operation 11 TEST OPERATION pati Is there a Eat less and went discrepancy exercise more Figure 1 2 A basic TOTE feedback loop example for weight loss The TOTE feedback loop was restated by Carver and Scheier 1990 as a cycle of outside impacts from the environment input functions or perceptions a comparator making use of reference values and output functions or behaviors In this self regulatory process an individual compares their perception to a standard and if a discrepancy exists the individual will adjust their behavior to reduce or eliminate the discrepancy The self regulation feedback loop requires three things to function 1 standards for a clear comparison point 2 monitoring in order to track the state of the current system and 3 a way to change behavior in the case of a discrepancy Baumeister et al 1994
20. e g working talking and cooking Additionally longer meal times may indicate meals eaten with others and thus they may reflect the social facilitation of energy intake Preliminary analyses from our research group for 38 meals indicated that bite count and meal time are very strongly correlated r 2 875 p lt 05 Research Question 4 Does meal duration predict the number of bites recorded during a meal Meal location One environmental factor that can affect consumption is meal location Many Americans consume meals outside of their homes at restaurants and fast food locations and the number of commercially prepared meals eaten per week has 52 increased in recent years Kant amp Graubard 2004 This increase in eating outside of the home is associated with an increase in kilocalories consumed Kant amp Graubard 2004 Increased energy intake outside of the home is partly the result of large portion sizes at these locations that are often much larger than recommended serving sizes Condrasky Ledikwe Flood amp Rolls 2007 Ledikwe Ello Martin amp Rolls 2005 Humans use environmental cues like portion size to guide food intake therefore restaurants portions may cue us to consume more food Wansink 2010 For example in a laboratory study that manipulated portion size participants ate 30 more kilocalories when offered a large portion of macaroni and cheese compared to a small portion Rolls Morris amp Ro
21. gt 05 indicated that the addition of the Kilocalories x Gender interaction did not significantly improve the model fit In addition the interaction term was nonsignificant 0 01 Therefore the varying Kilocalorie Bites slopes could not be explained by the Gender of the participant The cross level interaction term was dropped from subsequent models Exploration of Additional Level 2 variables With the significant random slope variance for the relationship between Kilocalories and Bites additional Level 2 variables individual difference variables were explored to determine if they might help explain this variation Hox 2010 Model 11 was determined to be the best model with five fixed predictors at level 1 Kilocalories Energy Density Location Social and Intake Day a Kilocalorie x Energy Density interaction at level 1 one fixed predictor at level 2 Gender and the significant random slope variance between Kilocalories and Bites All exploratory models were compared to model 11 to see if model fit would improve and if the random slope variance could be explained Model estimates are provided in Table 3 7 and Table 3 8 114 Table 3 7 Estimates of model fit and random effects for model 11 and exploratory models Model fit Random effects Model parameters 2LL e SE 100 SE TM 11 12 28312 93 378 09 9 74 164 56 28 26 00041 lt 001 17 14 28310 25 378 10 9 74 161 17 27 60 00041 lt 001 18 14 2831
22. of meals were eaten on weekends This is expected for 2 out of every 7 days being weekends 28 6 For Social 5 participants ate alone for over 90 of their meals and 1 participant ate with others for over 90 of their meals Across all meals for all participants 61 1 of meals were eaten alone and 38 9 of meals were eaten with others Because the majority of participants had acceptable variability for the dichotomous predictors all data was retained at this step Then the level 2 continuous variable Body Weight was examined for correct values outliers normality and linearity with descriptive statistics a histogram a q q plot and a bivariate scatterplot with Bites Skewness and kurtosis values and graphs indicated a normal distribution of body weight and no evidence of nonlinearity The level 2 dichotomous variable Gender was split almost evenly with 40 males and 43 females Next multivariate outliers among all level 1 predictors were identified within each participant using Mahalanobis distance Values were obtained by running a regression for each participant with all level 1 predictors entered and saving Mahalanobis distance values A Mahalanobis distance value greater than 20 515 the critical y value for p lt 001 and df 5 the number of IVs indicated the presence of a multivariate outlier Tabachnick amp Fidell 2007 Twenty meals were identified as multivariate outliers The sources of these outliers were examined and they
23. the dependent variable was meal level bite count This model has two levels of predictors Level 1 is meal level predictors features of the meals measured repeatedly across all meals which could impact bite count Level 2 is individual level predictors features of an individual that could impact bite count Main effects of each predictor at each level on bite count were tested MLM also allows within level and cross level interaction effects to be tested An example of the hierarchical data structure for two individuals for this two level model is shown in Figure 1 12 45 Individual Individual Figure 1 12 The two level model with meals at level land individuals at level 2 In the sections below possible predictors of bite count are identified at the two levels meal and individual When available previous research relevant to the selected predictors is described Because bites the dependent variable is a new construct in the literature empirical support is not always available However support for these predictors is drawn from research using calories or grams of food as outcome measures with the assumption that bites may serve as a proxy for the amount of food an individual consumes In particular the research by John de Castro and colleagues that investigated the predictors of energy intake in free living humans using a diet diary methodology is an excellent source that is used to support many of the research questio
24. this research is to detect food intake during the day Your participation will involve e completing a short form about yourself e completing a survey about your eating behavior e having your height weight body composition waist and hips measured e wearing a wrist worn watch like device called the Bite Counter during meals and throughout the day e completing daily questionnaires about what you ate and related behaviors during the previous day e completing a post study interview about your eating habits during the study and about the Bite Counter and diet questionnaires e eating one meal in the laboratory that will be video recorded The amount of time required for your participation will be about 1 hour day of participation up to 14 consecutive days You may be paid a maximum of 50 for participating You may also receive a data summary including Bite Count and dietary recall records Risks and Discomforts There are certain risks or discomforts associated with this research They include increasing sensitivity to food intake during the day For this reason individuals with a current or previous eating disorder are asked not to participate in this study Potential Benefits There are no direct benefits to you for participating in this study However this research may help us to understand food intake patterns during the day and improve our device for measuring food intake Protection of Confidentiality We will do everything
25. 0 people Logarithmic transformation of Number of People reduced skewness and kurtosis values somewhat and a histogram of Number of People revealed visible positive skew and positive kurtosis An inverse transformation of Number of People did not improve skewness and kurtosis and skew became highly negative Since neither transformation seemed to adequately correct the variable the decision was made to create a dichotomous predictor variable named Social with the groups Alone or With Others which could still represent social facilitation of eating The new variable Social is described in more detail with the other dichotomous predictors at level 1 below Meal Energy Density had positive skewness and kurtosis values within participants Removal of Bites and Meal Kilocalorie outliers did not improve Meal Energy Density skewness and kurtosis values Examination of plots revealed that the positive skewness and kurtosis were most likely due to a few high energy density meals reported by participants that differed from the energy density of the majority of their 94 meals In order to determine if transformation of this variable was appropriate and to examine linearity bivariate scatterplots of Bites and Meal Energy Density were examined within participants The scatterplots were mostly linear and oval shaped indicating that transformation of this variable was not necessary However the plots did reveal that for some participants there were a few ou
26. 16 254 1 Sugar white granulated or lump Figure 2 14 Example of an error in ASA24 that inflated the kcal value for a food 88 Another error found in the ASA24 data files were missing values for kcals and grams If the missing values were missing because the participant failed to report all food details or because the pathway of questions failed to prompt the participant these meals then had missing kcal and gram values However in one instance the missing food was the result of a database writing error for apple juice Although the participants reported apple juice type and amount consumed found in the My Selection file these drinks showed up as a missing value in the Individual Foods and Nutrients file Upon request from the author the ASA24 nutritionist provided information that could be used to replace missing values one ounce of apple juice was equal to 31 grams and 14 26 kcals Multiplying the amount reported by the participant resulted in amounts that could replace missing values For example if a participant reported drinking 100 of a 12 oz glass of apple juice then 372 grams and 171 12 kcals of apple juice were inserted to replace the missing values Multilevel Linear Modeling Analysis Data were analyzed using IBM SPSS Statistics 19 Data were cleaned using the guidelines provided by Tabachnick amp Fidell 2007 for cleaning grouped data The MLM analysis began with an intercepts only model null model without predicto
27. 19 37 36 46 14 15 2 06 45 13 Duration sec 02 26 06 03 25 20 18 05 06 05 91 43 14 Rate kcal min 26 04 13 33 35 49 21 16 22 11 41 96 47 15 Rate bites min 12 22 29 30 02 08 33 10 07 23 19 05 19 18 Note p lt 05 151 Body Measurements Height and weight were self reported during pre screening and BMI was calculated from height and weight as pounds inches x 703 Height weight BMI body fat percentage and waist to hip ratio WHR were measured at the beginning of the two week study and again at the end of the two week study Means standard deviations and results of within subjects t tests are reported in Table 3 26 for 82 study participants Participant BiteCD232 was excluded from body measurement comparisons due to third trimester pregnancy Overall participants overestimated their height and underestimated their weight resulting in an underestimation of BMI for self report Participants lost an average of 0 5 pounds over the course of the two week study equivalent to an average BMI reduction of 0 1 Table 3 26 Body measurements from self report pre study and post study Measurements Min Max M SD t Mean difference Self report Height inches 60 0 77 0 679 3 9 Weight pounds 102 0 275 0 168 7 39 8 BMI 17 7 39 5 25 6 5 0 Pre study Pre study Self report Height 600 760 67 5 37 497 0 4 Weight 102 4 288 4 171 5 42 0 4 75 2 8 BMI 17 1 424 2
28. 2 1 10 variance of the student level GPA values at job status 2 1 11 variance of the student level slopes 1 11 variance of the student level slopes Magnitude direction and statistical significance Magnitude direction and statistical significance Magnitude direction and statistical significance Greater than expected by chance Reduction in value from Q4 Greater than expected by chance Reduction in value from Q6 43 MLM and Eating Research This MLM analysis technique is particularly useful for eating research when the same participant s eating behaviors are measured at multiple meals In a traditional two level hierarchical structure each meal can be defined as Level 1 with multiple meal level predictors occurring at this level Then each individual can be defined as Level 2 with multiple individual level predictors occurring at this level The goal of the MLM analysis would be to determine the direct effect of meal and individual level explanatory variables on the Level 1 outcome e g bites and to determine if the individual level variables serve as moderators of the meal level relationships Hox 2010 MLM has been used to successfully analyze repeated measures eating behavior data For example O Connor Jones Conner McMillan and Ferguson 2008 used MLM to analyze daily diary reports of hassles and between meal snacking Using a two level hierarchical structure
29. 2005 Results from the usability questionnaire indicated that participants became much more aware of what they were eating how much they were eating and when they were eating This increased awareness could be attributed to completing the ASA24 dietary recall and using the bite counter daily although the unique effect of each one cannot be completely disentangled The ASA24 most likely made them more aware of food details and quantity whereas the bite counter most likely increased their awareness of when they were eating and meal duration since the device had to be turned on and turned off This awareness of food intake may have provided opportunities for individuals to make behavioral changes such as deciding to turn the device off and stop eating when feeling full or when a certain amount of food had been consumed 178 The weight loss observed in the study also may have resulted from decreased snacking Participants described not wanting to snack because they did not want to turn the device on again for something small and or because they did not want to report another meal in ASA24 It appears that the costs of the minimal effort to use the bite counter and or the greater effort of entering a snack into ASA24 sometimes outweighed whatever benefits participants might have obtained from snacking However suggesting that individuals reduce snacking to lose weight actually goes against current guidelines from the Academy of Nutrition and Dietet
30. 2009 Version 1 of ASA24 became available in September 2011 and is available free of charge to researchers A demo version of AS A24 can be found here http asa24demo westat com The ASA24 interview process has five steps 1 Meal based Quick List 2 Meal Gap Review 3 Detail Pass 4 Final Review and 5 Forgotten Foods During the first step the Meal based Quick List participants were asked to select an eating occasion breakfast brunch lunch dinner supper snack or just a drink specify the time and location of the meal indicate if a TV and or computer was used during the meal and indicate if the meal was eaten alone or with others Figure 2 1 Then the participants added the main foods and drinks for each meal to the Quick List Figure 2 2 In the second step the Meal Gap Review participants were asked if they consumed anything during all gaps between eating occasions that exceeded three hours Figure 2 3 If the 67 participants responded yes they returned to the Quick List to add the food and or drink In the third step the Detail Pass participants were asked to provide details for the foods and drinks recorded in the Quick List including the amount eaten and anything added to the main foods Figure 2 4 and Figure 2 5 During the Final Review participants were asked to review all foods drinks and details and to make edits if appropriate Figure 2 6 Next participants were asked if they consumed any commonly for
31. 83 5 79 01 24 3E 4 1 53 003 51 C8 87 002 26 4E 4 id 30 87 04 549 185 5 76 Q01 96 002 148 6003 51 C82 87 002 42 TE 4 Note 00 grand mean of bites y10 kilocalories bites slope y20 energy density bites slope y30 location bites slope y40 social bites slope y50 intake day bites slope y120 kilocalories x energy density interaction y01 gender bites slope y02 body weight bites slope y03 BMI bites slope y04 height bites slope y12 kilocalories x body weight interaction y13 kilocalories x BMI interaction y14 kilocalories x height interaction p 05 134 The Final Model for the Outliers Removed Sample Model 19 was the best fitting model for explaining variance in bites The overall effect size for this model was 0 431 Bickel 2007 Therefore the final model explained 43 1 of the overall variance in bites for the outliers removed sample This was an improvement over the model for the full sample which explained 23 3 of the variance in bites The fixed coefficients from model 19 shown in Table 3 17 above indicate the nature of the relationships between predictors and Bites for the final model for the outliers removed sample The positive relationship between Kilocalories and Bites and the negative relationship between Energy Density and Bites were main effects that were qualified by a significant interaction between Kilocalories and Energy Density The
32. AMPM interview method Another study that manipulated presentation of the serving size photographs found that eight photographs allow for more accurate estimations than four photographs and participants preferred seeing all serving size options at once rather than sequentially Subar et al 2010 Overall the ASA24 dietary recall was selected for use in the proposed study because it is based on a well validated intake measure the AMPM recall that results in accurate energy intake reports and because it will allow for inexpensive and practical 24 hour dietary recalls from a large sample The 24 hour recall is considered the best self report instrument available for estimating dietary intake and we can assume that the measure is unbiased across persons Kirkpatrick 2011 For the present study the ASA 4 recall data provided the number of kilocalories consumed at each meal the average energy density of each meal the date and time of the meal the meal location and whether the meal was consumed alone or with others Body Measurements Tanita WB 3000 Digital Beam Scale Body weight height and BMI were measured using the Tanita WB 3000 Digital Beam Scale Tanita Corp Arlington Heights IL 74 Omrom Body Logic Body Fat Analyzer Body fat percentage was measured using the Omrom Body Logic Body Fat Analyzer Omron Corp Kyoto Japan This hand held device analyzes the impedance of a small electrical current flowing between two elec
33. H amp Mermier C M 2000 Predictive accuracy of Omron Body Logic Analyzer in estimating relative body fat of adults International Journal of Sport Nutrition and Exercise Metabolism 10 216 227 Glick W H 1985 Conceptualizing and measuring organizational and psychological climate Pitfalls in multilevel research Academy of Management Review 10 3 601 616 Goetzel R Z Gibson T B Short M E Chu B Waddell J Bowen J et al 2010 A multi worksite analysis of the relationships among body mass index medical utilization and worker productivity Journal of Occupational and Environmental Medicine 52 S1 S52 S58 doi 10 1097 JOM 0b013e3181c95b84 Gokee Larose J Gorin A A amp Wing R R 2009 Behavioral self regulation for weight loss in young adults A randomized controlled trial International Journal of Behavioral Nutrition and Physical Activity 6 10 doi 10 1186 1479 5868 6 10 Goldstein D J 1992 Beneficial health effects of modest weight loss International Journal of Obesity 16 397 415 Goodpaster B H DeLany J P Otto A D Kuller L Vockley J South Paul J E et al 2010 Effect of diet and physical activity interventions on weight loss and cardiometabolic risk factors in severely obese adults Journal of the American Medical Association 34 1795 1802 doi 10 1001 jama 2010 1505 Grizzle J W Zablah A R Brown T J Mowen J C amp Lee J M 2009 Employ
34. Home 1 Not at Home Social coded 0 Alone 1 With Others Intake Day coded 0 Weekday Weekend Gender coded 0 Male 1 Female 101 Model 1 Is Nesting Present The Intercepts Only Model Model building began with an intercepts only model with Bites as the DV participants as the grouping variable and no predictors ICC1 the ratio of between participants variance to total variance was 0 24 This indicated that 2446 of the variance in bites was between participants and 76 of the variance in bites was within participants Nesting was present and MLM analysis could be used to explain variance at both levels Heck Thomas amp Tabata 2010 Model fit statistics and estimates of random effects for the intercepts only and subsequent models are presented in Table 3 5 to allow for comparison across models Similarly estimates of fixed effects for all models are presented in Table 3 6 As can be seen in Table 3 4 the null model consisted of both significant within participants variance 563 43 and significant between participants variance 180 57 that could be explained by the addition of level 1 and level 2 predictor to the model 102 Table 3 5 Estimates of model fit and random effects Model fit Random effects j x ZEE ei HOUSE um ee ToU d e E 1 3 29468 49 563 43 14 30 180 57 30 93 2 4 28722 97 442 76 11 24 192 22 32 18 3 5 28617 07 428 28 10 87 186 19 31 19 4 6 28611 13 427 43 10 8
35. InterCeptintercepts onty 1 408 13 162 59 563 434 180 57 0 233 Therefore the final model explained 23 3 of the overall variance in bites The fixed coefficients from model 19 shown in Table 3 8 above indicate the nature of the relationships between predictors and Bites for the final model The positive relationship between Kilocalories and Bites and the negative relationship between Energy 119 Density and Bites were main effects that were qualified by a significant interaction between Kilocalories and Energy Density The simple slopes of the Kilocalories x Energy Density interaction term for model 19 had the same values as the simple slopes for Model 11 and were still significant Thus the size and the nature of the interaction did not change and Figure 3 1 was still appropriate for interpretation of the interaction for the final model The relationship between Kilocalories and Bites depended on the Energy Density of the meal being eaten with a stronger relationship between Kilocalories and Bites for meals of lower Energy Density The relationship between Location and Bites remained nonsignificant in the final model Therefore when controlling for the effects of the other predictors Location was no longer a significant predictor of Bites The relationship between Social and Bites remained significant and indicated that on average participants took 5 73 more bites when eating with others than when eating alone The relation
36. Raynor amp Fava 2006 or observational VanWormer et al 2009 Thus it is possible that increased frequency of self weighing leads to weight loss or successful weight loss encourages an individual to self weigh more frequently A series of experimental studies have partially addressed this issue of causality by manipulating self weighing behavior with results indicating that more frequent self weighing is related to weight loss Gokee Larose Gorin amp Wing 2009 Levitsky Garay Nausbaum Neighbors amp DellaValle 2006 Strimas amp Dionne 2010 Interestingly Strimas and Dionne 2010 concluded that individual differences may moderate the relationship between self weighing frequency and weight loss Also interactions between self weighing and other parts of a weight loss program are important future directions for investigation VanWormer et al 2008 Self monitoring of body weight allows an individual to compare his or her weight to a goal weight However a limitation of this approach is that weight alone does not provide information about how to change the behaviors that impact weight change Weight can fluctuate one to two pounds per day which is similar to weight loss recommendations of one to two pounds per week which provides a challenge to an individual trying to assess the source of weight loss on a weekly basis Additionally the mechanisms behind weight change in self weighing studies are difficult to isolate because se
37. also include the time eating began and ended eating rate and perhaps even foods consumed and kilocalorie estimates if bite counter recordings are paired with an eating diary This combination of information could be very useful for an individual trying to change their eating patterns For example if someone sees that they typically eat all of the their daily meals in under 10 minutes and they would like to begin increasing their meal durations in order to slow their overall eating rate they could use the eating calendar to help them accomplish this goal The bite counter could also have an additional feature indicating how long someone has been eating like a stop watch to provide real time feedback about Meal Duration 165 Meal Location Research question 5 investigated if meal location could predict bite count Specifically meals eaten outside of the home were compared to meals eaten at home Within participant correlations between location and bites were 0 05 and 0 06 for the full sample and the outliers removed sample respectively Total correlations between location and bites were 0 04 and 0 07 for the full sample and the outliers removed sample respectively These small correlations indicated that participants might take more bites when eating outside of the home than when eating at home For both the full sample and the outliers removed sample the kilocalories by energy density interaction explained about 0 2 0 3 of the variance in b
38. applications pp 1 9 New York Guilford Press Wadden T A Berkowitz R I Womble L G Sarwer D B Phelan S Cato R K 2005 Randomized trial of lifestyle modification and pharmacotherapy for obesity The New England Journal of Medicine 353 2111 2120 Wadden T A amp Letizia K A 1992 Predictors of attrition and weight loss in patients treated by moderate and severe caloric restriction In T A Wadden amp T B Vanitallie Eds Treatment of the seriously obese patient New York Guilford Press Waldon H M Martin C K Ortego L E Ryan D H amp Williamson D A 2004 A new dental approach for reducing food intake Obesity Research 12 11 1773 1780 Wallinga D 2010 Agricultural policy and childhood obesity A food systems and public health commentary Health Affairs 29 405 410 doi 10 1377 hlthaff 2010 0102 Wansink B 2010 From mindless eating to mindlessly eating better Physiology amp Behavior 100 454 463 Welsh E M Sherwood N E VanWormer J J Hotop A M amp Jeffery R W 2009 Is frequent self weighing associated with poorer body satisfaction Findings from a phone based weight loss trial Journal of Nutrition Education and Behavior 41 6 425 428 Westertep Plantega M S Westerterp K R Nicholson N A Mordant A Scoffelden P F amp Ten Hoor F 1990 The shape of the cumulative food intake curve in humans during basic and manip
39. being able to do any other activities and while being video recorded They frequently told the experimenter about almost always doing something else while they eat and as a result the lab meal felt strange or uncomfortable to them Thus it is possible that participants ate quickly in order have the lab meal experience end as soon as possible Future research should aim to create a more natural eating environment in which participants are free to do other activities or perhaps eat with others Because bite size should remain constant within individuals introducing other activities and a social 177 element should not overly influence bite size although the features of laboratory meals that could impact bite size should also be topics of future research Mishra et al 2012 Weight Loss Participants lost an average of 0 5 pounds over the two week study period Weight loss was not a goal of the study and the study was not advertised as such However 42 4 of participants who completed the study were trying to lose weight and they used the study as an opportunity to help them self monitor their eating behaviors Recruitment at the beginning of January was particularly successful as some of these participants used the study to kick off their New Year s weight loss resolution The significant weight loss was most likely the result of self monitoring eating over two weeks a behavior that is consistently related to weight loss Wadden et al
40. bite size to the model eliminated the significant height main effect and interaction with kilocalories This suggests that bite size would be a better individual difference variable that could be used to calibrate a bite counter kilocalorie setting Individuals with smaller bite sizes could be given a smaller kilocalorie multiplier and individuals with larger bite sizes could be given a larger kilocalorie multiplier Also individuals with different bite sizes might need to be given different bite reduction goals Individuals with larger bite sizes may need to reduce their intake by fewer bites than individuals with smaller bite sizes in order to reduce energy intake These recommendations make an assumption that bite size is constant across meals for the same person There is evidence in the literature that bite size is fairly constant within individuals Medicis amp Hiiemae 1998 Westerterp Plantega et al 1990 with greater variation between individuals Hutchings et al 2009 This has been observed in our own laboratory study during which participants consistently took the same bite sizes kcals bite of the same food over three separate sessions but there was greater variation in bite size between participants Salley et al 2011 Hence it follows that a person s bite size could serve as a calibration step for the bite counter This idea is analogous to calibrating a pedometer or step counter for 175 running or walking Before
41. bites slope y120 kilocalories x energy density interaction y01 gender bites slope y04 height bites slope y14 kilocalories x height interaction p lt 05 138 Examining the fixed effects in Table 3 19 it can be seen that the interaction between Kilocalories and Energy Density became nonsignificant in the day level model for the outliers removed sample This indicated that when variability was reduced by aggregating to the day level the relationship between Kilocalories and Bites no longer depended on Energy Density Thus the main effects of Kilocalories and Energy Density were interpreted For every additional Kilocalorie consumed during a day participants took 0 04 more bites on average Stated in a more practical way for every 25 Kilocalories consumed participants took 1 more bite on average Also for every 1 point increase in daily energy density participants took 31 81 fewer bites on average Location also became nonsignificant in the day level model This indicated that when variability was reduced by aggregating to the day level Location was no longer a significant predictor of Bites Social remained significant and indicated that for each additional meal eaten with someone else participants took 3 80 more bites on average The significant positive relationship between Height and Bites was qualified by a significant cross level interaction between Height and Kilocalories 139 In order to examine the nature of th
42. counseling in an Internet weight loss program Archives of Internal Medicine 166 1620 1625 Tate D F Wing R R amp Winett R A 2001 Using Internet technology to deliver a behavioral weight loss program Journal of the American Medical Association 285 1172 1177 doi 10 1001 jama 285 9 1172 Thompson F E amp Subar A F 2008 Dietary assessment methodology In A M Coulston amp C J Boushey Eds Nutrition in the prevention and treatment of disease 2 ed pp 3 39 Boston Academic Press Turk M W Yang K Hravnak M Sereika S M Ewing L J amp Burke L E 2009 Randomized clinical trials of weight loss maintenance A review Journal of Cardiovascular Nursing 24 58 80 263 VanWormer J J French S A Pereira M A amp Welsh E M 2008 The impact of regular self weighing on weight management A systematic literature review International Journal of Behavioral Nutrition and Physical Activity 5 54 doi 10 1186 1479 5868 5 54 VanWormer J J Martinez A M Martinson B C Grain A L Benson G A Cosentino D L amp Pronk N P 2009 Self weighing promotes weight loss for obese adults American Journal of Preventive Medicine 36 1 70 73 doi 10 1016 j amepre 2008 09 022 Vohs K D amp Baumeister R F 2004 Understanding self regulation An introduction In R F Baumeister amp K D Vohs Eds Handbook of self regulation Research theory and
43. data folder This Word document is used to keep a record of what has been done to the data in Excel for this participant Bite Counter data Original Bite Counter data is in Dissertation Data Bite CounterRaw ParticipantID Files are named by participant number device number and download date e g BiteCD001_Device1413_Oct132011 There are typically two files per participant because data was downloaded twice Data was cleared off of the device after the first download Thus data will not repeat from the first file to the second file a Copy all of the original bite counter data and paste it into the first sheet of ParticipantID xls Each row on this sheet represents a recording period by the bite counter ultimately a meal Daily meals questionnaire data In Dissertation Data SurveyMonkey Daily meals questionnaire ParticipantID CSV open Sheet 1 csv and Sheet 2 csv for this participant originally downloaded as an Advanced Spreadsheet from Survey Monkey using a participant ID filter a The data is split into two csv files by Survey Monkey but can be combined into one to make merging the data easier Sheet 2 is just an extension of Sheet 1 Simply copy the data from Sheet 2 and paste it onto the end of Sheet 1 Save Sheet 1 yes keep it a csv file b Using meal date and time match the daily meals questionnaire data to the bite counter data This is made easier if the two spreadsheets are viewed side by side Copy each meal
44. data from meal to meal If they did see that they tended to take more bites when eating outside of the home they could try to target these locations as an opportunity to reduce the number of bites being taken Social Research question 6 investigated if eating with others versus eating alone could predict bite count Within participant correlations between social and bites were 0 25 and 0 28 for the full sample and the outliers removed sample respectively Total correlations between social and bites were 0 23 and 0 27 for the full sample and the outliers removed sample respectively These positive correlations for social were the second largest correlations with bites found for the tested model and indicated that participants took more bites when they ate with others than when they ate alone Social explained 1 9 and 2 1 of the within participants variance for the full sample and the outliers removed sample respectively This was the second largest unique effect at the meal level for the tested model The slopes between social and bites at the meal level were 5 73 and 5 76 167 for the full sample and the outliers removed sample respectively These slopes indicated that participants took about 5 to 6 more bites during meals that they ate with others compared to meals that they ate alone Translated to the average number of kilocalories consumed per bite this equates to eating 125 to 150 additional kilocalories during a meal eaten with oth
45. goals Classical and operant conditioning form the basis for behavior therapy Clark et al 2010 Associations among activities locations mental states eating behaviors and physical activity behaviors are identified i e behaviors that are classically conditioned are identified and behaviors are rewarded or punished based on how they affect these weight loss goals operant conditioning Self monitoring allows the individual to examine his or her own behaviors identify where changes can be made and then monitor the results of those behavioral changes Recording food intake activity weight types of food amounts of foods caloric values of foods times places and feelings can all provide insight into associations that may be contributing to an individual s obesity Clark et al 2010 For example an individual tracking her food intake may realize that she always eats ice cream when watching TV even when she is not hungry This individual can then set a goal of no longer eating ice cream when watching TV and only eating ice cream at a table when feeling hungry If the individual engages in behaviors that help her to reach this goal then the individual may see a positive result such as a 14 weight loss of one pound over a week This positive reinforcement leads to the continuation of this new eating behavior pattern The theoretical basis for self regulation theory and behavioral therapy both describe self monitoring as an essen
46. how to charge the bite counter The written instructions in Appendix H were reviewed in person and provided in a folder for the participant to take home The experimenter instructed the participant to record all meals and snacks However if a meal or snack was going to last for a very long time such as drinking coffee and nibbling on candy for over an hour at one s desk at work or drinking a glass of wine in the evening while making dinner the participant was told not record this intake because it would be too difficult to define a meal end time The participant was given a username and password for the ASA24 system The participant completed a demonstration of the ASA24 program by entering two meals from their previous day The experimenter was available for guidance and to answer questions The participant was also shown how to complete the daily meals questionnaire on Survey Monkey The participant was instructed to complete this questionnaire during the ASA24 Final Review so that meal details could be matched with the ASA24 entries The participant received basic written instructions for completing the ASA24 program and the daily meals questionnaire see Appendix I The participant was also given a 50 page spiral notebook 3 x 5 to make notes about meal times and foods Using this notebook was optional and participants were encouraged to use other methods for taking notes if more convenient such as on their mobile phone or personal comp
47. interpretations of each symbol are provided in Table 1 5 Tabachnick amp Fidell 2007 GPA Bo PB jobstatus e 1 1 Table 1 5 Symbols and Meanings for the Level 1 Equation Symbol Meaning i The measurement occasion nested within an individual j The individual GPA T The GPAs for measurement occasions i in individuals j the DV Bo j For an individual j the mean intercept of GPA P4 i For an individual j the slope of the relationship between GPA and job status jobstatus j The job status scores for measurement occasions i in individuals j the level 1 IV amp j Deviation of predicted GPA values from actual GPA values for measurement occasions i in individuals j the error term for the level 1 equation The level 2 model is shown in equations 1 2 and 1 3 The mean GPAs of the individuals Bp jp and the slopes of the relationship between GPA and job status for the individuals B4 p become DVs in equations 1 5 and 1 6 Bickel 2007 Hox 2010 The interpretations of each new symbol are provided in Table 1 6 Tabachnick amp Fidell 2007 40 Boj Yoo Yor gender Hoj 1 2 Bj Vio Y1i gender pa 1 3 Table 1 6 Symbols and Meanings for the Level 2 Equations Symbol Meaning Yoo The grand mean of GPA scores across all individuals when all predictors are zero Yo1 The overall regression coefficient for the relationship slope between gender and GPA gender The gen
48. level model and the day level model for the outliers removed sample Significant within participants variance between participants variance and random Kilocalories Bites slope variance remained in the day level model This differed from the day level model for the full sample which did not have significant random Kilocalories Bites slope variance Therefore the cross level interaction between Kilocalories and Height was retained 137 Table 3 18 Random effects for the meal level and the day level models for the outliers removed sample 101 Model ei SE t00 SE Kcalories Meal level 347 37 9 61 139 64 25 83 00034 001 Day level 1507 75 74 42 1860 48 340 78 00016 001 Note SE Standard Error eij residual within participant variance 100 random intercept between participants variance t10 random slope variance p lt 05 Table 3 19 Fixed effects for the meal level and the day level models for the outliers removed sample Model y00 y10 y20 y30 y40 y120 y04 yl4 SE SE SE SE SE SE SE SE 39 87 04 5 49 1 85 5 76 01 96 002 MERE 1 48 003 51 82 87 002 42 9E 4 ise 125 33 04 31 81 95 3 80 01 449 002 7 5 42 003 4 99 1 25 1 45 005 1 54 9E 4 Note 00 grand mean of bites y10 kilocalories bites slope y20 energy density bites slope y30 location bites slope y40 social bites slope y50 intake day
49. meal or snack Edit a meal or snack Delete a food or drink Move a food or drink Copy a food or drink Edit a food or drink Done entering all meals and snacks Undo Finish later was eaten with Find a Food or Drink Browse the categories or search using the box below Suggestions Coca Cola Coca Cola C2 Cereals and energy bars Chicken turkey poultry Dairy dairy substitutes p Desserts and sweets Eggs Fats Oils Dressings Spreads Fish shellfish Fruit p Meat Miscellaneous See Chan Medan Chili Other Pancakes waffles crepes Pasta noodles and spaghetti p Pizza calzones hot pockets No match Found l Figure 2 1 Selecting a meal time location computer and or TV use and who the meal My Foods and Drinks Me late and drank Yesterday Wednesday October Lunch 12 00 PM McDonald s Double Quarter Pounder with Cheese Figure 2 2 Adding foods and drinks to the Quick List for lunch 69 UJ Actions d Select an action below to edit your foods and drinks Add a meal or snack Delete a meal or snack Edit a meal or snack amp Delete a food or drink Move a food or drink _ Copy a food or drink s e Edit a food or drink 12 j Done entering all meals and sn A 4j Undo Finish later 00 PM to 08 00 PM W Find a Food or Drink Browse the categories or search
50. of my friends for their support here at Clemson and long distance via phone and e mail Special thanks go to Dr Stephanie Fishel Brown for providing feedback on manuscript drafts and never ending encouragement iv TABLE OF CONTENTS TITLE PAGE p i ABSTRACT ocupa ee Petes te eat ie E eh sok at ME Dd ae e ii DEDICATION pes ieren a Mi RR DUI a ae NEU E a a ge a iii ACKNOWLEDGMIN JES e Ras e a DIS O R DS iv LIST OF TABLES riac huno du ie b See ee dane a ae vii ETS OR FIGURES s ties nde ata uteha ona utu a ea esas X CHAPTER i NIDRODECPION s doses pret eta bau D mde ed 1 PUT POSE P 1 oci secede ERR 1 Sell MORBILOEIBE o osi eR One rear bead seq us OE ete S M cg uen fus 10 Whe Bite Counter iiie Et eb rer e debat ipod eel rec dees 20 Multilevel Linear Modeling of Eating Behavior 27 The Present Bite Counter Study stccicanuasercasncnass seated 45 I METHODS eet dh etg en aee d e a k sa ae iala 59 ParlctDanisutesadoenetu estet na biais s peeled iuo iet ato 59 Materni alsine aoa eto enl a eetptom buen aput dudes ab Letto AUnd 66 Procedure i insi absenin e e aS apad to cesis cod iet E E a Caldas 76 statistical Analyses ceste oisi esa ts eii ene eto ede pecie aad cited 8l Ir DRE SUITS dicdstaanonetetest Moped testae paid ME 92 Original Pri 92 MIM AIaDy SIS ierit disp scans sor tese pde ped
51. ooo tse tad A RE e EBEN EIN idS 31 Symbols and Meanings for the Level 1 Equation esses 40 Symbols and Meanings for the Level 2 Equations eesses 41 Symbols and Meanings for the Random Variance Components 42 Research questions with their corresponding parameter estimates figures and mnterpretattODS os ne cast irae carte tena coe Cond oam ie ones 43 Demographic statistics used to guide sample recruitment and selection 63 Demographic characteristics of the 83 study participants 65 Frequencies and percentages of participant meal reporting and meal featutes osea pesti SO DERE Ria e esae tes ue A Rod I Ae 98 Descriptive statistics for the meal level 1 and participant level 2 variables ctio oci La oce la R dere o is tau 99 Within participant correlations between level 1 variables 100 Total correlations between level 1 and level 2 variables 101 Estimates of model fit and random effects sss 103 Estimates of fixed effects for level 1 and level 2 predictors 104 Estimates of model fit and random effects for model 11 and exploratory Inadels aces cese quo ee putei qe b ere UETUUO ue M nra RIS eon ERR 115 vii List of Tables Continued Table 3 8 3 9 3 10 3 11 3 12 3 13 3 14 3 15 3 16 3 17 3 18 3 19 3 20 3 2
52. relationship between kilocalories consumed during a meal and number of bites recorded during a meal depend on the energy density of the food Meal duration Prior research has demonstrated a positive relationship between meal duration and the amount of food consumed For example in a laboratory study that manipulated meal duration participants were given either 12 or 36 minutes to eat a meal 51 consisting of pizza cookies and bottled water and they ate almost 100 kilocalories more during the longer meal Pliner Bell Hirsch amp Kinchla 2006 In another study that manipulated music playing during the meal listening to music was associated with longer meal times and increased food intake Stroebele amp de Castro 2006 Examination of diet diary studies has shown that meal size and meal time are positively correlated r 0 20 to r 0 54 de Castro 1991 de Castro 2010 Feunekes de Graaf amp van Staveren 1995 At a broader level over the past 30 years the amount of time Americans spend eating each day has increased about half an hour for men from 2 0 h to 2 4 h and almost an hour for women from 1 6 h to 2 5 h a finding that parallels rising obesity rates Zick amp Stevens 2011 As time elapses during a bite counter recording session it is likely that more bites are taken as people eat more food It is also possible that a longer meal will allow people to engage in more activities that could trigger false bite counts
53. sample size is 60 likely to be adequate For example Grizzle Zablah Brown Mowen and Lee 2009 examined predictors of employee customer oriented behavior and unit profits with a two level multilevel model Individuals were ate level 1 and restaurants were at level 2 An average of about 17 employees was nested within each of 38 restaurants for a total sample size of 671 Six variables and two cross level interactions were entered into the model As another example Erdogan and Bauer 2010 examined the effects of leader member exchange on employee outcomes and the moderating role of justice climate Individuals were at level 1 and stores were at level 2 An average of about 11 respondents was nested within each of 25 stores for a total sample size of 276 Seven variables and one within level interaction were entered into the model The present study had an average of 39 meals with Bite Counter and ASA24 data nested within 83 individuals and a total sample size of 3 246 meals with Bite Counter and ASA24 data This was much larger than these studies and was sufficient for running the MLM analysis which estimated up to 14 parameters see Results section for a description of the parameters Sample Recruitment and Compensation Clemson University students and employees were recruited using an e mail announcement sent to graduate students an Inside NOW e mail announcement and flyers hung on announcement boards in campus buildings Community
54. simple slopes for the outliers removed sample were calculated in accordance with Cohen et al 2003 using the fixed effects coefficients at high 1 SD and low 1 SD values of Kilocalories These slopes were significant at low B 0 05 SE 0 003 t 15 13 p lt 05 moderate B 0 04 SE 0 003 t 14 90 p lt 05 and high B 0 03 SE 0 003 t 9 87 p lt 05 values of Energy Density As can be seen in Figure 3 5 the relationship between Kilocalories and Bites depended on the Energy Density of the meal being eaten with a stronger relationship between Kilocalories and Bites for meals of lower Energy Density 135 60 50 a 40 e 30 ss m Low ED i 20 Average ED 10 4 HighED 0 4 Low Kilocalories High Kilocalories Figure 3 5 The Kilocalorie x Energy Density interaction for the outliers removed model demonstrating that the relationship between Kilocalories and Bites is strongest for meals with lower Energy Density The relationship between Location and Bites remained significant in the final model This differed from the model for the full sample for which Location became a nonsignificant effect The relationship between Location and Bites indicated that on average participants took 1 85 more Bites when eating out of the home than when eating at home The relationship between Social and Bites remained significant and indicated that on average participants took 5 76 more bites when eating with othe
55. that can be medical e g a heart attack death of a spouse or emotional e g a hurtful comment about one s weight Klem et al 1997 Second an individual forms goals and engages in self regulation that involves an eating plan regular exercise and regular self weighing Third the weight loss goal has been achieved and an individual actively maintains weight loss through self regulation and cognitions about food and weight maintenance strategies Fourth an individual reaches transcendence or an integration of weight maintenance into one s lifestyle In theory behaviors that once took much effort are now automatic and easier for the weight maintainer 1 Ahah epiphany moment triggering event relevant to one s weight 2 Goal formation and self regulation of eating exercise and weight 3 Maintenance of weight through self regulation and cognitions 4 Transcendence integration of weight maintenance into one s lifestyle Figure 1 1 The four stage process of weight maintenance described by Haeffele 2008 Continuous efforts to describe successful weight loss maintenance are led by the National Weight Control Registry NWCR The NWCR established in 1994 is the largest ongoing study of successful weight loss maintenance with over 5 000 contributing individuals NWCR 2011 This registry tracks people who have entered at least the third stage in the weight loss maintenance process they have lost
56. that you are eligible to participate in the study and I look forward to your participation At this time all of the wrist worn devices for the study are in use or are reserved I have added you to the study waiting list As soon as a device becomes available for you I will contact you to set up a time for our first meeting This is an ongoing study and you may be contacted anytime from current month year to April 2012 Sincerely Jenna Scisco Department of Psychology Clemson University 864 656 1144 5 a Add the participant to the waiting list in the ParticipantIDinfo spreadsheet 6 If the participant is not eligible send the following e mail Dear insert participant s name here Thank you for completing the eligibility survey for our research study being conducted by the Department of Psychology at Clemson University Your responses have indicated that you are not eligible to participate in the study Sincerely Jenna Scisco Department of Psychology Clemson University 864 656 1144 7 When a device becomes available select the next participant from the waiting list and send the following e mail Dear name Good news We currently have an opening in our study and would like to begin your participation I would like to schedule a meeting with you to provide participation instructions and your wrist worn device This meeting will take approximately one hour Please let me know some times that you are avail
57. the x axis Given this plot the second question that can be asked of the data set is does gender predict GPA As seen in Figure 1 8 on average females have higher GPAs than males A line has been fit to the data to demonstrate that this would typically be shown for variables with more than two values and to demonstrate the group differences 32 3 5 4 q 34 2 5 3 2 4 1 Male Female Gender Figure 1 8 Scatterplot demonstrating the average difference in GPA between the genders Q3 Does the relationship between job status and GPA depend on gender Figure 1 9 shows all of the GPA measurements for all students and all years with GPA on y axis and job status on the x axis Given this plot the third question that can be asked of the data set is does the relationship between job status and GPA depend on gender It can be seen in Figure 1 9 that the relationship between job status and GPA does appear to depend on gender with an overall increase in GPA for males when they work more hours per day and an overall decrease in GPA for females when they work more hours per day 23 3 5 A m Males amp 3 3 Females C 25 B ee 3 elt s ESSAS Linear Males L 777 a 2 T T l i Linear 0 1 2 3 4 Females Job Status Figure 1 9 Scatterplot demonstrating how the relationship between job status and GPA depends on gender The next four questions Q4 Q7 discussed in
58. this test the device goes into Sensor Test mode During the sensor test you should slowly roll the device away from you and then back towards you as if it were being rolled on the wrist The numbers on the display should go positive and then negative and a corresponding auditory cue will go high and low in pitch You should do this rolling motion once or twice and at some point stop the rolling motion in any position When the motion is stopped and the device held steady the number should stay within 10 and the sound will cease Charge the device overnight Day of orientation l Prepare participant s take home folder It should include a eo Bes ASA24 Dietary Recall and Daily Meals Survey Instructions i Assign the participant a password from the password excel spreadsheet Write password and unique participant ID on these instructions Bite Counter instructions Appointment slip Small notebook Extra copy of consent form 2 Prepare participant s in lab folder a Consent form 218 b Download Survey Monkey prescreening data for the participant as a PDF Include open ended responses Print and add to participant folder c Add prescreening sheet Label pre screening sheet with participant number date and time Add age bite counter number ASA24 user name ASA24 password email and phone number if provided to the sheet 3 Get out scale MyoTape and body fat analyzer Confirm that they a
59. using the box below 4 3 Enter search term 4 Breads other baked goods i Bagels p Biscuits Breads U My Foods and Drinks Y ne late and drank Yesterday Wednesday October Lunch 12 00 PM McDonald s Double Quarter Pounder with Cheese Y French fries y Coca Cola i Dinner 08 00 PM HU Chicken Caesar salad Tea hot or iced regular Chips Ahoy Snack 10 30 PM Did you eat or drink anything that you haven t already reported between 12 00 PM and 08 00 PM Pies tarts B Rolls buns Sweet breads coffee cakes sweet rais Tortillas taco shells other shells p Other QQ No match Found Chicken Caesar salad How much of the salad did you actually eat Select the image that best represents the amount you ate at dinner No image avacbe 4 cups Undo amp Finis 3 1 2 cups amp 3 cups Less than y 2 029 1 2 cup cups V 1 2 cup e 1 cup bed 11 2 cups Amount eaten 9 2 1 2 cups 2 cups 4 Previous Don t Know gt Figure 2 4 Portion size question for salad during the Detail Pass 70 Did you add anything to your Tea hot or iced that you haven t already reported Select all that apply Search or browse to find foods added to your Tea hot or iced Use the arrows to add or remove additions If nothing was added or you have already reported the additions select Nothing Added below Whole m
60. was to examine predictors of number of bites taken during a meal by humans in their natural environments Participants wore bite counters and recorded bite count during daily meals Participants also recorded their daily dietary intake using 24 hour recalls Predictors of bite count were explored at the meal level and person level using multilevel linear modeling This was one of the first studies to provide long term bite count data an essential first step for determining sources of variance in bite count Obesity Obesity has been identified as a major public health problem worldwide The World Health Organization WHO has declared obesity a global epidemic with an estimated 1 6 billion overweight adults and 400 million obese adults in 2005 WHO 2011 The WHO predicts that by 2015 2 3 billion adults will be overweight and 700 million adults will be obese WHO 2011 In the United States the National Health and Nutrition Examination Survey NHANES data collected in 2007 2008 indicated that 33 9 of Americans were obese and 68 3 of Americans were overweight Flegal Carroll Ogden amp Curtin 2010 Throughout the continuous NHANES data collection from 1999 to 2008 obesity rates have remained fairly steady at about one third of the US population These WHO and NHANES population estimates are based on the current standards for measuring obesity and overweight A body mass index BMI kg m of 30 or greater defines obesity and
61. 00 students each of whom is a member of 50 different classrooms If a researcher was interested in predicting academic performance there may be individual characteristics such as socio economic status SES of the child which might predict academic performance However there may also be features of the classroom such as teacher experience that might predict performance as well Thus it would be important to consider the relationship between SES and academic performance within the context of the teacher experience in each of the classrooms The students are considered nested or grouped within the classrooms Data can also be nested when it comes from repeated measurements for the same individuals over time Cohen Cohen West amp Aiken 2003 For example in the present study human eating behavior is being measured over time The variation in the number of bites recorded by the bite counter may be due to differences in the eating occasions such as the energy of the food eaten at each occasion However there may also be differences between individuals that affect how many bites are recorded such as gender or body weight Therefore it is important to consider the relationship between bites and the amount of energy consumed within the context of each individual s gender and body weight The eating occasions are nested or grouped within the individuals An analysis technique that allows for nested data is multilevel linear modeling MLM al
62. 07 Additionally repeated measures data is likely to have missing values due to participant drop out or a participant missing a measurement occasion In repeated measures analysis a participant with a missing measurement occasion would be removed from the data set completely In MLM this participant can remain in the data set Hox 2010 In the immediate text that follows a simple example is used to conceptually demonstrate the research questions that can be answered with MLM Starting with the raw data shown in Table 1 4 there are five students whose GPA was measured at five different years 2007 2008 2009 2010 and 2011 When GPA was measured job status 29 was also measured and defined as the average number of hours worked per day 0 hours unemployed 1 hour 2 hours 3 hours or 4 or more hours The gender of each student is also known The data is nested because the GPA and job status measurements can be grouped by the individual student who provided the data GPA is the dependent variable DV job status is the level 1 independent variable IV and gender is the level 2 IV Level 1 refers to a variable measured at the lowest level of analysis in this case the measurement occasion level Level 2 refers to a variable measured at the second level of analysis in this case the individual level The first three questions Q1 Q3 discussed in this example reflect the fixed effects in MLM Fixed effects examine the overall relatio
63. 1 Estimates of fixed effects for level 1 and level 2 predictors for model and exploratory inodels ift pe ese tee eH HM D RI eoa RESET 116 ICC2 values for level 1 Variables eue eee eae 121 Random effects for the meal level and day level models 123 Fixed effects for the meal level and the day level models 123 Within participant correlations between level 1 variables for outliers removed model nien ette teteeed 127 Total correlations between level 1 and level 2 variables for the outliers removed model 2 idiota enue etra deb tesa mnes 129 Estimates of model fit and random effects for the outliers removed iri orari Ne T OT a A a 132 Estimates of fixed effects for the level 1 and level 2 predictors for ihe outhers removed model e Gees 133 Estimates of model fit and random effects for model 11 and exploratory models for the outliers removed model sess 134 Estimates of fixed effects for level 1 and level 2 predictors for model 11 and exploratory models for the outliers removed model 134 Random effects for the meal level and the day level models for the outliers removed sample ederet ee ead o deb iun dde 138 Fixed effects for the meal level and the day level models for the ou tliers removed samples ue astedose Mdh meten noes e etra e 138 Within participant correlations between level 1 variables for bite size model
64. 1080 08870449808407422 Barte J C M ter Bogt N C W Bogers R P Teixeira P J Blissmer B Mori T A et al 2010 Maintenance of weight loss after lifestyle interventions for overweight and obesity a systematic review Obesity Reviews 11 899 906 doi 10 1111 j 1467 789X 2010 00740 x Basiotis P P Thomas R G Kelsay J L amp Mertz W 1989 Sources of variation in energy intake by men and women as determined from one year s daily dietary records American Journal of Clinical Nutrition 50 448 453 Baumeister R F Heatherton T F amp Tice D M 1994 Losing control How and why people fail at self regulation San Diego Academic Press Beasley J 2007 The pros and cons of using PDAs for dietary self monitoring Letter to the Editor Journal of the American Dietetic Association 107 739 Beaton G H Milner J Corey P McGuire V Cousins M Stewart E Little J A 1979 Sources of variance in 24 hour dietary recall data Implications for nutrition study design and interpretation American Journal of Clinical Nutrition 32 2546 2559 Bell E A Castellanos V H Pelkman C L Thorwart M L amp Rolls B J 1998 Energy density of foods affects energy intake in normal weight women American Journal of Clinical Nutrition 67 412 420 Bell E A amp Rolls B J 2001 Energy density of foods affects energy intake across multiple levels of fat content in l
65. 2 04 378 14 9 74 163 22 28 03 00041 lt 001 19 14 28306 08 378 09 9 74 157 83 26 94 00037 001 Note 2LL 2 log likelihood SE Standard Error eij residual within participant variance 100 random intercept between participants variance 110 random slope variance Significant model improvement from previous significant model using the Chi square deviance difference test p 05 115 Table 3 8 Estimates of fixed effects for level 1 and level 2 predictors for model 11 and exploratory models y00 y10 y20 y30 y40 y50 y120 y01 y02 y03 y04 y12 y13 yl4 d Parameters SE SE SE SE SE SE SE SE SE SE SE SE SE SE T B 40 27 04 5 84 75 5 76 1 85 0l1 3 14 1 47 002 50 82 85 82 002 2 50 T T 40 05 04 5 85 76 5 75 1 87 O1 1 92 06 8E 5 1 46 003 50 82 85 0 2 002 2 78 04 6E 5 i 40 27 04 5 86 75 5 76 1 85 01 2 82 24 1E 4 1 46 003 50 82 85 82 002 2 50 25 4E 4 i 4 40 94 04 5 81 79 5 73 188 01 1 50 83 002 1 45 003 50 82 86 82 002 3 70 54 7E 4 Note 00 grand mean of bites y10 kilocalories bites slope y20 energy density bites slope y30 location bites slope y40 social bites slope y50 intake day bites slope y120 kilocalories x energy density interaction y01 gender bites slope y02 body weight bites slope y03 BMI bites slope y04 height b
66. 25 3 Neither liked nor disliked 38 45 8 Disliked somewhat 12 14 5 Disliked very much 1 1 2 Problems wearing physical discomfort Yes 19 22 0 No 64 77 1 Experienced problems with bite counter Yes 36 43 4 No 47 56 6 Bite counter changed eating behavior Yes 43 51 8 No 40 48 2 Preferred tool Bite counter 63 75 9 AS A24 dietary recall 20 24 1 Overall the most difficult aspect of the bite counter was remembering to turn it on and off Some participants found it harder to remember as they became more accustomed to wearing the bite counter when at social functions or when engaged in other activities while eating Some participants had trouble remembering to charge the 157 device at night and some participants had difficulty remembering to wear the device The device was also frustrating when it would shut off automatically during meals and when the display malfunctioned Participants disliked that it was not waterproof that it could not be worn during exercise that it got in the way of long sleeves and jackets and that they did not receive bite count or charging feedback from the device In terms of physical discomfort and appearance the device was described as unattractive uncomfortable too big bulky cumbersome not trendy and ugly Some participants found the Velcro to be irritating and some participants disliked having something on their wrist A few participants wanted a longer wristband so th
67. 3 and a US quarter 23 Figure 1 6 The ambulatory bite counter used in the current study Possible applications of the bite counter for weight loss are numerous In the first study of a bite counter application the bite counter s utility for slowing bite rate and reducing energy intake was explored Scisco Muth Dong amp Hoover 2011 The study was a within participants design with three conditions Thirty university students ate three meals in the laboratory while wearing the bite counter a baseline meal without feedback Baseline a meal during which participants received bite rate feedback Feedback and a meal during which participants followed a 50 slower bite rate target Slow Bite Rate Bite rate feedback was provided by displaying participant s bites in real time on a step graph with the x axis representing time elapsed and the y axis representing number of bites taken Overall participants ate 70 fewer kilocalories during the Slow Bite Rate condition compared to the Feedback condition Additionally when baseline energy intake was added post hoc as a grouping variable participants who ate over 400 kilocalories at baseline n 11 ate 164 fewer kilocalories during the Slow Eating condition compared to Baseline and 142 fewer kilocalories in the Feedback 24 condition compared to Baseline However the Slow Bite Rate condition did not significantly affect participants who ate under 400 kilocalories at baseline n
68. 3 4 Definitely false Mostly false Mostly true Definitely true 200 10 11 12 13 14 Only at meal times Sometimes between meals 15 16 17 18 When I feel lonely I console myself by eating us false VS false true I consciously hold back at meals in order not to weight gain iets false Misi false s true I do not eat some foods because they make me fat Uses false E false d true I am always hungry enough to eat at any time Definitely false Mens false js true How often do you feel hungry 1 2 3 Often between meals How frequently do you avoid stocking up on tempting foods 3 Almost never Seldom Usually How likely are you to consciously eat less than you want 1 2 3 Unlikely Slightly likely Moderately likely Do you go on eating binges though you are not hungry 2 3 Never Rarely Usually 4 Definitely true 4 Definitely true 4 Definitely true 4 Definitely true 4 Almost always 4 Almost always 4 Very likely 4 Almost always On a scale of 1 to 8 where 1 means no restraint in eating eating whatever you want whenever you want it and 8 means total restraint constantly limiting food intake and never giving in what number would you give yourself 1 2 3 4 5 6 7 201 8 Appendix C Daily Meals Questionnaire The following questions will help the researchers link your questionnaire responses to the AS A24 dietary recall l Please en
69. 31 off during a few 12 Good meals One over BiteCD242 47 66 2 608 Good 13 esuumatce meal corrected Participant tried to hold down the button BiteCD245 40 90 9 407 to get past 14 Good calibration for the first week BiteCD246 46 100 0 419 Good 14 Good Bite counter turned BiteCD251 51 100 0 769 off during a few 14 Good meals Bite counter turned BiteCD258 53 96 4 572 off during a few 14 Good meals BiteCD260 71 87 7 652 Good 14 Good BiteCD261 35 92 1 613 Good 13 Good BiteCD266 43 87 8 643 Good 13 Good BiteCD268 44 66 7 402 Good 14 Good BiteCD270 15 51 7 423 Good 10 Good Note Outlier with a Bites Kilocalories correlation 0 31 Some participants completed extra recalls to make up for missing Bite Counter days 248 REFERENCES Academy of Nutrition and Dietetics 2012 Adult Weight Management Evidence Based Nutrition Practice Guideline Evidence Analysis Library http www adaevidencelibrary com topic cfm cat 3014 Retrieved 21 March 2012 Baker R C amp Kirschenbaum D S 1993 Self monitoring may be necessary for successful weight control Behavior Therapy 24 377 394 Baker R C amp Kirschenbaum D S 1998 Weight control during the holidays Highly consistent self monitoring as a potentially useful coping mechanism Health Psychology 17 367 370 Bandura A 1998 Health promotion from the perspective of social cognitive theory Psychology and Health 13 623 649 doi 10
70. 49 6 77 1 54 001 41 83 87 2 38 62 03 4 11 1 05 720 232 154 001 41 85 88 86 6 38 84 04 6 12 84 646 2 02 01 1 54 001 50 84 88 85 002 T 38 92 04 611 88 6 53 L82 01 2 38 154 000 50 84 88 C87 000 170 r t 39 16 04 6 3 81 641 202 01 6 18 1 51 001 05 84 88 85 00D 3 02 in 39 14 04 6 14 80 642 203 01 4 18 05 149 00 50 84 88 85 001 3 344 04 T i 40 27 04 584 75 5 76 1 85 01 3 14 1 47 002 50 82 85 82 002 2 50 1s 40 33 04 5 86 27 5 87 1278 Q01 3 31 148 003 50 97 86 83 002 2 62 T i 40 33 04 5 75 75 574 170 Q01 4 70 146 002 50 83 1 06 82 002 2 65 Y i 40 27 04 583 74 5 76 191 01 3 31 147 003 50 82 85 83 002 2 51 T 40 56 04 584 79 571 L87 01 6 03 01 147 003 50 82 85 82 002 2 93 01 Note Model 12 estimates were unstable and thus were not included y00 grand mean of bites y10 kilocalories bites slope y20 energy density bites slope y30 location bites slope y40 social bites slope y50 intake day bites slope y120 kilocalories x energy density interaction y560 social x intake day interaction y01 gender bites slope y02 body weight bites slope y11 gende
71. 5 186 86 31 30 5 7 28550 99 419 41 10 64 183 99 30 81 6 8 28543 71 418 42 10 62 184 26 30 84 7 9 28497 60 412 27 10 46 184 17 30 80 8 10 28495 70 411 98 10 46 184 79 30 90 9 10 28493 51 412 27 10 46 174 64 29 31 10 11 28491 73 412 27 10 46 170 72 28 69 11 12 28312 93 378 09 9 74 164 56 28 26 0004 001 127 13 28479 86 407 46 10 40 175 23 29 30 a 13 13 28308 11 374 35 9 99 167 01 28 99 0004 001 18 90 17 72 14 13 28305 56 373 29 9 78 160 94 27 65 0004 001 28 16 15 52 15 13 28312 81 2377 91 9 76 164 84 28 30 0004 001 0 45 2 79 16 13 28309 40 378 03 9 74 162 45 27 69 0004 lt 001 Note Model 12 estimates were unstable and thus were not included 2LL 2 log likelihood SE Standard Error ej residual within participant variance 100 random intercept between participants variance 110 random slope variance Significant model improvement from previous significant model using the Chi square deviance difference test p 05 103 Table 3 6 Estimates of fixed effects for level 1 and level 2 predictors y00 y10 y20 y30 y40 y50 y120 y560 y01 y02 yll Mousse A Parameters SE SE SE SE SE SE SE SE SE SE SE f 3 40 24 1 54 38 51 04 2 4 1 57 C001 s 38 50 04 428 155 000 4D F P 38 51 04 431 204 1 55 001 41 84 3 38 65 03 413 1
72. 5 19 39 26 3 7E 08 1200 64 2011 10 31 12 42 34 20 00 Lunch 12 42 PM 50 3 7E 08 1302 91 2011 11 5 23 45 18 21 42 Dinner 10 45PM 56 3 7E 08 1343 40 2011 11 7 16 46 29 22 23 Justa drin 4 46PM 36 3 7E 08 1348 96 2011 11 2 22 25 42 22 28 Dinner 10 25 PM 17 3 7E 08 1439 98 2011 10 29 23 21 31 23 59 Dinner 11 20 PM 53 3 7E 08 1460 72 2011 11 6 23 00 54 24 20 Dinner 11 00 PM Figure 2 11 Example of bite counter data with corrected duration and bite count sorted by meal duration Examining the number of bites recorded also allowed for detection of possible errors For example participant BiteCD051 had a recording of 8 bites for meal 4 The 86 associated meal data was then examined to see if the bite count value was reasonable This meal was a breakfast of 250 8 kcal of white bread that lasted 3 minutes and 30 seconds In Figure 2 12 it can be seen that this participant had a number of shorter meals with similar kcal and or bite values Based on all of this associated information it was decided that this data was most likely correct and the meal was retained MeallD 31 5 2011 10 12 11 24 53 AM 1 37 Snack 99 28 Cookie brownie fat free without icing 24 6 2011 10 10 12 34 41 PM 1 32 Snack 94 64 Apple raw 45 pi 2011 10 17 2 50 46PM 6 40 Missing st 77 48 Apple raw 4 8 2011 10 5 9 51 15 3 30 Breakfast 250 8 Bread white made from home recipe or purcha 30 9 2011 10 12 7 38 35 1 59 Breakfast 215 3905
73. 6 117 159 Snijders T A B amp Bosker R J 2011 Multilevel analysis An introduction to basic and advanced multilevel modeling Qo ed Thousand Oaks CA SAGE Sobal J amp Nelson M K 2003 Commensal eating patterns A community study Appetite 41 2 181 190 doi 10 1016 S0195 6663 03 00078 3 Sobal J amp Wansink B 2008 Built environments and obesity In E M Blass Ed Obesity Causes Mechanisms Prevention and Treatment Sunderland MA Sinauer Associates Sperduto W A Thompson H S amp O Brien R M 1986 The effect of target behavior monitoring on weight loss and completion rate in a behavior modification program for weight reduction Addictive Behaviors 11 337 340 Spiegel T A Kaplan J M Tomasinni A amp Stellar E 1993 Bite size ingestion rate and meal size in lean and obese women Appetite 21 131 145 Strimas R amp Dionne M M 2010 Differential effects of self weighing in restrained and unrestrained eaters Personality and Individual Differences 49 1011 1014 Stroebele N amp de Castro J M 2006 Listening to music while eating is related to increases in people s food intake and meal duration Appetite 47 285 289 262 Strychar I Lavoie M E Messier L Karelis A D Doucet E Prud homme D et al 2009 Anthropometric metabolic psychosocial and dietary characteristics of overweight obese postmenopausal women with a histo
74. 6 AM 46 995 12 25 PM 52 095 4 47 PM 77 48 10 52 AM 77 48 1 Milk malted dry mix fortified t 1 Chicken noodle soup 1 Grapes raw NS as to type 1 Apple raw 1 Apple raw Figure 2 13 Example of a low kcal value that was removed from the data set sorted by kcal values Additional errors identified in the ASA24 data were large kcal values that stemmed from food entry errors or ASA24 program errors When participant BiteCD056 s data was sorted by meal kcal a large meal of 1678 kcal was found the largest meal for this participant Inspection of the food kcal values as shown in Figure 2 14 indicated that 1269 of the kcal came from a report of two cups of whole dry milk This participant frequently reported drinking whole milk but not dry milk Additionally two cups of dried milk was judged to be an excessive amount to consume at one meal so it was assumed that the participant reported this food incorrectly Therefore the values were converted to two cups of whole milk 296 kcal and the meal was reduced to 705 kcal mou uw NO NONO ON 3 6 00PM 421 516 3 6 00 PM 3 6 00PM 3 6 00PM 3 6 00 PM 256 52 167827 138 32 76 32 125 1 19119 vu vu vn v2 vai ription 1 Bread wheat or cracked wheat 219 64 1 Chicken patty fillet or tenders breaded cooked 27 9488 1 Tomato sauce 6 34904 1 Margarine like spread tub salted N 3 6 00PM 4 2 1269 76 1 Milk dry whole not reconstituted
75. 64 55 6 77 0 77 WHR 0 67 1 06 0 84 0 09 Body fat percent 48 447 26 3 9 30 Post study Post study Pre study Weight 103 0 285 4 171 0 41 8 2 06 0 54 BMI 17 3 41 5 263 55 2 13 0 1 WHR 0 68 1 11 0 84 0 09 0 34 0 00 Body fat percent 7 1 443 264 092 0 97 0 10 Note p lt 0 05 All r test df 81 152 Usability Questionnaire At the end of the study participants had the opportunity to provide feedback about their experience in the study specifically about their impressions of the ASA24 dietary recall program and the bite counter Table 3 27 shows the frequency of responses for questions about the ASA24 dietary recall program The majority of participants 67 5 reported completing the ASA24 for most foods and beverages they consumed In associated open ended responses participants with a favorable view of ASA24 described the interface as simple straight forward well organized user friendly and easy to follow They liked the comprehensive list of food choices the food categories the search feature the good layout the pictures of the foods being able to add forgotten foods at any time the prompting pathway of questions being able to see the meal breakdown and summary its presence on the Internet and being able to use a computer to complete it the instructions provided and the e mail reminders with links Participants described that the ASA24 became routine that it was easy to comple
76. 8 Second future research should investigate what type of dietary intake reporting is most accurate and acceptable for participants in bite counter studies It may be that real time recording of intake with a mobile Internet capable device would be a better approach A study comparing participant perceptions of their reporting accuracy and usability of different dietary intake tools while simultaneously recording meals with the bite counter could inform future bite counter validation studies The tool selected should also have an accurate kilocalorie database be a validated measure of energy intake and provide data in a way that can be managed by researchers It may be that the ASA24 will be the best tool available considering all of these factors especially as improvements are made to ASA24 over time but further exploration is necessary Third bite counter training and feedback could be provided to participants in order to improve the quality of the bite counter recordings It may be the case that participants should refrain from other activities while eating in order to reduce the occurrence of false positives Perhaps participants should be able to see when bites are being recorded on the device in real time so that they can adjust their behavior to make sure that bites are being recorded during meals This training and feedback could take a number of forms from a small manual provided with the bite counter at the beginning of the study to vide
77. 998 has also described the process of behavioral self regulation in similar terms Self regulation begins with self observation that can vary in its informativeness regularity temporal proximity and accuracy Then a judgment process allows the individual to compare what he or she has learned from self monitoring to his or her own standards standard norms and social standards The individual will also judge the monitored activity as important to them not important or relatively neutral and determine if their performance is the result of their own actions or the actions or assistance of others Finally during a self reaction phase an individual evaluates performance positively or negatively and provides a tangible reward or punishment According to Bandura 1998 successful self regulation depends on successful self monitoring because it is the self monitoring process that provides the information 13 necessary for an individual to set goals and to evaluate his or her progress toward those goals Baumeister et al 1994 applied this idea to the self regulation of eating behavior when they stated that the first key to successful self regulation of eating is to self monitor food intake p 180 Self monitoring can also be described from the perspective of behavior therapy Clark et al 2010 The goal of behavior therapy in weight loss is to develop healthy eating and exercise habits that will allow individuals to reach their weight
78. Cheese ci Bread pit Milk cow s fluid 196 fat 13 11 2011 10 7 7 57 25 2 44 Missing survey data 1 17 11 2011 10 8 12 32 49PM 3 01 Justadrin 138 348 Milk cow s fluid 1 fat 29 11 2011 10 11 6 26 52PM 5 02 Snack 94 64 Apple raw 32 12 2011 10 12 3 53 41PM 10 22 Snack 77 48 Apple raw 1 13 2011 10 4 10 06 00 AM 2 49 Breakfast 107 2014 Cheese s Bread pita wheat or cracked wheat 5 13 2011 10 5 10 11 35 AM 3 12 Snack 58 5 Peach raw 9 13 2011 10 5 10 35 28 PM 4 58 Snack 5 74 Carrots raw 6 16 2011 10 5 12 19 54 PM 3 31 Snack 121 365 Pudding canned chocolate fat free Figure 2 12 Example of screening for a low bite count error with data sorted by bite count Possible ASA24 program database and reporting errors were identified by screening the data file for abnormal values For example when participant BiteCD014 s data was sorted by total meal kcal a snack of 37 kcal of Ovaltine amp powder was found with an associated bite count of 20 and duration of 5 minutes 33 seconds This meal number 8 can be seen in Figure 2 13 Typically Ovaltine powder would be reconstituted with a liquid such as milk but no reconstituting liquid was reported This was judged to be an error in either participant reporting or the ASA24 program and the meal kcal data was removed from the data set 87 8 11 63 13 34 333 362 112 174 368 20 26 7 7 20 5 33 6 02 1 52 2 54 6 08 Snack Snack Snack Snack Snack 9 50 PM 37 344 10 5
79. Combined with the possible errors from the bite counter recordings it can be assumed that the average within participants correlation when outliers were removed of 0 51 is just a starting point As improvements are made to the bite counter and the ASA24 or other dietary intake recording tools it is possible that error in bite and kilocalorie recordings could be reduced thus potentially improving the correlation between these two variables Lack of Bite Counter Training and Feedback Another limitation of this study was that participants did not receive bite counter training They were simply told how to use the device to record their meals Participants were encouraged to eat as they normally would which could have included engaging in other activities while eating and use of the non dominant hand Participants did not receive feedback from the device other than an on message and beeping when it was turned on and off so they did not develop an understanding of when the device was recording bites and when it was not This could have resulted in participants using the device in a way that would differ from someone who knows how the device works and what is being detected Perhaps more knowledgeable participants that are given meaningful device feedback would use the device correctly and the correlation between bites and kilocalories could possibly improve 187 Study Sample The majority of participants in the study were students o
80. G Kopelman I D Caterson amp W H Dietz Eds Clinical obesity in adults and children 33 ed pp 351 365 West Sussex UK Wiley Blackwell 253 Enders C K amp Tofighi D 2007 Centering predictor variables in cross sectional multilevel models A new look at an old issue Psychological Methods 12 2 121 138 Erdogan B amp Bauer T N 2010 Differentiated leader member exchanges The buffering role of justice climate Journal of Applied Psychology 95 6 1104 1120 Feunekes G I J de Graaf C amp van Staveren W A 1995 Social facilitation of food intake is mediated by meal duration Physiology amp Behavior 58 3 551 558 Flegal K M Carroll M D Ogden C L amp Curtin L R 2010 Prevalence and trends in obesity among US adults 1999 2008 Journal of the American Medical Association 303 235 241 doi 10 1001 jama 2009 2014 Flegal K M Graubard B I Williamson D F amp Gail M H 2007 Cause specific excess deaths associated with underweight overweight and obesity Journal of the American Medical Association 295 2028 2037 doi 10 1001 jama 298 17 2028 Finkelstein E A Trogdon J G Cohen J W amp Dietz W 2009 Annual medical spending attributable to obesity Payer and service specific estimates Health Affairs 28 w822 w831 doi 10 1377 hlthaff 28 5 w822 Fishel Brown S R 2010 The relationship between energy balance understanding and me
81. H Masters T M 2012 The influence of bite size on quantity of food consumed A field study Journal of Consumer Research 38 5 791 795 doi 10 1086 660838 Moshfegh A J Rhodes D G Baer D J Murayi T Clemens J C Rumpler W V Cleveland L E 2008 The US Department of Agriculture Automated Multiple Pass Method reduces bias in the collection of energy intakes American Journal of Clinical Nutrition 88 324 332 230 M ller M J Bosy Westphal A amp Krawczak M 2010 Genetic studies of common types of obesity A critique of the current use of phenotypes Obesity Reviews 11 612 618 National Cancer Institute 2011 Automated Self administered 24 hour Dietary Recall ASA24TM Retrieved from http riskfactor cancer gov tools instruments asa24 National Weight Control Registry 2011 The National Weight Control Registry Retrieved from http www nwcr ws O Connor D B Jones F Conner M McMillan B amp Ferguson E 2008 Effects of daily hassles and eating style on eating behavior Health Psychology 27 1 S20 S31 Paeratakul S Ferdinand D P Champagne C M Ryan D H amp Bray G A 2003 Fast food consumption among US adults and children Dietary and nutrient intake profile Journal of the American Dietetic Association 103 1332 1338 Palmer M A Capra S amp Baines S K 2009 Association between eating frequency weight gain and health Nutrition Re
82. In A Jacobs amp L B Sachs Eds The psychology of private events pp 39 59 New York Academic Press Kanfer F H amp Gaelick L 1986 Self management methods In F H Kanfer amp A P Goldstein Helping people change A textbook of methods 3 ed pp 283 345 New York Pergamon Press Kant A K amp Graubard B I 2004 Eating out in America 1987 2000 Trends and nutritional correlates Preventive Medicine 38 243 249 257 Kayman S Bruvold W amp Stern J S 1990 Maintenance and relapse after weight loss in women Behavioral aspects American Journal of Clinical Nutrition 52 800 807 Kirk S F L Penney T L amp McHugh T L F 2010 Characterizing the obesogenic environment The state of the evidence with directions for future research Obesity Reviews 11 109 117 doi 10 1111 j 1467 789X 2009 0061 1 x Kirkpatrick S 2011 Accounting for measurement error in dietary intake data Measurement Error Webinar Series http riskfactor cancer gov measurementerror Klem M L Wing R R McGuire M T Seagle H M amp Hill J O 1997 A descriptive study of individuals successful at long term maintenance of weight loss American Journal of Clinical Nutrition 66 239 246 Kruger J Blanck H M amp Gillespie C 2006 Dietary and physical activity behaviors among adults successful at weight loss maintenance nternational Journal of Behavioral Nutrition and Physica
83. N i d y i m D a VER m hs A Pd N I H i y y i Figure 1 3 Positive wrist roll when taking a bite Initial research with the bite counter was completed with a tethered sensor an InertiaCube3 InterSense Inc Bedford MA with an attached athletic wrist band see Figure 1 4 To test the bite counter concept a controlled study focused solely on a single food Scisco 2009 Fifty one participants were presented with 870 kcal 276 grams of Kellogg s Eggo cinnamon toast waffles and allowed to eat as much as they liked using a fork The waffles were pre cut into uniform bite size pieces The 21 participant was seated at a table and the bite counter was placed on the wrist of the dominant hand and connected to an external computer A video camera was positioned to record the person while eating The computer recorded the raw sensor data and the times at which bites were detected The raw sensor data and bite detection times were correlated with the recorded video in order to evaluate the performance of the device The participants ate a range of 8 to 95 bites 34 bites on average The sensitivity of the device was 94 and only 6 of the actual bites were undetected The positive predictive value was 80 While the conditions in this test were restrictive in terms of food type eaten and utensil used it showed that our technique works across a large number of participants Figure 1 4 The tethere
84. Rate aen C NUM RO Pe 92 Additional Two Level Model sd Ret see i eun eR RNV dn 121 Additional Model with Outlier Participants Removed 126 Bite Size MOGel ideni qase ab ete tieu kae a AD M ORT e SER e 141 Lab Meal eiue ge obit deed eie Md deine tete 148 Table of Contents Continued Body Measurements oi ecr p aess is oe evite ce Soo o ease evo e das eso 152 Usability Questionnaire usc descen Eod ege seus eo DR mM EE 153 IV DISCUSSION 2 stis SED DNI aite qe NND Sd f ED 160 Sources of Variance in Bite Count sese 160 oa Kora Meee ie re mm qoe dire peut E ta E Dus ede 177 bugP 178 Implications of ASA24 and Bite Counter Usability 180 Study Strengths ibendum andes tas ear ieee anise eae 183 Study Limitat ons 5o mt v ar eheu er tou sence east A ARS 186 Future Research DIFectofis ences iso epo ere aco V Pene ees 188 The Future Bite Counter uii eni peteret pred teet need 191 COHCIUSIOIE eid oerte Sze cide octal E adest ge ee Plau 192 APPENDICES life teste nmco wisst uisi done V ata ous e afta esp Uo Paride 194 A Demographics QUeSUOBDAITG ecole eode tene Hb HE THU pee RES 195 B Three Factor Eating Questionnaire R 18 TFEQ R 18 200 C Daily Meals Questionnaire ene essaie te enar epe e R ceensncssatesedaneteasis 202 D gt Usability OuestiOtnitialfe e SG eke iaa er e a Baa eed e RAS 208 E Initial Participant Co
85. SOURCES OF VARIANCE IN BITE COUNT A Dissertation Presented to the Graduate School of Clemson University In Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy Human Factors Psychology by Jenna Lori Scisco May 2012 Accepted by Dr Eric R Muth Committee Chair Dr Thomas R Alley Dr Adam W Hoover Dr Patrick J Rosopa ABSTRACT The obesity epidemic affects millions of individuals worldwide New tools that simplify efforts to self monitor energy intake may enable successful weight loss and weight maintenance The purpose of this study was to examine predictors of the number of bites recorded by the bite counter device during daily meals in natural real world settings Participants N 83 used bite counters to record daily meals for two weeks Participants also recorded their daily dietary intake using automated computer based 24 hour recalls Predictors of bite count were explored at the meal level and individual level using multilevel linear modeling A positive relationship between kilocalories and bites was moderated by energy density such that participants took more bites to consume greater kilocalorie meals when energy density was low than when energy density was high The positive relationship between kilocalories and bites was also moderated by participants average bite size during a laboratory meal such that participants with smaller bite sizes took more bites to consume greater kilocalorie meal
86. South Asian Open ended responses included eating local organic and whole foods limiting eating out refined sugars starches fats fried foods carbs junk food sodium snacking eating healthier diets including Weight Watchers Type I diabetes Type B blood type figure competitor yogi and macrobiotic counting calories eating smaller meals and using smaller plates following vegetarian practices including lacto lacto ovo and pescetarian increasing fiber fruits vegetables lean protein seafood and eating complex carbs fats and protein in every meal 65 Materials Bite Counters Bite counters were 1400 through 1700 series devices from Bite Technologies see Figure 1 6 Each device series used the same equipment and design with improvements made over time to increase the daily battery life The device was a 2 5 x 1 5 inch 64 x 38 mm plastic rectangle that was 1 inch 25 mm thick and weighed 2 7 oz 75 grams A 1 inch 25 mm wide 6 5 8 5 inches 165 216 mm long wrist band was attached to the device The battery in the device ideally allowed for 14 hours of bite counting use per charge approximately 2 weeks of regular use It took 3 hours to fully recharge the battery The bite counter stored data for up to 320 eating sessions A USB connection was used for downloading data and recharging These bite counters operated as a typical watch when not in use as a bite counter Prior to each eating session the user p
87. a BMI of 25 or greater defines overweight Obesity is associated with increased rates of type 2 diabetes mellitus hypertension dyslipidaemia heart disease cerebrovascular disease respiratory disease osteoarthritis kidney disease and cancer Malnic amp Knobler 2006 Obesity is also associated with greater mortality from cardiovascular disease some cancers diabetes and kidney disease Flegal Graubard Williamson amp Gail 2007 The increased prevalence of health problems in the obese population naturally leads to increased health care costs In the United States the medical costs of obesity were estimated to be 147 billion per year in 2008 doubling from 78 5 billion in 1998 Finkelstein Trogdon Cohen amp Dietz 2009 Additionally obese and overweight employees are estimated to cost their employers 641 and 201 respectively more per employee per year due to increased doctor visits emergency room visits and productivity losses Goetzel et al 2010 It is imperative that obesity rates be reduced to improve the health of these individuals and to decrease associated health care costs Stated simply obesity is the result of an energy imbalance in the body Sharma amp Padwal 2010 The energy consumed in the form of food and drink is greater than the amount of energy expended through physical activity and basal metabolism and this tipping of the energy scales toward excess intake results in weight gain Dulloo 2010 While
88. a summary xls Copy the data from the Merged Data sheet in ParticipantID xls and paste into ParticipantID data summary xls Delete rows so that MealID Bites Year M D Duration Meal or snack Meal time MealKCAL and Food Description remain Create a new column names calories per bite Calculate for each meal with matching data as MealKCAL Bites Calculate the average number of bites calories and calories bite for each column Highlight each average at the bottom of the respective columns for the participant to see easily Email ParticipantID data summary xls to the participant as an attachment with the following message Dear first name Attached please find your data summary from the Bite Counter study This spreadsheet contains all of the meals for which Bite Counter data and or ASA24 data were recorded Each row is a meal Your average number of bites per meal calories per meal and calories per bite are highlighted at the bottom of the spreadsheet Thank you for your participation Jenna Scisco Department of Psychology Clemson University 240 Appendix P Description of Data Quality for Each Participant Bites Bite Counter ASA24 ASA24 ID matched matched Kilocalories problems data problems data completed meals meals correlation quality quality First bite counter had time drift and display problems Second One meal was BiteCD001 20 60 6 171 bite counter turned 11 overestimated
89. able to meet from 5 days 216 here and I will select a time for this meeting There is currently a waiting list for this study and a prompt reply is appreciated Sincerely Jenna Scisco Department of Psychology Clemson University 864 656 1144 7 a When the participant responds send the e mail described in 4 a When a participant has been scheduled add their session to the lab calendar as ParticipantID orientation and reserve their bite counter on the Bite Counter Status white board Record the date and time of the orientation in the ParticipantIDinfo spreadsheet 217 Appendix F Orientation Protocol 1 day before participant arrives 1 Bite Counter preparation a b e Record the participant s bite counter number on ParticipantIDinfo spreadsheet Connect the device to the bite counter software i Download and save all previous data Clear the data from the device ii Sync the time with the computer time ili Verify that the display settings are set to on with no review of calories bites or charge iv Disconnect the device Confirm the on setting and no review of calories bites or charge Run the device Diagnostics You do this by holding the device steady pressing and holding the right button down and pressing the left button and then releasing both buttons The first diagnostic is a Display Test During this test you should see the entire display activated Following
90. accuracy of the kilocalorie estimations or the user could skip this energy density input step knowing that its effect on day level or greater kilocalorie sums will not be as great as it averages out over time Meal Duration Research question 4 investigated if meal duration could predict bite count During data exploration Meal Duration was identified as a variable with an almost perfect correlation with Bites This indicated that Meal Duration and Bites were representing the same construct The longer the device was on the more bites either true 164 detections or false positives were recorded by the device There are two practical implications of this finding First for over half of the meals participants were engaged in other activities while eating and some of these activities could involve the use of the hands such as using a computer Thus while the device is on it could be detecting these activities false positives in addition to true bites which may explain why there was such a strong correlation between Meal Duration and Bites Second Meal Duration itself could potentially be used as an outcome variable from the bite counter It is possible that the detection of Bites could be used as one indicator of eating activity which might enable automatic detection of eating behavior by the device Dong Hoover Scisco amp Muth under review Then Meal Duration could be used by an individual as part of an eating diary which might
91. ake regulation Neuroscience and Biobehavioral Reviews 26 581 595 de Lauzon B Romon M Deschamps V Lafay L Borys J Karlsson J the Fleubaix Laventie Ville Sante FLVS Study Group 2004 The three factor eating questionnaire R18 is able to distinguish among different eating patterns in a general population Journal of Nutrition 134 2372 2380 Diaz V A Mainous A G amp Everett C J 2005 The association between weight fluctuation and mortality Results from a population based cohort study Journal of Community Health 30 3 153 165 doi 10 1007 s10900 004 1955 1 Dixon M A amp Cunningham G B 2006 Data aggregation in multilevel analysis A review of conceptual and statistical issues Measurement in Physical Education and Exercise Science 10 2 85 107 Dong Y Hoover A W Scisco J L amp Muth E R 2012 A new method for measuring meal intake in humans via automated wrist motion tracking Applied Psychophysiology and Biofeedback Dong Y Hoover A Scisco J amp Muth E under review Detecting the eating activities of a free living human by tracking wrist motion Dulloo A G 2010 Energy balance and body weight homeostasis In P G Kopelman I D Caterson amp W H Dietz Eds Clinical obesity in adults and children 33 ed pp 67 81 West Sussex UK Wiley Black well Elfhag K amp R ssner S 2010 Weight loss maintenance and weight cycling In P
92. als with both bite counter and ASA24 data for each participant ranged from 15 to 100 M 39 SD 15 3 190 meals had complete data from all three sources 78 5 The number of meals with data from all three sources for each participant ranged from 13 to 99 M 38 SD 15 These frequencies and additional features of these meals are 97 described in Table 3 1 Participants engaged in other activities for at least 68 of their reported meals Talking using a computer and watching TV were the most common activities engaged in while eating Participants ate most often with their hands a fork or a spoon Table 3 1 Frequencies and percentages of participant meal reporting and meal features Meals N of analysis data set All meals 4065 100 Bite counter data 3606 87 7 Daily meals questionnaire DMQ 3794 93 3 ASA24 data 3691 90 8 Bite counter and ASA24 data 3246 79 9 Bite counter ASA24 and DMQ data 3190 78 5 Engaged in other activities during the meal 2712 68 2 Talking conversation 1012 24 9 Using a computer 758 18 6 Watching TV movie 719 17 7 Reading 176 4 3 Driving 141 3 5 Cooking food preparation 31 0 8 Feeding a child or pet 23 0 6 Using phone to talk or text 25 0 6 Utensil used Hands 2354 57 9 Fork 1221 30 0 Spoon 885 21 8 Knife 412 10 1 Chopsticks 29 0 7 Straw 17 0 4 Toothpick 2 0 05 98 Descriptive statistics for main Level and Level 2 analysis variables are presented in Table 3 2 ICC1 repre
93. ammarstrand Hemmingsson Neovius amp Johansson 2008 Weight maintenance is another challenge presently facing behavioral modification programs A recent review of the weight maintenance literature for lifestyle modifications indicated that only half of the individuals who lost weight using this approach maintained the weight loss a year or more after supervision ceased Barte et al 2010 In order to improve weight maintenance success and reduce obesity rates over the long term the behaviors of individuals who have successfully lost weight and maintained the weight loss weight maintenance experts can be studied and described Effective behaviors that are common across these weight maintenance experts can be extended to the development of weight loss and maintenance programs Successful Weight Loss and Weight Maintenance Successful weight loss and weight maintenance do not have standard definitions in the literature Generally weight loss is defined as losing a percentage of one s body weight and weight maintenance is defined as maintaining that weight loss for a period of time Specific definitions from the literature are provided in Table 1 1 Obesity research focuses on intentional weight loss as opposed to unintentional weight loss resulting from disease or negative health behaviors McGuire Wing Klem amp Hill 1999 Individuals may experience periods of weight fluctuation with repeated attempts to lose weight followed by
94. an usual but others described these meals a larger than usual Participants became more aware of when they ate how often they ate and more aware of mealtime vs not mealtime Some participants described becoming more aware of how fast they ate and one participant described paying more attention to when they became full One participant would not eat at night after the device had been plugged in to charge Some participants described eating more often with their dominant hand trying not to move their dominant hand around too much for other activities while eating and noticing that they sometimes ate with their non dominant hand When asked which tool they preferred the majority of participants 75 9 reported preferring the bite counter because it took less time was easier and simpler and because it was new and different For those who preferred using the ASA24 dietary recall they preferred this tool because it allowed them to receive feedback about what foods they were eating and how much they were eating 159 CHAPTER FOUR DISCUSSION The purpose of this study was to identify sources of variance in bite count during meals from people using the bite counter during their daily lives In the discussion that follows the results for each variable of interest are summarized Practical implications for the bite counter study strengths limitations and future research directions are discussed Sources of Variance in Bite Cou
95. andom Social Bites slope variance significantly improved the model fit However the random Social Bites slope variance 28 16 did not significantly differ by participant Due to the small increase in model fit but non significant slope variation the random Social Bites slope variance was dropped from subsequent models Model 15 Does the Relationship between Intake Day and Bites vary by participant In model 15 the relationship between Intake Day and Bites was allowed to vary by participant random Intake Day Bites slope variance The y deviance difference test comparing model 15 to model 11 28312 93 28312 81 0 12 df 13 12 1 p gt 05 indicated that the addition of the random Intake Day Bites slope variance did not significantly improve the model fit In addition the random Intake Day Bites slope variance 0 45 did not significantly differ by participant Because the random Intake Day Bites slope variance did not improve model fit and did not vary by participant it was dropped from subsequent models 113 Model 16 Can the varying Kilocalorie Bite slopes by explained by Gender Because the relationship between Kilocalories and Bites varied significantly by participant a cross level interaction between Kilocalories and Gender was added to the model to examine if Gender could explain some of this random slope variance The a deviance difference test comparing model 16 to model 11 28312 93 28309 40 3 53 df 13 12 1 p
96. articipant x BIeCRZIY gt Ap 203 calibration found study Many zero bite overwhelming recordings 246 Bites Bite Counter 4 ASA24 ASA24 ID matched matched Kilocalories problems data problems data completed meals meals correlation quality quality BiteCD215 36 94 7 254 Bite co nter famed ud Good off frequently Bite counter turned us BiteCD216 41 89 1 543 off during a few 12 to Internet meals access Bite counter turned Over estimated BiteCD217 46 75 4 617 off during a few 12 two meals meals corrected Bite counter turned BiteCD218 31 70 5 591 off during a few 13 Good meals First device had BiteCD219 39 70 9 576 eplay prone 14 Good Second device was good Many missing recalls and did BiteCD222 20 57 1 099 Good 9 T S understand purpose of the study First device turned One over BiteCD224 22 64 7 338 Ole queat 12 eoumo Second device was meal good corrected First device turned BiteCD227 69 93 2 492 ott during afew 14 Good meals Second device was good Bite counter turned BiteCD231 42 91 3 529 off during a few 13 Good meals Bite counter turned BiteCD232 59 92 2 767 off during a few 14 Good meals 247 Bites Bite Counter 4 ASA24 ASA24 ID matched matched Kilocalories problems data problems data completed meals meals correlation quality quality BiteCD237 21 41 2 578 Good 11 Good BiteCD240 36 87 8 584 Good 14 Good Bite counter turned BiteCD241 25 78 1 5
97. articipant instructions Final meeting and meal The protocol for the final meeting and meal are described in detail in Appendix L After 14 days of data collection the participant returned to the laboratory to return the bite counter and complete the Usability Questionnaire on Survey Monkey Weight body fat percentage waist circumference and hip circumference were measured again In addition the participant ate a meal in the laboratory in order to measure average bite size The participant ate Amy s brand macaroni and cheese This meal was selected because it is easy to prepare in the laboratory is acceptable for either lunch or dinner and is amorphous and thus can be eaten in different sized bites Amy s brand received the highest taste ratings when compared to nine other commercially available 80 macaroni and cheese varieties by three research assistants A soy cheese variety was available for vegans and a rice pasta variety was available for those allergic to gluten The participant was seated at the laboratory eating station set with a fork napkin plate macaroni and cheese on top of the plate in its original container and a glass of 500 mL of water An Ohaus Scout Pro Balance SP4001 Ohaus Corp Pine Brook NJ with an RS232 interface was concealed under a tablecloth and sampled the weight of the meal every three seconds Data was collected using TAL WinWedge RS232 data acquisition software TAL Technologies Inc Philadelphia
98. asures of wellness Doctoral dissertation Retrieved from ProQuest Dissertations and Theses database UMI No 3402719 Fisher E B Lichtenstein E Haire Joshu D Morgan G D amp Rehberg H R 1993 Methods successes and failures of smoking cessation programs Annual Review of Medicine 44 481 513 Fisher J O Rolls B J amp Birch L L 2003 Children s bite size and intake of an entr e are greater with large portions than with age appropriate or self selected portions American Journal of Clinical Nutrition 77 1164 1170 French S A Story M amp Jeffery R W 2001 Environmental influences on eating and physical activity Annual Review of Public Health 22 309 335 Fry J P amp Neff R A 2009 Periodic prompts and reminders in health promotion and health behavior interventions Journal of Medical Internet Research 11 2 e16 254 Fujimoto K Sakata T Etou H Fukagawa K Ookuma K Terada K amp Kurata K 1992 Charting of daily weight pattern reinforces maintenance of weight reduction in moderately obese patients The American Journal of the Medical Sciences 303 3 145 150 Fulton J E Dai S Steffen L M Grunbaum J A Shah S M amp Labarthe D R 2009 Physical activity energy intake sedentary behavior and adiposity in youth American Journal of Preventive Medicine 37 1 S40 S49 doi 10 1016 j amepre 2009 04 010 Gibson A L Heyward V
99. at men eat more kilocalories than women during a single meal with the degree of difference varying across studies For example in three different studies Grunberg amp Straub 1992 Pliner et al 2006 Rolls Morris amp Roe 2002 men have been found to eat 30 70 more kilocalories than women This gender difference has also been found in humans in their natural eating environments An analysis of a decade of diet diary research has indicated that about 16 of the variance in daily energy intake is due to the gender of the individual de Castro 1996 If bites and kilocalories are strongly correlated one may predict that males will have higher 56 bite counts than females However if males take larger bites than women in order to consume more food Burger et al 2011 it is possible that a reverse gender effect could be found for bite count Research Question 9 Does gender predict bite count Weight Individuals with larger body weights require more kilocalories to maintain their body weight McArdle et al 2005 Body weight has been found to be more strongly correlated with energy intake than BMI Periwal amp Chow 2006 This is because two people can have the same BMI but different heights and weights For example Jane is 5 3 and weighs 200 pounds her BMI is 35 4 Greg is 5 9 and weighs 240 pounds his BMI is 35 4 However Greg is a much larger individual and thus requires more kilocalories at each meal If bi
100. at they could slide their hand through the band without having to separate the two ends As described above friends and coworkers often asked about the device but some participants disliked describing their weird looking watch to others When asked how the device could be improved participants suggested a smaller device with a curved back a thinner non Velcro wristband optional beeping less frequent charging different colors additional watch features like the date and a stop watch syncing to devices like the iPhone water resistance impact resistance and automatic detection of eating For the 43 4 of participants who experienced bite counter problems the main problem with the bite counter was that it would sometimes shut off during meals and would need to be turned back on A few participants thought that 18 88 was an error message although this indicated that the device was calibrating Some participants thought they had to hold down the button to pass through the 18 88 message which resulted in difficulty getting the device to stay on Finally the devices did not 158 automatically adjust for daylight savings time which was a small inconvenience until the experimenter could adjust the time for them For the 51 8 of participants who described changing their eating behavior as a result of using the bite counter many participants described snacking less and eating fewer meals Sometimes these meals were described as smaller th
101. became significant r 0 05 p lt 05 and indicated that females took fewer bites than males A positive correlation emerged between Body Weight and Bites r 0 05 p 05 which indicated that people with heavier body weights took more bites The correlation with Bites was similar for BMI r 0 04 p 05 which reflected the overall near perfect correlation between Body Weight and BMI r 0 92 p lt 0 05 128 Table 3 13 Total correlations between level 1 and level 2 variables for the outliers removed model Variable 1 2 3 4 5 6 7 8 9 1 Bites 2 Kilocalories 0 46 3 Energy Density 0 14 0 06 4 Location 0 07 0 08 0 07 5 Social 0 27 0 31 0 02 0 12 6 Intake Day 0 03 0 06 0 01 0 15 0 17 7 Gender 0 05 0 29 0 01 0 06 0 04 0 01 8 Body weight 0 05 0 22 0 05 0 07 0 06 0 02 0 45 9 BMI 0 04 0 13 0 08 0 06 0 04 0 03 0 19 0 92 10 Height 0 01 0 28 0 05 0 07 0 07 0 01 0 72 0 54 0 17 Note p 0 05 Location coded 0 Home 1 Not at Home Social coded 0 Alone 1 With Others Intake Day coded 0 Weekday 1 Weekend Gender coded 0 Male 1 Female Bite size calculated as kilocalories per bite during the lab meal 129 All predictors were centered at the grand mean and model building was conducted on this outliers removed sample in the same manner as described for the previous model building with all parti
102. between kilocalories and bites was used The kilocalorie bite relationship is the most theoretically meaningful for the bite counter project In order for the bite counter to be understood and well accepted by the weight loss community as a measure of energy intake it should provide a reasonable estimate of the number of kilocalories consumed Therefore at minimum the current analysis should be appropriately powered to detect this effect Preliminary analyses from free living humans in our research group suggest a correlation of about 0 7 between kilocalories and bites We can assume that this correlation will decrease with a larger sample size as more variance is introduced but we still expect this effect to be large Following Cohen s guidelines a large effect size is 0 5 Cohen et al 2003 The necessary sample size to detect the relationship between two variables with an expected effect size of 0 5 with an alpha level of 0 5 and a power level of 0 80 is 28 Cohen 1992 Therefore collecting data from at least 28 meals per participant is sufficient for detecting the expected relationship between kilocalories and bites A final approach to confirming that the sample size selected for the current study is appropriate is to examine articles that have used MLM analyses with similar numbers of variables entered into the model If the sample sizes are comparable or smaller than the proposed sample sizes and the model was able to converge then our
103. bove example when an individual is eating watermelon for breakfast she may take 60 bites to eat 500 kilocalories When that same individual is eating sausage for breakfast she may only take 20 bites to eat 500 kilocalories That is it takes fewer bites to eat the same number of kilocalories when the energy density of the food is high indicating that the relationship between bites and kilocalories is not as strong for high energy density foods compared to low energy density foods This hypothetical relationship is shown in Figure 1 10 As can be seen in Figure 1 13 the slope of the line for high energy density foods is less steep because it takes fewer bites to eat more 50 kilocalories compared to low energy density foods The slope of the line for low energy density foods is steeper because it takes more bites to eat more kilocalories compared to high energy density foods Low High Linear Low Linear High Bites 0 j T T T 1 0 200 400 600 800 Kilocalories Figure 1 13 Hypothetical interaction between kilocalories and energy density Conversely it is possible that an interaction with the opposite pattern could emerge if the individual takes more bites of an energy dense food and fewer bites of a less energy dense food Then the relationship between bites and kilocalories would be weaker for high energy density foods compared to low energy density foods Research Question 3 Does the
104. by a number of studies that manipulate the number of 54 people present at a meal Redd amp de Castro 1992 For example when children ate a snack in groups of nine they consumed 30 more food than when they ate in groups of three Lumeng amp Hillman 2007 In another study adults eating with friends ate 18 more than when they ate alone Hetherington Anderson Norton amp Newson 2006 These experimental studies provide more support for a link between the number of people present at a meal and the amount of food consumed Redd amp de Castro 1992 Following from this social facilitation literature one can assume that eating with more people will result in higher bite counts if bite counts reflect increase energy intake Additionally eating with other people involves more talking and gesturing which may trigger additional bite recordings by the bite counter device Research Question 6 Does the number of people an individual eats with predict the number of bites recorded during a meal Day of the week The day of the week a meal is eaten on is a cultural influence that may impact the amount of food consumed Weekdays are typically devoted to routine work activities that constrain eating behavior whereas weekends are reserved for leisure activities or celebrations that are associated with more food intake e g birthday parties picnics social gatherings Basiotis et al 1989 de Castro 1991 Daily diary studies have sh
105. chedule BiteCD129 27 87 1 285 Device minted oil 13 Good once BiteCD132 36 97 3 507 Time drift 14 Good BiteCD138 31 73 8 721 Display problems 13 Good Difficulty BiteCD148 18 52 9 666 remembering 13 Good to wear and use Good Reported One over BiteCDIS1 33 71 7 081 cecal 14 counted remembering to turn meal on and off corrected BiteCD152 39 73 6 ATS Fost mea y conne Sj Good by participant 245 Bites Bite Counter 4 ASA24 ASA24 ID matched matched Kilocalories problems data problems data completed meals meals correlation quality quality BiteCD153 27 57 4 A71 Good 12 Good BiteCD170 26 40 0 548 Good 14 Good First bite counter One over BiteCD175 49 94 2 553 turned off frequently 14 estimated Second bite counter meal had no problems corrected Bite counter turned off during a few BiteCD178 30 81 1 136 meals Not reported 12 Good by participant but seen in data Bite counter turned BiteCD196 43 82 7 636 off during a few 14 Good meals Bite counter turned BiteCD197 60 88 2 749 off during a few 14 Good meals Bite counter turned BiteCD208 49 98 0 626 off during a few 14 Good meals BiteCD210 54 98 2 631 BiecCoumenimied vua Good off once Bite counter turned off during a few BiteCD211 36 72 0 454 meals 14 Good Display problems Time drift BiteCD213 36 92 3 i988 CPU COUR ed iy Good off frequently Participant tried to hold down the button Good to get past P
106. cipants Results from Models 1 through 16 are presented in Tables 3 14 and 3 15 Exploratory models 17 through 19 are compared to model 11 in Tables 3 16 and 3 17 The results of model 1 indicated that there was still significant nesting ICC1 0 22 with 22 of the variance in bites occurring between participants Models 2 through 5 7 through 8 and 10 through 16 were in line with full sample findings Examining the unique effect of each level 1 predictor for explaining within participants variance in bites it was found that Kilocalories explained 28 4 Energy Density explained 2 7 Kilocalories x Energy Density explained 1 4 Location explained 0 3 and Social explained 2 1 However in Model 6 the significant effect of Intake Day found in the full sample was non significant for the outliers removed sample did not improve model fit and explained 046 of the within participants variance This indicated that the number of Bites taken during meals did not differ between Weekends and Weekdays Therefore Intake Day was dropped from further models Additionally in Model 9 the significant effect of Gender found in the full sample was non significant for the outliers removed sample did not improve model fit and explained 0 of between participants variance This indicated that the number of Bites taken during meals did not differ between males and females Thus Gender was dropped from future models Results of exploratory models 17 and 18 were i
107. cise behaviors Additionally some researchers can choose to address the obesity problem from an individual behavior modification perspective These researchers can work to provide individuals with strategies and tools to resist the many forces in the environment that promote weight gain Hill Wyatt Reed amp Peters 2003 p 854 Many lifestyle change programs have been developed to help people increase their physical activity reduce their energy intake and ultimately lose weight Often this behavioral modification results in modest weight loss success For example Goodpaster et al 2010 reported that a one year lifestyle modification program for the severely obese that included reducing energy intake with a prescribed diet and increasing activity to 60 minutes of walking 5 days per week resulted in 30 of participants achieving at least a 10 weight loss As another example Rock et al 2010 examined the effectiveness of a commercial weight loss program for overweight and obese women Results indicated that a low fat reduced energy diet and 30 minutes of exercise on at least 5 days per week led to a one year weight loss of about 10 and a 2 year weight loss of about 7 In general these lifestyle modification programs are typically preferred over bariatric surgery and pharmacotherapy due to their fairly promising success rates much lower financial expense relative safety and wide availability to the general public R ssner H
108. clusion that more consistent self monitoring is related to greater weight loss is a recurring trend in the self monitoring of exercise and food intake literature Wadden et al 2005 However similar to the self monitoring of body weight literature the direction of the relationship between self monitoring physical activity and food intake and weight loss is unknown Self monitoring these behaviors may lead to weight loss or weight loss may encourage self monitoring practices 17 Once a relationship between more consistent self monitoring and weight loss was established researchers began to investigate the many factors that could improve adherence to a self monitoring protocol with the assumption that improved adherence would be related to increased weight loss After a thorough literature review a number of common factors that improve self monitoring were identified These are summarized in Table 1 2 Simplified diaries Internet technology PDAs PEDs and mobile phones SMS can be used as self monitoring tools that can increase self monitoring adherence Beasley 2007 Burke et al 2009 Burke et al 2011 Cushing Jensen amp Steele 2010 Helsel Jakicic amp Otto 2007 Micco et al 2007 Morgan Lubans Collins Warren amp Callister 2011 Patrick et al 2009 Tate Wing amp Winett 2001 Yon et al 2007 Counselor support and feedback accountability human counseling and reminders to self monitor are features of self m
109. d InertiaCube3 attached to an athletic wristband In a follow up study a much smaller and less expensive sensor was used the STMicroelectronics LPR530al as shown in Figure 1 5 Dong Hoover Scisco amp Muth 2012 Participants wore this smaller sensor and the InterCube3 in order to compare performance between sensors In this laboratory study with less control over the eating 22 situation 47 participants were recorded eating a meal that they brought with them to the study using the utensil s of their choice and given no particular instructions as to how to eat the meal The meals chosen ranged from noodles eaten with a spoon to chicken tenders and french fries eaten with fingers to a pasta dish eaten with a fork As with the controlled meal a video camera was positioned to record the person while eating and the bite counter was placed on the person s dominant wrist and connected to an external computer Data were also recorded and analyzed in the same manner as with the controlled meal The sensitivity of the STMicroelectronics device was found to be 86 with a positive predictive value 81 The sensitivity of the InertiaCube sensor was found to be 85 with a positive predictive value 81 The first non tethered ambulatory bite counters using the smaller sensor were developed by Bite Technologies and became available in summer 2011 Figure 1 6 Figure 1 5 The smaller MEMS sensor center compared to the InterSense IneritaCube
110. d also take into consideration the number of kilocalories being consumed at a meal 162 Although no prior research has investigated a relationship between energy density and bites previous research has investigated the relationship between energy density and kilocalories finding that people tend to consume more kilocalories when they eat more energy dense foods e g Bell et al 1998 de Castro 2004a In this case bites cannot be substituted for kilocalories when describing the relationship with energy density because more bites are associated with meals consisting of overall lower energy density Again this points to the importance of examining the kilocalorie by energy density interaction as discussed below Kilocalories by Energy Density Interaction Research question 3 investigated if the relationship between kilocalories and bites would depend on the average energy density of the foods being consumed For both the full sample and the outliers removed sample the kilocalories by energy density interaction explained about 1 5 of the variance in bites indicating that this effect was relatively small compared to the overall effect of kilocalories For both the full sample and the outliers removed sample the simple slopes revealed that when energy density was at its mean across all meals 1 18 kcals g low energy density about 25 kilocalories were consumed per bite When energy density was one standard deviation below its mean 0 18
111. d effect at level 2 and a BMI x Kilocalorie interaction term were added to create Model 18 The a deviance difference test comparing model 18 to model 11 28312 93 28312 04 0 89 df 14 12 2 p gt 05 indicated that the addition of the BMI fixed effect and the BMI x Kilocalorie interaction did not significantly improve model fit Random slope variance was not reduced 00041 117 00041 0 indicating that BMI did not explain any of the random Kilocalorie Bite slope variance Finally the BMI x Kilocalorie interaction term 0 0001 was non significant Therefore the BMI fixed effect and the BMI x Kilocalorie interaction term were dropped from further exploratory models Model 19 Can the varying Kilocalorie Bite slopes be explained by Height It was thought that Height could be another individual difference variable that could explain some of the random Kilocalorie Bite slope variance A Height fixed effect at level 2 and a Height x Kilocalorie interaction term were added to create Model 19 The a deviance difference test comparing model 19 to model 11 28312 93 28306 08 6 85 df 14 12 2 p lt 05 indicated that the addition of the Height fixed effect and the Height x Kilocalorie interaction significantly improved model fit Height explained 9 896 0 0004 1 0 00037 0 00041 100 of the random Kilocalories Bites slope variance The Height fixed effect 0 83 was non significant indicating no direct relationship between Heig
112. d not be reproduced in a sample that had higher quality bite counter and ASA24 data overall Practically future bite counter users seeking to reduce bite counts would not need to focus on whether intake occurs on a weekend or a 169 weekday and instead should focus on if they are eating with other people as this would indicate greater potential for taking more bites Social by Intake Day Interaction Research question 8 investigated if the relationship between eating with others and bites depend on whether it is a weekend or a weekday No significant interaction between social and intake day was found for any of the models with the interaction explaining close to 0 of the within participants variance This finding did not coincide with previous research that found greater social facilitation of eating on weekends compared to weekdays de Castro 1991 Practically this finding indicates that bite counter users should be cognizant of their bite count when eating with others every day of the week Gender Research question 9 investigated if gender could predict bite count Overall correlations between gender and bites were very small negative and only significant for the outliers removed sample This indicated that females may take fewer bites than males during meals Gender explained 5 2 of the between participants variance in full sample model but none of the between participants variance in the outliers removed model Slopes betwe
113. der for individuals j Hoj The deviation of the mean GPA intercept of an individual j from the overall mean GPA An error component for the Level 2 equations Y1o The overall regression coefficient for the relationship slope between job status and GPA ua The degree to which the relationship between job status and GPA depends on gender The cross level interaction term Hij The deviation of each individual j slope from the overall slope An error component for the Level 2 equations Combining the level one and level two equations through substitution results in equation 1 4 GP Yoo Yoa gender Hoj Yio Yi gender p jobstatus e 1 4 Rearranged this becomes the full model shown in equation 1 5 4 GPA Yoo Yo1 gender y ojobstatus y gender jobstatus p jobstatus Hoj ej 1 5 It can be seen that By j and Pa j have been dropped from the overall equation These coefficients are not fixed values because they vary by the individual j Thus they are called random effects MLM provides an estimate of the variance of each random effect Tabachnich and Fidell 2007 These two variances are described in Table 1 7 Table 1 7 Symbols and Meanings for the Random Variance Components Symbol Meaning Too The variance of the random means intercepts Tip The variance of the random slopes When an MLM analysis in conducted the three fixed coefficients in e
114. e 2002 Similarly when participants were offered two portions of pasta the larger portion size resulted in participants consuming 26 more kilocalories Burger et al 2011 However at home we have familiar environmental cues such as the consistent sizes of our plates and bowls that can help us to regulate our portion sizes and our subsequent food intake Sobal amp Wansink 2008 Increased energy intake outside of the home is also the result of increased energy density due to greater fat content in restaurant and fast food meals Paeratakul Ferdinand Champagne Ryan amp Bray 2003 In support of these relationships a daily diary study conducted in the US indicated that meals eaten in restaurants are 38 larger than meals eaten at home and 44 larger than meals eaten in other locations de Castro et al 1990 A 24 hour dietary recall study with children and adolescents in the US found that meals eaten at restaurants were 55 larger than meals eaten at home and meals eaten in restaurants contained significantly more calories from fat Zoumas Morse Rock Sobo amp Neuhouser 2001 Given increased energy intake at 53 locations outside of the home it is possible that more bites will be taken during meals eaten outside of the home than meals eaten at home Research Question 5 Does the location of a meal predict the number of bites recorded during a meal Social facilitation Meals are frequently eaten with other people a
115. e y50 intake day bites slope y120 kilocalories x energy density interaction y560 social x intake day interaction y01 gender bites slope y02 body weight bites slope y11 gender x kilocalories interaction p lt 05 133 Table 3 16 Estimates of model fit and random effects for model 11 and exploratory models for the outliers removed model Model fit Random effects Model parameters 2LL e SE 100 SE pe 11 10 24060 94 347 34 9 61 151 43 27 87 00038 lt 001 17 12 24058 08 347 40 9 61 145 62 26 81 00036 lt 001 18 12 24060 04 347 39 9 61 149 49 27 50 00038 001 19 12 24055 14 347 37 9 61 139 64 25 83 00034 lt 001 Note 2LL 2 log likelihood SE Standard Error eij residual within participant variance 100 random intercept between participants variance t10 random slope variance Marginally significant model improvement from Model 11 using the Chi square deviance difference test p 05 Table 3 17 Estimates of fixed effects for level 1 and level 2 predictors for model 11 and exploratory models for the outliers removed model y00 y10 y20 y30 y40 y120 y02 y03 y04 y12 y13 y14 4 Parameters SE SE SE SE SB SE SE SE SE SB SE SE Fe M 39 60 04 5 50 1 84 5 80 01 154 00 5 C8 87 002 is i5 39 60 04 5 51 184 5 8 01 05 9E 5 151 6003 5D C8 87 00 03 5E 5 T T 30 60 04 5 52 1
116. e associated with increased energy intake in previous research Periwal amp Chow 2006 body weight does not seem to be associated with the number of bites taken during a meal This indicates that body weight is most likely not an individual difference characteristic that could guide bite counter kilocalorie calibration settings 171 Body weight and BMI were highly correlated Thus BMI had similar small positive correlations with bites that were only significant for the outliers removed model This indicated that a higher BMI might be associated with taking more bites during meals In exploratory analyses BMI did not significantly predict bites or explain individuals differences in the relationship between kilocalories and bites This indicates that BMI is most likely not an individual difference characteristic that could guide bite counter kilocalorie calibration settings Height In exploratory analyses participant height did not have a positive correlation with the number of bites taken during a meal However for the outliers removed sample at the meal level the slope between height and bites was 0 96 and significant indicating that as height increased by one inch participants took about one fewer bite per meal on average When aggregated to the day level for the outliers removed model the slope between height and bites was 4 49 and significant indicating that as height increased by one inch participants took about 4 to 5 fe
117. e change in model fit was assessed by comparing model 7 to model 6 Results of the y deviance difference test 28543 71 28497 60 46 11 df 2 9 8 1 p lt 05 indicated that the addition of the Kilocalorie x Energy Density interaction significantly improved model fit Next the change in within participants variance from model 6 to model 7 was examined The Kilocalorie x Energy Density interaction explained an additional 1 596 418 42 412 27 418 42 100 of the within participants variance Lastly the Kilocalorie x Energy Density interaction term was negative and significant 0 01 In order to examine the nature of the interaction simple slopes were calculated in accordance with Cohen et al 2003 using the fixed effects coefficients at high 1 SD and low 1 SD values of Kilocalories These slopes were significant at low B 0 05 SE 0 002 t 21 92 p 05 moderate B 0 04 SE 108 0 001 t 30 97 p lt 05 and high B 0 03 SE 0 002 t 17 97 p lt 05 values of Energy Density Figure 3 1 shows that the relationship between Kilocalories and Bites is strongest for low Energy Density meals The Kilocalorie x Energy Density interaction was retained in all subsequent models Low ED 20 k Average ED 10 4 amp High ED Low Kilocalories High Kilocalories Figure 3 1 The Kilocalorie x Energy Density interaction demonstrating that the relationship between Kilocalories and Bites is strongest for
118. e interaction simple slopes were calculated in accordance with Cohen et al 2003 using the fixed effects coefficients at high 1 SD and low 1 SD values of Kilocalories These slopes were significant at low B 0 047 SE 0 005 t 9 16 p lt 05 moderate B 0 040 SE 0 003 t 12 07 p lt 05 and high B 0 033 SE 0 004 t 8 72 p lt 05 values of Height The magnitude and the direction of the slopes did not change from the meal level model Figure 3 6 to the day level model Figure 3 7 shows that the positive relationship between Kilocalories and Bites is stronger for shorter participants and weaker for taller participants 200 180 160 140 120 100 a0 c 7 I Low Height 60 40 20 Average Height 4A High Height Low Kilocalories High Kilocalories Figure 3 7 The Kilocalorie x Height interaction for the outliers removed model at the day level demonstrating that the relationship between Kilocalories and Bites is strongest for shorter participants 140 Bite Size Model Because Height was a significant moderator of the Kilocalories to Bites relationship it was hypothesized that Height was a proxy for Bite Size That is taller participants might have had larger mouths capable of holding more food and thus taller participants might have taken larger bites Therefore participants with a measure of average Bite Size kilocalories per bite from the lab meal were retained in the
119. e recall itself has a large number of questions and steps and recalling more meals and more foods requires a greater time investment by the participant Participants cited the time needed to complete the recall as one of their main frustrations This could have resulted in participants trying to get through the recall process quickly which might have led to incorrect responses to questions about foods details and portion sizes Incorrect responses as well as difficulty finding food items could have led to error in estimation of kilocalories from ASA24 Additionally ASA24 uses pictures to help participants estimate portion sizes However these pictures could lead to perceptual errors and subsequent over or under estimation of the amount of food that was actually consumed Scisco Blades Zielinski amp Muth under review Furthermore the ASA24 is designed for participants to use their memory to complete the recall and reviews of 24 hour dietary recall approaches indicate that participants can typically remember most of their meals with a tendency to underreport Thompson amp Subar 2008 Although the interviewer prompts and the multiple pass method of the ASA24 are designed to reduce underreporting Thompson amp Subar 2008 the time needed to complete the recall electronically could potentially lead to underreporting Also participants in this study expressed great difficulty remembering 180 their meal details unless they used anot
120. e taken your last bite press the left button again to turn off Bite Count mode A beep will indicate that the device has turned off Your data will save automatically and the display will return to Time mode What is a meal A meal is anytime that you are eating and or drinking that has a definite beginning and end That is you know that you will begin eating and or drinking and you can predict when the eating or drinking will end either by finishing all of the food drink or becoming full or satisfied 224 What should I do during a multi course meal If you are eating a multi course meal with extended periods of no eating in between turn the bite counter on and off for each course For example at a restaurant you might turn the bite counter on and off three different times if there are breaks in between each course once for the appetizer once for the entr e and once for the dessert How do I charge the Bite Counter To charge the Bite Counter insert the large end of the USB cable into the power supply and plug the small end of the USB cable into the Bite Counter Plug the power supply into an electrical outlet The display will read chr when the battery is charging and will display Time mode when charging is complete How often should I charge the Bite Counter You should charge the bite counter every night while you are sleeping The bite counter will not work properly if it is not fully charged every 24 hours
121. e with the participant number and date to the JennaDissrtn Scaledata folder e Press the start stop button on the video camera to stop recording Then turn off the video camera f Take off the two bite counters Ask the participant to move back to the conference table Have the participant complete the SLIM scale and LAM scale Now that you have finished the meal I would like you to fill out two quick scales One scale will ask you about your feelings of hunger or fullness and one will ask you how much you liked the meal Ask the participant if there is any other feedback they would like to provide about their experience in the study Record comments on final meeting sheet Debrief the participant We re all done Now I can tell you about the purpose of the study As you know this study is trying to describe the relationship between the number of bites detected by the bite counter during a meal and the number of calories in that meal Additionally I am interested in a number of other predictors of bites including the energy density of the food the duration of the meal the number of people someone eats with where the meal was eaten day of the week gender and body weight Additionally I will use the data from today s meal to calculate your average bite size which may play a role in these relationships Do you have any questions about your participation in the study Ask the participant to f
122. ean and obese women American Journal of Clinical Nutrition 73 1010 1018 249 Bickel R 2007 Multilevel analysis for applied research It s just regression New York The Guilford Press Bishop K L 2002 Longitudinal predictors of weight fluctuation in men and women Doctoral dissertation Retrieved from ProQuest Dissertations and Theses database UMI No 3088037 Black J L amp Macinko J 2008 Neighborhoods and obesity Nutrition Reviews 66 2 20 doi 10 1111 j 1753 4887 2007 00001 x Blanton C A Moshfegh A J Baer D J amp Kretsch M J 2006 The USDA automated multiple pass method accurately estimates group total energy and nutrient intake Journal of Nutrition 136 2594 2599 Bosker R J Snijders T A B amp Guldemond H 2003 PINT Power IN Two level designs user s manual version 2 1 Retrieved from http stat gamma rug nl multilevel htm Boutelle K N amp Kirschenbaum D S 1998 Further support for consistent self monitoring as a vital component of successful weight control Obesity Research 6 3 219 224 Boutelle K N Kirschenbaum D S Baker R C amp Mitchell M E 1999 How can obese weight controllers minimize weight gain during the high risk holiday season By self monitoring very consistently Health Psychology 18 4 364 368 Burger K S Fisher J O amp Johnson S L 2011 Mechanisms behind the portion size effect Visibility and b
123. ecause Intake Day was naturally a day level variable an ICC2 value did not need to be calculated Table 3 9 ICC2 values for level 1 variables Variable ICC2 Bites 44 Kilocalories 42 Energy Density 19 Location 50 Social 49 All ICC2 values were less than 0 60 and typically one would not aggregate these variables up to the day level because important variability would be lost Nonetheless in order to explore a model with level 1 representing the day these variables were aggregated up to the day level 121 In the day level model the sum of day level values within a participant were used for each aggregated variable Meal energy density for this model was calculated as the sum of the kilocalories for the day divided by the sum of the grams for the day All rows in the data set represented a day thus day become level 1 and participant remained level 2 The sums were used in this model because this might be a practical way for an individual to interpret bite counter data 1 e someone might want to know how the total number of bites for a day is related to the total number of kilocalories for a day All predictor variables were centered at the grand mean Model 19 the final model at the meal level was run using the data at the day level The random Kilocalories Bites slope variance became non significant in the day level model ro 2 1E 4 SE 8E 5 Wald Z 1 70 p gt 05 This indicated that the relationship betw
124. ed to enter the meals during the day instead of all at once the following day When asked how they changed their eating behavior as a result of using ASA24 participants described that becoming more aware of what they were eating and portion sizes helped them to eat healthier and eat smaller portions Participants reported not eating foods that were difficult to find in the database or unnecessary snacks so that they would not have to enter them into ASA24 later Some participants focused on consuming 155 food and beverage during meals and snacking less between meals One participant stated I felt like I had to be sitting down and have organized meals One participant reported eating simpler foods with fewer ingredients which would make the food easier to report in ASA24 One participant described eating more than usual in order to provide more data for the study Participants recorded their meals in a variety of places other than ASA24 to aid their daily reporting Many participants used the small notebook provided which was described as invaluable Others used their day planners calendars tablets phones computer sticky notes and e mails chains to themselves to record details about their intake during the day In addition to ASA24 other recall type programs were used by some participants including Fat Secret Livestrong and My Fitness Buddy Table 3 28 shows the frequency of responses for questions about the bite counter Wh
125. ee customer orientation in context How the environment moderates the influence of customer orientation on performance outcomes Journal of Applied Psychology 94 5 1227 1242 doi 10 1037 a0016404 Grunberg N E amp Straub R O 1992 The role of gender and taste class in the effects of stress on eating Health Psychology 11 97 100 235 Haeffele J 2008 A grounded theory approach to the process of successful weight maintenance Doctoral dissertation Retrieved from ProQuest Dissertations and Theses database UMI No 3340331 Harvey Berino J Pintauro S Buzzell P DiGiulio M Gold B C Moldovan C amp Ramirez E 2002 Does using the Internet facilitate the maintenance of weight loss International Journal of Obesity 26 1254 1260 Heck R H Thomas S L amp Tabata L N 2010 Multilevel and longitudinal modeling with IBM SPSS New York NY Routledge Helsel D L Jakicic J M amp Otto A D 2007 Comparison of techniques for self monitoring eating and exercise behaviors on weight loss in a correspondence based intervention Journal of the American Dietetic Association 107 1807 1810 doi 10 1016 jada 2007 07 014 Herman C P Roth D A amp Polivy J 2003 Effects of the presence of others on food intake A normative interpretation Psychological Bulletin 129 6 873 886 Heron K E amp Smyth J M 2010 Ecological momentary interventions Incorporating mobile
126. een Bites and Kilocalories at the day level did not vary between participants This random effect was subsequently removed from the model as was the cross level interaction between Kilocalories and Height A final model at the day level was evaluated with Kilocalories Energy Density Kilocalories x Energy Density Location Social Intake Day Gender and Height in the model as fixed effects Table 3 10 provides the random effects for the final meal level model and the final day level model and Table 3 11 provides the fixed effects for the final meal level model and the final day level model to aid in comparison across the models 122 Table 3 10 Random effects for the meal level and the day level models 101 Model ei SE t00 SE Kcalories Meal level 378 09 9 74 157 83 26 94 00037 lt 001 Day level 1615 17 73 36 2086 17 345 05 n a Note SE Standard Error eij residual within participant variance 100 random intercept between participants variance t10 random slope variance p lt 05 Table 3 11 Fixed effects for the meal level and the day level models Model 00 Y10 Y20 y30 y40 y50 y120 y01 y04 vid SE SE SE SE SE SE SE SE SE SE Meallevel 40 94 04 5 81 79 5 73 1 88 01 1 50 83 002 1 45 003 50 82 86 82 002 3 70 54 7E 4 RENE 121 77 03 27 57 1 30 5 30 8 15 01 26 31 1 52 ify id 5 18 003 4 57 1 30 1 43 3 06
127. el was very low when returned which indicated a possible user error Many long duration meals Bite counter turned off during a few meals First bite counter turned off frequently Second bite counter had no problems Good 242 12 Good 16 Good 14 Good 11 Good 14 Good 17 Good 14 Good 13 Good 14 Good Bites Bite Counter 4 ASA24 ASA24 ID matched matched Kilocalories problems data comblsted problems data meals meals correlation quality P quality Bite counter turned BiteCD034 25 78 1 223 off during a few 13 Good meals Time drift First bite counter turned off during a few meals Second bite counter was BiteCD038 26 65 0 513 better but a number 13 Good of errors long duration meals with few bites were removed No bite counter c Pu Me large which BiteCD041 15 55 6 066 recordings 13 ee very long durations parucipa and high bile description of conis eating one large meal per day BiteCD043 39 83 0 321 Good 14 Good BiteCD0S1 32 76 2 481 P ORE MEET Good off once Difficulty reporting protein shakes modified eating BiteCD055 42 95 5 207 Good 14 Org protein shakes abnormal eating less food for 3 4 days due to ear infection BiteCD056 40 87 0 548 Good 14 Good Bite counter turned MR BiteCD060 25 69 4 494 off during a few 11 meals 243 in sleeping schedule Bites Bite Counter 4 ASA24 ASA24 ID matched matched Kilocalori
128. en An observational validation study Journal of the American Dietetic Association 104 595 603 doi 10 1016 j jada 2004 01 007 251 Conway J M Ingwersen L A Vinyard B T amp Moshfegh A J 2003 Effectiveness of the US Department of Agriculture 5 step multiple pass method in assessing food intake in obese and nonobese women American Journal of Clinical Nutrition 77 1171 1178 Cope D R Loonen M J J E Rowcliffe J M amp Pettifor R A 2005 Larger barnacle geese Branta leucopsis are more efficient feeders a possible mechanism for observed body size fitness relationships Journal of Zoology 265 37 42 Crawford D Jeffrey R W amp French S A 2000 Can anyone successfully control their weight Findings of a three year community based study of men and women International Journal of Obesity 24 1107 1110 Cuntz U Leibbrand R Ehrig C Shaw R amp Fichter M M 2001 Predictors of post treatment weight reduction after in patient behavioral therapy International Journal of Obesity 25 S1 S99 S101 Cushing C C Jensen C D amp Steele R C 2010 An evaluation of a personal electronic device to enhance self monitoring adherence in a pediatric weight management program using a multiple baseline design Journal of Pediatric Psychology doi 10 1093 jpepsy jsq074 de Castro J M 1991 Weekly rhythms of spontaneous nutrient intake and meal pattern of humans Physiolog
129. en Vinyard amp Moshfegh 2003 The AMPM was also found to accurately reflect total energy intake in free living humans with underreporting of energy intake increasing for those with greater BMIs Moshfegh et al 2008 In addition the AMPM has been shown to provide a more valid measure of total energy intake compared to other energy intake measures such as the Block food frequency questionnaire and National Cancer Institute s Diet History Questionnaire Blanton Moshfegh Baer amp Kretsch 2006 Drawbacks to the AMPM recall are the costs associated with training interviewers and the impracticality of interviewers administering recalls in person or over the telephone in a large sample study Subar et al 2007 The ASA24 dietary recall addresses these problems by allowing participants to complete recalls unassisted at any time during a recall day using an Internet based recall program The majority of ASA24 development has been guided by experts in the field of dietary assessment Zimmerman et al 2009 Some studies with users of the ASA24 system have also guided software 13 development A pilot study of the Quick List indicated that participants preferred recalling by meal e g breakfast lunch rather than recalling all foods for one day together Subar et al 2007 Additionally this pilot study indicated that the act of scrolling through food lists helped to trigger memories of foods and beverages eaten an advantage over the
130. en gender and bites in the final models were not significantly different from zero Gender also did not explain any differences in the relationships between kilocalories and bites between participants It is possible that men might take 170 more bites in order to consume more kilocalories McArdle et al 2005 but women might take more bites if they are taking smaller bites Burger et al 2011 These two effects could possibly counteract each other resulting in no consistent relationship between gender and bites found in the present study This indicates that gender is most likely not an individual difference characteristic that could guide bite counter kilocalorie calibration settings Body Weight and BMI Research question 10 investigated if body weight could predict bite count Overall correlations between body weight and bites were very small positive and only significant for the outliers removed sample This indicated that a higher body weight might be associated with taking more bites during meals Body weight explained 2 2 of the between participants variance in full sample model and 5 0 of the between participants variance in the outliers removed model Slopes between body weight and bites in the final models were not significantly different from zero Body weight also did not explain any differences in the relationships between kilocalories and bites between participants Thus although a higher body weight has been found to b
131. er Blanck and Gillespie 2006 surveyed 1 958 people who had tried to lose weight and reported that 30 maintained a weight loss whereas 70 failed to maintain a weight loss They found that regular exercise differentiated the two groups with successful weight maintainers exercising more often Interestingly successful weight maintainers also reported more self monitoring including planning meals tracking calories tracking fat and measuring the food on their plate on most days of the week When reviewing the literature on successful weight loss and weight maintenance it becomes clear that self monitoring is an essential part of the weight loss and weight maintenance process Accurate and reliable tools may help individuals self monitor consistently Relatively new technologies including the Internet lightweight data loggers such as pedometers and accelerometers and short message service SMS via cellular phones have the potential to improve self monitoring efforts Svensson amp Lagerros 2010 Our research group has developed a new self monitoring tool the bite counter device which has the potential to change the way individuals self monitor their food intake Hoover Muth amp Dong 2009 In order for the bite counter to be an effective self monitoring tool we must understand how an individual should use the device We can begin to develop this understanding with a thorough review of the self monitoring literature and exi
132. ers compared to a meal eaten alone a finding that is very similar to the de Castro and de Castro 1989 finding that meals eaten with others contained about 180 more kilocalories than meals eaten alone The slopes between social and bites at the day level were 5 76 and 3 80 for the full sample and the outliers removed sample respectively This indicated that for every additional meal eaten with others during a day participants took between 4 and 6 additional bites per day The reduction in the number of additional bites taken for the day level model for the outliers removed sample suggests that eating with others may not have as strong of a relationship with number of bites taken for the entire day compared to number of bites taken during a meal This is similar to the finding by de Castro 1996 that social facilitation is a stronger predictor of meal size than daily food intake Also findings were very similar for Social in the meal level and day level models with Bite Size that included 60 participants The practical implication of this finding is that the bite counter may provide individuals with some information about their eating patterns when they eat with others If individuals are made aware of the tendency to take more bites when eating with others they could try to monitor bites during these meals and keep their number of bites taken during meals eaten with others similar to the number of bites taken during meals eaten 168 alone Tha
133. es option being able to save commonly eaten foods for quick entry They sometimes had trouble 154 finding foods especially if the food was international cuisine and thought that some options were incomplete or unclear Many participants described frustration with the penguin interviewer providing instructions and slowing down the recall process When ASA24 was initially released the penguin would provide instructions for every recall About halfway through data collection December 28 2011 ASA24 was updated so that participants were asked on their second and all subsequent recalls if they wanted the penguin s help or if they wanted to turn him off This appeared to eliminate frustration with the penguin Participants described the interface as unwieldy and not stream lined with too much mouse clicking and not enough opportunity to use the keyboard Needing Internet access was sometimes troublesome and sometimes the program would slow down or freeze which was the source of many of the reported problems with ASA24 Participants who wanted to use Apple products e g iPhone iPad or the Linux operating system were disappointed to learn that ASA24 was not compatible Downloading the new version of Microsoft Silverlight was difficult for some participants but this problem was always resolved through troubleshooting Finding the time to complete the recall was difficult for participants with busy schedules Some participants want
134. es problems data problems data completed meals meals correlation quality quality BiteCD063 44 74 6 644 Time drift 13 ED EPIS out of 17 BiteCD065 31 79 5 667 Good 12 Good Bite counter turned BiteCD069 37 88 1 323 off during a few 13 Good meals Bite counter turned BiteCD073 43 78 2 517 off during a few 13 Good meals BiteCD074 50 66 7 539 Pile COMME iae Good off once Bite counter turned BiteCD075 32 86 5 247 off during a few 12 Good meals Carnation instant Bite counter turned breakfast errors BiteCD077 25 59 5 314 off during a few 11 pathway of meals questions errors removed Bite counter turned Incomplete BiteCD078 28 62 2 660 off during a few 12 recalls meals removed BiteCD083 32 88 9 543 Time drift 13 Good BiteCD084 44 75 9 381 Good 14 Good BiteCD094 35 68 6 419 Good 13 Good 244 Bites Bite Counter 4 ASA24 ASA24 ID matched matched Kilocalories problems data problems data completed meals meals correlation quality quality Bite counter turned BiteCD095 32 84 2 017 off during a few 14 Good meals Time drift BiteCD096 38 63 3 409 Good 14 Good BiteCD097 50 86 2 678 Good 14 Good BiteCD100 30 88 2 519 Good 14 Good BiteCD101 45 95 7 580 Ee coumarins qu Good off twice BiteCD104 35 89 7 553 Good 14 Good Good Extended data BiteCD108 39 90 7 532 Good 15 collection due to personal emergency Missed 8 recalls BiteCD125 18 72 0 575 Good 12 outer 20 due to exam s
135. ess continues until the likelihood does not improve by more than an amount known as the convergence criterion Cohen et al 2003 There is no analytic solution to ML estimation meaning that there is not a set of equations from which the coefficients are directly calculated given its iterative nature Cohen et al 2003 ML estimation is made possible by high speed computers and an iterative computational procedure that can run hundreds to thousands of estimations until convergence is reached Bickel 2007 Cohen et al 2003 Hox 2010 Restricted maximum likelihood REML or RML is a preferred method of ML for smaller samples because it uses a likelihood function to take into consideration the number of parameters being estimated in the model Bickel 2007 and is less biased Hox 2010 REML includes only the variance components in the likelihood function and the parameter estimates are estimated separately Hox 2010 ML which includes the variance components and the parameter estimates in the likelihood function Hox 2010 should be used when comparing fit across incremental models Tabachnick amp Fidell 2007 MLM Equations Returning to the present example with level 1 job status and level 2 gender predicting GPA the full MLM regression equation can be built from a series of equations at each level The level one model is represented by equation 1 1 using conventional 39 notation for MLM Bickel 2007 Hox 2010 The
136. evice that can provide real time kilocalorie feedback to the user Imagine a bite counter that is shipped to a future user along with a microwavable calibration meal The user would eat this low energy density calibration meal while recording bites with the device The kilocalories bites ratio determined with this low energy density calibration meal would be used to set the bite counter s kilocalorie conversion setting for that individual Kilocalories Low ED Kilocalories bites ratio Bites For example if a person eats a 500 kilocalorie calibration meal in 20 bites the kilocalories to bites ratio would be 500 20 25 Inserted into the above equation Kilocalories Low ED 25 Bites This equation would then be modified by the person before eating meals by entering the energy density of the meal into the bite counter For example if four categories of 191 energy density are used Very Low Low Medium and High the user would select the energy density of the meal using an energy density menu feature and one of four equations would be used to adjust the kilocalories bites relationship Kilocalories Very Low ED 0 8 25 Bites Kilocalories Low ED 1 25 Bites Kilocalories Medium ED 1 3 25 Bites Kilocalories High ED 2 25 Bites The coefficients for these equations are based on the simple slopes obtained from the kilocalorie energy density interaction and these coefficients would need to be replicated and tested in future st
137. evice to turn on If there may be a need to remove or correct the data flag the data c Was there a delay in turning on the bite counter or turning off the bite counter If so flag the data 2 Bite Counter data a Bite Counts Flag values 10 and 50 b Meal duration Flag values 1 minute and 45 minutes 3 ASA24 data a MealKCAL Flag values lt 50 and gt 1000 b MealED Flag values 0 and gt 4 0 c Hag incomplete recalls d Flag incomplete foods 4 Data sheets a Did the participant report any problems at either the data download meeting or the final meeting If so flag affected meals 5 E mails a Did the participant report any problems at any time via e mail If so flag affected meals 6 Usability questionnaire a Did the participant report any new problems in their usability questionnaire If so flag affected meals 7 Go back to the flagged meals Using the decision making flow charts in Figure 2 10 and 2 11 decide if data should be removed corrected or kept the same Take the appropriate action a When a meal is removed add it to the removed tab This will allow you to keep all of the data if you decide to use it later b Record all actions in ParticipantID data merging and screening history docx 239 Step 3 Create data summary for the participant 1 In the Dissertation Data Merged and screened datalParticipantID folder create a new Excel workbook named ParticipantID dat
138. f a clear hunger satiation goal in the restaurant and the absence of this goal in the laboratory That is the laboratory environment was more artificial and participants may not have sought to reduce hunger which made them more susceptible to anchoring on the bite size cue However in a restaurant they may have seen the small bite size as feedback that they were not making much progress on reducing their hunger and thus they ate more in order to reach visual cue based satiation Therefore it is possible that there is a positive relationship negative relationship or no relationship between kilocalories and bites Because there is no published research examining the relationship between bites and kilocalories in humans eating in their daily environments this study will be the first to explore this kilocalorie bite relationship Research Question 1 Do kilocalories consumed during a meal predict number of bites recorded during a meal Energy density Energy density is defined as the number of kilocalories per gram in a given food Rolls Ello Martin amp Ledwicke 2005 Differences in water and fat contents between foods tend to have the largest impact on energy density Yao amp Roberts 2001 More water in a food is associated with decreased energy density due to 48 water s zero energy content whereas more fat in a food is associated with increased energy content because fats are roughly twice as energy dense as proteins a
139. f computer use Dietary restraint Cognitive restraint emotional eating and uncontrolled eating were measured using the Three Factor Eating Questionnaire R 18 TFEQ R 18 Appendix B de Lauzon et al 2004 Daily meals questionnaire Additional features of the meal not described by the bite counter data or the ASA24 data were obtained with an additional survey Appendix C The survey asked participants to report their bite counter usage and technical problems additional activities they engaged in while the bite counter on the utensils used hunger fullness palatability the number of people they ate with for each meal and who prepared the meal The survey also asked participants to estimate their daily physical activity Usability Participants completed a usability questionnaire during their last visit to the laboratory on Survey Monkey Appendix D This questionnaire assessed problems difficulties likes dislikes and preferences for the ASA24 dietary recall and the bite counter Procedure Pre screening The procedures for online pre screening are described in Appendix E When the participant contacted the researcher to participate in the study the researcher sent the 76 participant a link to complete an online consent form the demographics questionnaire and the TFEQ R18 on Survey Monkey Participants with a history of an eating disorder were excluded from the study as using the bite counter and completing dietary recalls
140. for the full sample and the outliers removed sample respectively This indicated that as the average number of kilocalories per gram for a day increased by 1 participants took about 27 to 32 fewer bites per day To make these results more meaningful it is important to put them in the context of average food energy densities Rolls 2007 describes four energy density categories 1 Very Low Energy Density 0 0 6 kcals g foods such as non starchy fruits and vegetables nonfat milk and broth based soups 2 Low Energy Density 0 6 1 5 kcals gram foods such as starchy fruits and vegetables grains breakfast cereals with low fat milk low fat meats beans and legumes and low fat mixed dishes such as chili and spaghetti 3 Medium Energy Density 1 5 4 0 kcals gram foods such as meats cheeses pizza French fries salad dressings bread pretzels ice cream and cake and 4 High Energy Density 4 0 9 0 foods such as crackers chips chocolate candies cookies nuts butter and oils Applying this information to the study results if a person was eating a very low energy density meal e g 0 5 kcals gram consisting of a fruit and vegetable salad it could be expected that they would take about 18 more bites compared to eating a medium energy density e g 3 5 kcals gram meal of a burger and fries However a significant interaction between kilocalories and energy density as described below indicates that the main effect of energy density shoul
141. from the Daily meals questionnaire and paste into the Merged Data sheet next to the associated bite counter data If data from the 237 questionnaire is missing write missing data in the empty cells Ifthe bite counter data is missing create a new row and insert the questionnaire data ASA2A data In Dissertation Data ASA24 BiteCD_Request196_AllData open BiteCD 776_INF csv Copy and paste all of the data for the participant into the INF sheet in ParticipantID xls a Hide cells so that the following are visible UserName RecallNo RecallStatus IntakeDate IntakeDay Occ No Occ Time FoodAmt KCAL FoodComp Food Description If foods are incomplete check the MS file in Dissertation Data ASA24 BiteCD_Request196_AllData for the food portion and detail responses Insert any known values into the INF file based on this information from MS If values are unknown and the data set is thus missing necessary KCAL and gram data mark this as missing data in the INF sheet in ParticipantID xls Create MealFoodAmt and MealKCAL columns Sum up FoodAmt and KCALs for each meal i SUMIFS FoodAmt range RecallNo range RecallNo Occ No range Occ No ii SUMIFS KCAL range RecallNo range RecallNo Occ No range Occ No iii The first row for each meal will have the correct totals iv Move additional food descriptors up to the first row for each meal using copy and paste transpose v Hide rows below each meal s first ro
142. gotten foods or drinks questions to which they must have responded yes or no Figure 2 7 If they responded yes they returned to the Quick List to add the foods or drinks Before finishing the Last Chance option was provided for additions or changes to be made The Last Chance question was followed by a Trailer Question that asked the participants to report if the amount of food consumed was more than usual usual or much less than usual A number of features make the ASA24 program unique and comprehensive including a tutorial on how to complete the recall an animated audible character to guide participants through the interview a penguin Show Me video clips for major sections allowing participants to find foods by browsing through defined food groups or by searching for keyed text using photographs to assist participants in reporting portion size a module to asses who a participant was eating with and a module to assess where a meal was consumed 68 Meal Details Enter the details of the first meal or snack you would like to report e Meal or snack Lunch Time of meal or snack VAM 22 0 oi Location Fast food restaurant TV and computer use while eating and drinking Usingacomputr 00000000000 Did you eat with anyone yes No Don t know Family Member s L Other s Actions Select an action below to edit your foods and drinks Add a meal or snack Delete a
143. he ASA24 recall missing data incomplete data database error pathway of questions error or user entry error Errors were either corrected or removed from the dataset A flowchart describing the decision making process for bite counter data error identification correction and removal is shown in Figure 2 8 A flowchart describing the decision making process for ASA24 data error identification correction and removal is shown in Figure 2 9 The red parallelograms at the top of each figure refer to the possible errors that could be flagged when following the screening steps in Appendix O 82 Bite Counter Data Correct data increase or decrease by percentage of missing or Figure 2 8 Bite counter data decision making process for error identification correction and removal 83 ASA24 Data Figure 2 9 ASA24 data decision making process for error identification correction and removal 84 In order to demonstrate the decision making process for error identification correction and removal a number of examples are provided Starting with the bite counter data a turning off data series was a frequent error identified in the raw bite counter data For example participant BiteCD012 had a small snack of 108 kcal of Captain Crunch cereal reported at 11 07PM on November 20 When this was matched with the bite counter data three lines of data were found at that time meals 97 98 and 99 As can be seen in Figure 2 10 the
144. he model one at a time and their unique contribution to the model was assessed If predictors did not improve model fit explain bite variance or have significant fixed coefficients they were dropped from subsequent models After running the intercept only model as described above level 1 variables were entered into the model as fixed effects one at a time After each level 1 variable was added the level 1 interactions were added Model fit was compared using the 2 log 90 likelihood x deviance difference test with degrees of freedom as the number of added parameters Hox 2010 If the a difference between two models was above the critical value for the associated number of degrees of freedom this was evidence of improved model fit The change in residual variance as level 1 variables were added to the model indicated the unique amount of within participants variance explained by each predictor The fixed coefficient for each predictor was examined for significance using its associated t test Then level 2 variables were entered into the model as fixed effects one at a time In addition to examining the X deviance difference test and the significance of the fixed coefficient the change in intercept variance indicated the unique amount of between participants variance explained by each level 2 predictor Next the slopes between level 1 predictors and Bites were allowed to vary one at a time and random slope variance that significantly i
145. her method to record their meal details at the time of the meal such as the invaluable small notebook This recording of details in the notebook most likely reduced another cited benefit of 24 hour recalls that they have less of an influence over eating behavior at the time of the meal Thompson amp Subar 2008 An alternative to the ASA24 for future studies could be a dietary intake recording tool that allows meal details to be entered at the time of the meal This might be preferred by some participants because they could input their meal information during smaller time periods throughout the day rather than dedicating a larger single period of time in the morning or evening trying to remember details from the previous day There are a number of popular programs available for mobile devices such as FatSecret and LiveStrong but the accuracy of the kilocalorie databases would need to be examined prior to use in a research study An advantage of the ASA24 is that it uses the USDA s Food and Nutrient Database for Dietary Studies FNDDS Although the 24 hour recall is considered the best self report instrument available for estimating dietary intake Kirkpatrick 2011 there may be other methods that participants in future bite counter studies may find easier to complete If participants are already turning the bite counter on and off and making notes about details of bite counter use taking a few more minutes to record the foods eaten may no
146. high levels of physical activity The NWCR researchers have also addressed how changes in popular diets over time have affected successful weight loss maintenance Phelan Wyatt Hill and Wing 2006 tracked dietary intake of registry members from 1995 2003 Dietary trends were found to reflect popular diets As dieters transitioned from low fat diets to low carbohydrate diets registry members obtained a greater percentage of their calories from fat consumed more saturated fat and obtained a lower percentage of their calories from carbohydrates However over 75 of the registry members were still at or below recommended levels of fat intake Vegetable consumption and dietary fiber from vegetables fruits and beans also increased during this time period The researchers concluded that individuals can lose and maintain weight loss on a variety of diets Overall the NWCR has identified common behaviors that result in successful weight loss maintenance a low calorie low fat diet consuming breakfast regularly engaging in high levels of physical activity about 1 hour per day walking is the most common activity regular self weighing and being mindful of one s diet and physical activity Hill Wyatt Phelan amp Wing 2005 Wing amp Phelan 2005 Maintaining weight loss is associated with maintaining these behavioral changes long term and consistently across weeks weekends holidays and non holidays Hill et al 2005 Wing amp Phelan
147. ht and Bites However the Height x Kilocalories interaction term 0 002 was negative and significant In order to examine the nature of the interaction simple slopes were calculated in accordance with Cohen et al 2003 using the fixed effects coefficients at high 1 SD and low 1 SD values of Kilocalories These slopes were significant at low B 0 047 SE 0 004 t 12 51 p lt 05 moderate B 0 040 SE 0 003 t 15 62 p lt 05 and high B 0 033 SE 0 003 t 9 67 p lt 05 values of Height Figure 3 2 shows that the positive relationship between Kilocalories and Bites is stronger for shorter participants and weaker for taller participants 118 B Low Height 20 Average Height 10 4r High Height Low Kilocalories High Kilocalories Figure 3 2 The Kilocalorie x Height interaction at the meal level demonstrating that the relationship between Kilocalories and Bites is strongest for shorter participants The Final Model Model 19 was the best fitting model for explaining variance in bites In order to calculate the overall effect size for the model all predictors in the model needed to be fixed with no random slopes Bickel 2007 Therefore model 19 was run without the random Kilocalories Bites slope variance For this model the residual variance was 408 13 and the intercept variance was 162 59 The overall effect size was calculated as l residualgxeq interceptrixea residualintercepts only
148. icipantID 231 10 11 a Name the file ParticipantID_DeviceNumber_MonthDayY ear b Check the data for errors and ask the participant about any error like data For example if there are a lot of zeros or short meals with few bites is the device turning off or are they testing the device c Record any problems on the final meeting sheet Measure the participant s height and weight body fat percentage and waist and hip circumference Record values on the final meeting sheet Ask participant to complete the usability survey on the computer While participant completes the survey prepare macaroni and cheese according to package instructions in the microwave Pour 500 mL of water into a glass for drinking Place the water and macaroni and cheese on the table The macaroni and cheese should be placed in its container on top of the plate that is on the scale When the participant is done with the survey explain the purpose of the meal Today we will be collecting some data on feelings of hunger and fullness and enjoyment of a meal I have prepared macaroni and cheese for you to eat today The session will be video taped Additionally you will be wearing two different bite counters on your dominant wrist Before we begin with the meal I would like you to fill out a quick scale asking about feelings of hunger or fullness Please make a slash mark crossing the vertical line to indicate your current feeling of hunger or fullness
149. ics 2012 that recommend distributing caloric intake throughout the day in 4 5 meals and snacks Research supporting this official recommendation seems to be mixed In a review of cross sectional and longitudinal studies of adults snacking behaviors were found to be associated with increased body weight Mesas Munoz Pareja Lopez Garcia amp Rodriguez Artalejo 2012 However in a review of weight loss and weight maintenance interventions eating frequency one definition of snacking behavior was not associated with body weight or related health outcomes Palmer Capra amp Baines 2009 Identifying relationships between snacking body weight and health is difficult because definitions of snacking are not consistent in the literature and changes in eating frequency may be difficult for individuals to sustain over time Palmer Capra amp Baines 2011 Therefore if reduced snacking was a mechanism by which this study led to weight loss it is possible that this effect may not persist over the long term 179 Implications of ASA24 and Bite Counter Usability The usability questionnaire provided important insights into participant s impressions of the study tools The ASA24 dietary recall is a new Internet based automated recall system designed by the National Cancer Institutes to be a dietary intake research tool Most participants completed about 12 to 14 recalls which indicated that the AS A24 was acceptable for daily use However th
150. ienced any technical difficulties or had questions Data download meeting The protocol for this 15 minute meeting is described in detail in Appendix K After about 7 days of data collection the participant came to the laboratory for data downloading and bite counter reset If minor bite counter problems were seen in the data typically trouble getting the bite counter to stay on which looked like a series of zero or one bites followed by a full recording the experimenter reviewed 79 the correct way to turn the bite counter on and off with the participant and provided recommendations for getting the bite counter to stay on These recommendations included charging the device overnight every night not wearing the device too tightly on the wrist and waiting an additional 10 seconds after the device said on to begin moving the wrist If severe bite counter problems were detected many zero and one bite sessions with few full recordings the experimenter gave a new bite counter and charger to the participant to use for the remaining week The experimenter also gave the recommendations described above for minor problems because the data errors could have been due to device failure user error or a combination of the two In both cases the experimenter also ran the device test mode to check that the sensor was operational and to check the battery level If a low battery level was detected this guided the experimenter s troubleshooting and p
151. ies and Bites vary by participant Models 11 15 allowed the slopes between Bites and a level 1 predictor to vary by participant one variable at a time Hox 2010 If a significant random slope variance was found this was retained in the model and a cross level interaction was added to try to explain these varying slopes with a level 2 predictor In model 11 the relationship between Kilocalories and Bites was allowed to vary by participant random Kilocalories Bites slope variance First the change in model fit was assessed by comparing model 11 to model 9 Model 11 was not compared to model 10 because model 10 was not a significant improvement over model 9 and its predictor was dropped from subsequent models Results of the y deviance difference test 111 28493 51 28312 93 180 58 df 12 10 2 p lt 05 indicated that the addition of Kilocalories Bites slopes varying by participants significantly improved model fit The random Kilocalories Bites slope variance of 0 0004 was significant indicating that the relationship between Kilocalories and Bites did vary by participant Therefore the random Kilocalories Bites slope variance was retained for all subsequent models Model 12 Does the Relationship between Energy Density and Bites vary by participant In model 12 the relationship between Energy Density and Bites was allowed to vary by participant random Energy Density Bites slope variance The a deviance difference test co
152. ifference test 28497 60 28493 51 4 09 df 10 9 1 p lt 05 indicated that the addition of Gender significantly improved model fit Next the change in between participants intercept variance from model 7 to model 9 was examined Gender explained 5 2 184 17 174 64 184 17 100 of the between participants variance Lastly a significant positive relationship between Gender and Bites was observed in the Gender Bites slope of 6 18 On average 6 18 more bites per meal were recorded for females compared to males Gender was retained as a level 2 predictor for all subsequent models 110 Model 10 Does Body Weight predict Bites Body weight was added to the model as a fixed effect at level 2 in order to address research question 10 Does body weight predict bite count First the change in model fit was assessed by comparing model 10 to model 9 Results of the a deviance difference test 28493 51 28491 73 1 78 df 11 10 1 p gt 05 indicated that the addition of Body Weight did not significantly improve model fit Next the change in between participants intercept variance from model 9 to model 10 was examined Body Weight explained 2 24 174 64 170 72 174 64 100 of the between participants variance However the Body Weight Bites slope was non significant 0 05 Because Body Weight did not improve the model or its interpretation it was dropped from subsequent models Model 11 Does the Relationship between Kilocalor
153. ijlstra N de Wijk R A Mars M Stafleu A amp de Graaf C 2009 Effect of bite size and oral processing time of a semisolid food on satiation American Journal of Clinical Nutrition 90 2 269 275 doi 10 3945 ajcn 2009 27694 Zimmerman T P Hull S G McNutt S Mittl B Islam N Guenther P M Subar A F 2009 Challenges in converting an interviewer administered food probe database to self administration in the National Cancer Institute automated self administered 24 hour recall ASA24 Journal of Food Composition and Analysis 22S S48 S51 doi 10 1016 j jfca 2009 02 003 Zoumas Morse C Rock C L Sobo E J amp Neuhouser M L 2001 Children s patterns of macronutrient intake and associations with restaurant and home eating Journal of the American Dietetic Association 101 8 923 925 265
154. ile 26 5 of participant wore the bite counter all day as instructed 30 1 wore it only during meal times and 42 4 found a middle ground between all day and just mealtimes Participants found the bite counter easy to use because they only had to press a button to turn it on and off Some people liked that it was on the wrist easily portable functioned as a watch and could be strapped to a lunch bag or the refrigerator handle They described using the device as not rocket science a no brainer user friendly and that it became second nature The audible and visual feedback was helpful for knowing when the device was turned on and off Some participants liked being asked about the device by friends and coworkers so that they could tell them about their participation in the study Participants liked that it increased their awareness of eating 156 Table 3 28 Responses to usability questions about the bite counter Question N of total sample Frequency of wearing bite counter All day everyday from morning to evening 22 26 5 Only part of the day more often than meals 35 42 2 Only during meals took it off other times 25 30 1 Did not wear it during some meals 1 1 2 Ease or difficulty of use Extremely easy 26 31 3 Very easy 38 45 8 Somewhat easy 11 13 3 Neither easy nor difficult 5 6 0 Somewhat difficult 2 2 4 Very difficult 1 1 2 Liked or disliked Extremely liked 2 2 4 Liked very much 9 10 8 Liked somewhat 21
155. ilk Cream creamers Common Additions Honey 1 milk Lemon 2 milk Lemonade Buttermilk Milk Skim milk Soy milk Sugars and sugar substitutes Dry milk 1 fat Other Additions Acidophilus milk Calcium fortified milk skim or nonfat Coconut milk Dry milk lowfat Dry milk skim or nonfat Dry milk unknown type Q4 Previous REVIEW Are these ALL the foods and drinks you had Yesterday Lunch 12 00 PM McDonald s Double Quarter Y Pounder with Cheese b French fries Coca Cola v Dinner 08 00 PM gt Chicken Caesar salad P Tea hot or iced regular B Chips Ahoy y Snack 10 30 PM gt Cheez It crackers regular gt Lemonade regular Show Food Detail Make Changes Noa Figure 2 6 Final review of foods drinks and details 71 FREQUENTLY FORGOTTEN FOODS Certain foods and drinks are frequently forgotten Did you forget to report any of the following foods and drinks Please respond to each item by selecting Yes or No In addition to the foods and drinks you already reported did you have any Water including tap faucet bottled water fountain Coffee tea soft drinks milk or juice Beer wine cocktails or other drinks Cookies candy ice cream or other sweets Chips crackers popcorn pretzels nuts or other snack foods Fruits vegetables or cheese Breads rolls or tortillas Anything else Figure 2 7 Forgotten foods prompt
156. ill out the compensation form Tell the participant they will receive their data summary via e mail within 4 weeks 233 After the participant leaves l Save the Intertia cube data It is important to do this step immediately after the participant leaves because the data will be written over if the files are not renamed and moved a Goto Computer Local disk C Jenna b There will be two files BiteDetect txt and OriginalData txt c Rename the files BiteDetect_Participant txt and OriginalData_Participant txt d Cut the files and paste them to the Desktop JennaDissrtn Bite Counter Data folder Upload to dropbox Save the bite counter data Measure the remaining water by pouring it into the graduated cylinder Record the total amount of water remaining on the final meeting sheet Weigh the macaroni and cheese container and record the weight on the final meeting sheet Transfer the video from the video camera to the computer Plug in the power cord and the USB cord for the video camera Turn on and rotate mode button Open up the video camera on the computer Canon HDD AVCHD BDMV gt Stream Select the latest video MTS and rename ParticipantNumber_Date MTS Copy the file into the Videos folder on the desktop Unplug from the computer Turn off video camera g moaogp Watch the video and record the number of bites taken manually on the sheet Transfer the information from the final meeti
157. included abnormal dichotomous predictor values for the participant e g the only meal eaten with someone else or the only meal eaten at home and high values for continuous predictors e g 96 highest Meal Energy Density value for a participant All 20 multivariate outlier meals were removed from the data set Then correlations among the remaining variables of interest were examined for evidence of multicollinearity r s gt 0 90 Tabachnick amp Fidell 2007 All correlations were 0 50 so no additional evidence of multicollinearity was found Finally homogeneity of variance of the DV Bites was examined using the ratio of the largest participant variance to the smallest participant variance Fmax The variance ratio for bites was 62 47 indicating a severe violation of homogeneity of variance value much higher than 10 Tabachnick amp Fidell 2007 As a result Bites variances were allowed to vary by person or be heterogeneous by using the Compound Symmetry Heterogeneous covariance type when multi level linear modeling analyses were performed Snijders amp Bosker 2011 Data for MLM analysis After outlier removal 4 065 meals remained 95 596 of the original meals Of these remaining meals 3 606 meals had bite counter data 88 796 3 794 meals had Daily Meals Questionnaire responses 93 396 and 3 691 meals had complete ASA24 data 90 8 3 246 meals had both bite counter and ASA24 data 79 9 The number of me
158. increases awareness of eating behavior Participants with incomplete survey responses were also excluded for example skipping the last page of the survey Participants were also excluded from the study if they did not have daily access to an Internet connected computer with at least a 10 inch screen and the ability to install Microsoft Silverlight this was necessary for completion of the dietary recalls Eligible participants were added to a waiting list if no bite counters were available Participants selected for the study were contacted by the researcher to attend an individual orientation meeting Orientation Meeting The protocol for the Orientation meeting is described in detail in Appendix F Upon arrival at the meeting the participant read and signed a Clemson University IRB approved written consent form see Appendix G The experimenter stated that the purpose of the study was to investigate how well a new device the bite counter was able to estimate energy intake during a meal The experimenter emphasized the importance of compliance with daily bite counter use and dietary recalls and confirmed that the participant would be able to complete these tasks for two weeks Then the experimenter measured the participant s height weight body fat percentage hip circumference and waist circumference T The participant was given a bite counter and told how to wear the bite counter during the day how to record bites during a meal and
159. ite size Obesity 19 3 546 551 Burke L E Conroy M B Sereika S M Elci O U Styn M A Acharya S D Glanz K 2011 The effect of electronic self monitoring on weight loss and dietary intake A randomized behavioral weight loss trial Obesity 19 2 338 344 Burke L E Styn M A Glanz K Ewing L J Elci O U Conroy M B Sevick M A 2009 SMART trial A randomized clinical trial of self monitoring in behavioral weight management design and baseline findings Contemporary Clinical Trials 30 540 551 Burke L E Swigart V Turk M W Derro N amp Ewing L J 2009 Experiences of self monitoring Successes and struggles during treatment for weight loss Qualitative Health Research 19 6 815 828 250 Burnett K F Taylor C B amp Agras W S 1985 Ambulatory computer assisted therapy for obesity A new frontier for behavior therapy Journal of Consulting and Clinical Psychology 53 5 698 703 Butryn M L Phelan S Hill J O amp Wing R R 2007 Consistent self monitoring of weight A key component of successful weight loss maintenance Obesity 15 12 3091 3096 Cardello A V Schutz H G Lesher L L amp Merrill E 2005 Development and testing of a labeled magnitude scale of perceived satiety Appetite 44 1 13 Carels R A Young K M Coit C Clayton A M Spencer A amp Hobbs M 2008 Can following the caloric res
160. ites indicating that this effect was very small The slopes between location and bites were 79 and 1 03 for the meal level models for the full sample and the outliers removed sample respectively Although in the expected direction these slopes were not significantly different from zero The slope between location and bites was 1 85 for the meal level for the outliers removed sample and indicated that when this sample ate a meal outside of the home they took about 2 additional bites during the meal compared to eating a meal at home This translates into consuming about 50 additional kilocalories when eating outside of the home compared to eating at home However the day level model slope of 0 95 between location and bites was not significantly different from zero Location was a non significant predictor in the model with Bite Size that included only 60 participants 166 Taken together these results suggest that people may take a few more bites when they eat meals outside of the home which may be an indicator of increased energy intake during these meals and larger portion sizes available when eating outside of the home e g Condrasky et al 2007 de Castro et al 1990 However location was not a very strong or reliable predictor of bites across models Therefore individuals using the bite counter could be made aware of a tendency to take more bites outside of the home and they could watch for this pattern in their personal bite count
161. ites slope y12 kilocalories x body weight interaction y13 kilocalories x BMI interaction y14 kilocalories x height interaction p lt 05 116 Model 17 Can the varying Kilocalorie Bite slopes be explained by Body Weight First although Body Weight did not explain bites directly it was possible that Body Weight might have been an individual difference that could explain some of the random Kilocalorie Bite slope variance A Body Weight fixed effect at level 2 anda Body Weight x Kilocalorie interaction term were added to create Model 17 The X deviance difference test comparing model 17 to model 11 28312 93 28310 25 2 68 df 14 12 2 p gt 05 indicated that the addition of the Body Weight fixed effect and the Body Weight x Kilocalorie interaction did not significantly improve model fit Random slope variance was not reduced 00041 00041 0 indicating that Body Weight did not explain any of the random Kilocalorie Bite slope variance Finally the Body Weight x Kilocalorie interaction term 0 00008 was non significant Therefore the Body Weight fixed effect and the Body Weight x Kilocalorie interaction term were dropped from further exploratory models Model 18 Can the varying Kilocalorie Bite slopes be explained by BMI It was thought that BMI the ratio of a participant s weight to their height might be an individual difference variable that could explain some of the random Kilocalorie Bite slope variance A BMI fixe
162. kcals g very low energy density about 20 kilocalories were consumed per bite When energy density was one standard deviation above its mean 2 18 kcals g medium energy density about 33 kilocalories were consumed per bite The strength of this interaction was reduced for the day level model for the full sample it was eliminated 163 in the day level model for the outliers removed sample but it remained the same in the day level model that included Bite Size as a predictor These results indicate that if individuals use the bite counter to monitor energy intake in the future at the meal level the energy density of the meal should be considered A smaller kilocalorie multiplier could be applied to meals with lower energy densities and a larger kilocalorie multiplier could be applied to meals with higher energy densities Rolls 2007 four categories of energy density could serve as a guide for future bite counter features For example a participant could enter 1 through 4 into the bite counter to indicate the energy density of the meal and the appropriate multiplier could then be applied However if an individual is going to use the bite counter to monitor energy intake at the day level or higher then the variability in energy density might be reduced such that it would have a smaller impact on the relationship between kilocalories and bites In this case the user could continue to input the energy density of the meal to improve overall
163. ked Disliked somewhat O Disliked very much O Extremely disliked What did you like or dislike about completing the 24 hour dietary recall In the past two weeks did you have any problems using the 24 hour dietary recall Yes O No Please describe any problems you had with the 24 hour dietary recall Did you feel that completing the 24 hour dietary recall changed your eating behavior Yes O No How did you feel the 24 hour dietary recall changed your eating behavior Did you record your dietary intake anywhere other than the Internet based ASA24 system Yes No If you did record your intake in another way please explain how you recorded your intake 210 15 In the past two weeks how often did you wear the bite counter Select the option that most applies O All day everyday from morning to evening O Only part of the day more often than just meal times Only during meal times the other times I took it off I did not wear it during some meals O I did not wear it during many meals O I did not wear it for one or more days 16 In the past two weeks how easy or difficult did you find it to use the bite counter O Extremely easy O Very easy Somewhat easy Neither easy nor difficult Somewhat difficult Very difficult Extremely difficult
164. l it is possible that eating more kilocalories will be associated with taking more bites of the meal For example if an individual takes 15 bites to eat 300 kilocalories of a sandwich we could predict that it might take 5 more bites to eat 100 kilocalories assuming that bite size stays relatively constant This prediction is supported by preliminary analyses from our research group Across 38 meals bite count and kilocalories at the meal level were positively related r 723 p 05 However there is some research to suggest that when an individual eats more of the same food larger bites are taken and the number of bites does not increase In a within subjects laboratory study Burger Fisher and Johnson 2011 found that when adult participants ate 220 more kilocalories of a pasta entr e they did not take significantly more bites This increase in food consumption was explained by the participants taking 47 larger bites Similarly Fisher Rolls and Birch 2003 found that when children ate 25 more food at lunch they did not take significantly more bites Again this increase in food consumption was explained by an increase in bite size Also Mishra Mishra and Masters 2012 used fork size as a proxy for bite size and found that restaurant patrons ate more food with smaller forks compared to larger forks and lab participants ate more from larger forks compared to smaller forks The authors attributed this result to the presence o
165. l Activity 3 17 Ledikwe J H Ello Martin J A amp Rolls B J 2005 Portion sizes and the obesity epidemic The Journal of Nutrition 135 905 909 Levitsky D A Garay J Nausbaum M Neighbors L amp DellaValle D M 2006 Monitoring weight daily blocks the freshman weight gain A model for combating the epidemic of obesity International Journal of Obesity 30 1003 1010 Linde J A Jeffery R W French S A Pronk N P amp Boyle R G 2005 Self weighing in weight gain prevention and weight loss trials Annals of Behavioral Medicine 30 3 210 216 Lombard D N Lombard T N amp Winett R A 1995 Walking to meet health guidelines The effect of prompting frequency and prompt structure Health Psychology 14 2 164 170 Lovasi G S Hutson M A Guerra M amp Neckerman K M 2009 Built environments and obesity in disadvantaged populations Epidemiologic Reviews 31 7 20 doi 10 1093 epirev mxp005 Lowry R Galuska D A Fulton J E Wechsler H Kann L amp Collins J L 2000 Physical activity food choice and weight management goals and practices among US college students American Journal of Preventive Medicine 18 1 18 27 258 Lumeng J C amp Hillman K H 2007 Eating in larger groups increases food consumption Archives of Disease in Childhood 92 384 387 Malnick S D H amp Knobler H 2006 The medical complications of obesit
166. l eat a meal in the laboratory Please bring your bite counter USB cord and charger with you to this meeting to return them Please do not eat or drink anything other than water for at least two hours prior to this meeting Thanks Jenna Scisco Department of Psychology Clemson University 864 656 1144 Day of meeting amp meal 1 Check food allergies to see if a special meal is needed 2 Turn on the desktop computer a b C Check the IntertiaCube3 by double clicking the Blue T indicator on the right of the Windows Taskbar The InterSense Server should show that the IntertiaCube3 is operational There will be a green circle and the yaw pitch and roll will be responsive to sensor movement Look in the C Jenna folder and make sure there are no Original Data Bite Detect or Human Detect data files If there are rename and move them Put a stop watch next to the computer 3 Setup the video camera a b Put the camera in the tripod stand It can be plugged in or unplugged if the battery indicator is full Make sure the camera is positioned so that you can see as much of the area where the participant will be sitting as possible 230 c Turn off video camera Set up the food scale a Pull back the tablecloth b Turn the scale on Wait until the scale reads 0 0g c Put an empty plate on top of the scale Make sure it is centered and not touching any wood Wait a few seconds for the weight
167. le 1 2 3 4 5 1 Bites 2 Kilocalories 0 45 3 Energy Density 0 14 0 07 4 Location 0 05 0 08 0 03 5 Social 0 25 0 30 0 02 0 11 6 Intake Day 0 01 0 08 0 01 0 16 0 18 Note p lt 0 05 Location coded 0 Home Not at Home Social coded 0 Alone 1 With Others Intake Day coded 0 Weekday 1 Weekend Total correlations are presented in Table 3 4 for all level 1 and level 2 variables These total correlations represent the relationships for the complete meal level data set without taking into account within participant nesting Snijders amp Bosker 2011 Correlations between level 1 predictors and Bites remained similar in size and direction compared to the within participant correlations The level 2 variables Gender Body Weight BMI and Height were not related to the number of bites taken during a meal 100 Table 3 4 Total correlations between level 1 and level 2 variables Variable 1 2 3 4 5 6 7 8 9 1 Bites 2 Kilocalories 0 39 3 Energy Density 0 14 0 09 4 Location 0 04 0 07 0 06 5 Social 0 23 0 29 0 01 0 12 6 Intake Day 0 01 0 06 0 01 0 15 0 17 7 Gender 0 02 0 30 0 00 0 03 0 02 0 00 8 Body weight 0 01 0 22 0 04 0 04 0 05 0 01 0 47 9 BMI 0 00 0 14 0 07 0 01 0 04 0 01 0 19 0 91 10 Height 0 00 0 26 0 04 0 07 0 04 0 00 0 72 0 48 0 07 Note p 0 05 Location coded 0
168. lease report all problems you experience with the bite counter This will help the researchers troubleshoot bite counter problems for you How are the ASA24 dietary recall daily meals survey and bite counter data linked Researchers will link these three using your unique participant ID number and the TIME of the meal Because time is so important please enter the meal times into the ASA24 dietary recall and the daily meals survey as accurately as possible 226 Appendix J Appointment Slip You re scheduled for two more Bite Counter meetings Please come to Brackett Hall room 422 on at AM PM and at AM PM Please bring your Bite Counter USB cord and charger to both meetings A meal will be provided for you to eat at the last meeting Please refrain from eating or drinking anything other than water for at least 2 hours prior to this last meeting Questions Contact Jenna Scisco E mail jscisco clemson edu or call 864 656 1144 227 Appendix K Data Download Meeting Protocol One day before meeting 1 Send participant a reminder e mail Dear name This is a reminder that we will have our first bite counter data download meeting on date at time in Brackett Hall room 422 Please bring your bite counter USB cord and charger to this meeting The meeting will last approximately 15 minutes Thanks Jenna Scisco Department of Psychology Clemson University 864 656 1144 Day of meeting 1 Add Da
169. lf weighing is often correlated with tracking food intake and physical activity VanWormer et al 2008 It is possible that self monitoring of physical activity and food intake has unique utility for an individual trying to lose weight or maintain a weight loss By tracking the specific behaviors that impact weight changes the individual may 16 begin to understand the patterns of physical activity and food intake that result in weight loss or weight maintenance Self Monitoring of Food Intake and or Physical Activity Early studies of self monitoring of food intake and physical activity revealed that tracking eating behaviors keeping a paper and pencil food diary and entering food intake and exercise into a computer are related to weight loss Burnett Taylor amp Agras 1985 Fujimoto et al 1992 Sperduto Thompson amp O Brien 1986 As a next step researchers investigated how consistency of self monitoring affects weight loss efforts A series of self monitoring intervention studies had participants record their eating behaviors food intake and physical activity using a paper and pencil self monitoring booklet and found that more frequent self monitoring is related to greater weight loss Baker amp Kirschenbaum 1993 Baker amp Kirschenbaum 1998 Boutelle amp Kirschenbaum 1998 Boutelle and Kirschenbaum 1998 suggested self monitoring all foods eaten on 75 or more of days in order to successfully lose weight The con
170. light version 4 0 or the ability to install this program O Yes O No O I don t know 196 10 11 12 13 Have you ever been diagnosed with an eating disorder e g Anorexia Bulimia O Yes No What hand do you use most often for eating a meal For example what hand do you use most often for eating with a fork Right hand Left hand What is your height in feet and inches Feet _ Inches What is your weight in pounds pounds Please indicate the normal or typical time at which you eat the following meals during a weekday If you do not eat one of more of these meals during a weekday please enter 00 00AM for that meal s time HH MM AM PM Breakfast Morning snack Lunch Afternoon snack Dinner Evening snack 197 14 15 16 17 Other Please indicate the normal or typical time at which you eat the following meals during a weekend If you do not eat one of more of these meals during a weekend please enter 00 00AM for that meal s time HH MM AM PM Breakfast Morning snack Lunch Afternoon snack Dinner Evening snack Other Are you currently trying to lose weight O Yes 0 No Are you currently trying to gain weight L Yes O No Do you have any food allergies Yes O No 198 If yes please list the foods you are allergic
171. ll that apply C O Other please specify Fork Knife Spoon Chopsticks Hands 205 15 How hungry were you before you ate this meal CI CI CI 16 How full CI C Not hungry at all O Somewhat hungry Moderately hungry Very hungry Extremely hungry were you after you ate this meal Not full at all LI Somewhat full Moderately full Very full Extremely full 17 How much did you like your meal in terms of its taste I did not like it at all I liked it somewhat O I liked it moderately O I liked it very much O I liked it extremely 18 How many people did you eat with during this meal If you ate alone enter zero 206 19 Who prepared this meal Select all that apply O I prepared the meal 1 A family member prepared the meal _ A friend prepared the meal A restaurant cafeteria grocery store or other location prepared the meal After answering all of the above questions for each meal the participant will be asked 20 How physically active were you yesterday O I was sedentary O I was somewhat active I was moderately active I was very active O0 Iwas extremely active 207 Appendix D Usability Questionnaire Please enter your unique participant ID provided by the researcher If you do not remember your participant ID plea
172. localorie estimates more if they know that the device has been calibrated to them Bite size may be one very important key to a bite counter that can accurately estimate kilocalories consumed during meals Fifth future research should explore adding an energy density feature to the bite counter in order to adjust kilocalorie estimates to the energy density of the meal being eaten There are numerous research questions in this area as well It is unknown if people can accurately estimate the energy densities of meals Meals are sometimes comprised of many different foods and beverages making energy density estimates potentially very difficult The heuristics that could be used to guide energy density judgments should be identified and tested The Volumetrics categories Rolls 2007 may 190 be appropriate or there might be different categories that could be applied to overall meal judgments Accurate meal energy density input from the user may be another key to a bite counter that can accurately estimate kilocalories consumed during meals The Future Bite Counter The future goal of the bite counter is to be a device that can not only count bites but also can count kilocalories during a meal Based on the main findings from this research energy density and bite size are two features that should be implemented into a future bite counter in order to provide a user with more accurate kilocalorie estimates A future bite counter is imagined as a d
173. low Energy Density meals Model 8 Do Social and Intake Day interact to predict Bites An interaction between Social and Intake Day was added to the model in order to address research question 8 Does the relationship between number of people an individual eats with and bite count depend on whether it is a weekend or a weekday First the change in model fit was assessed by comparing model 8 to model 7 Results of the y2 deviance difference test 28497 60 28495 70 1 9 df 10 9 1 p gt 05 indicated that the addition of the Social X Intake Day interaction did not improve model 109 fit Next the change in within participants variance from model 7 to model 8 was examined The Social X Intake Day interaction explained an additional 0 0007 412 27 411 98 412 27 100 of the within participants variance Finally the Social X Intake Day interaction term was non significant 2 38 Because the Social X Intake Day interaction did not improve the model or its interpretation it was dropped from subsequent models Model 9 Does Gender predict Bites Gender was added to the model as a fixed effect at level 2 in order to address research question 9 Does gender predict bite count First the change in model fit was assessed by comparing model 9 to model 7 Model 9 was not compared to model 8 because model 8 was not a significant improvement over model 7 and its interaction term was dropped from subsequent models Results of the a deviance d
174. meals or on after the conclusion of meals may have been incorrect Any of these possibilities could have led to under or overestimation of bite counts Some participants also found the device uncomfortable and unattractive and chose not to wear it during the day This could have led to forgetting to use the device to record meals Future device design improvements should make the bite counter more attractive and comfortable for daily use This would help participants to remember to wear the bite counter and record their meals with the device Additionally research on the ability to 182 automatically detect eating should continue as this could potentially eliminate the need for participants to activate and deactivate the device Dong et al under review However any recording errors associated with detecting meals automatically should be less severe than the recording errors associated with participants forgetting to turn the device on and off in order for automatic detection of meals to improve device accuracy This is another area for future research Study Strengths Large Sample Size and Success of Data Collection Overall data collection efforts were successful This was one of the first studies to collect eating behavior data from naturalistic settings with the bite counter The study required a significant time commitment by participants who used the bite counter for 14 consecutive days while spending up to an hour each day completi
175. members were recruited using flyers hung in Fike Recreation Center community centers fitness centers libraries and coffee shops Study announcements were put on the Clemson psychology 61 department webpage the Applied Psychophysiology Lab webpage and the Bite Technologies Facebook page All participants received 50 for two weeks of participation 25 for less than 2 weeks of participation drop outs and a free data summary The data summary included foods kilocalories bites and average kilocalories per bite for each meal reported The data summary was e mailed as a Microsoft Excel file to the participant within four weeks after completing the study Sample Characteristics The present study recruited and selected a representative sample of participants based on gender BMI and age Demographic statistics for Clemson University surrounding counties South Carolina and the US were gathered to guide recruitment and selection and these are described in Table 2 1 Based on these demographic statistics the present study aimed to recruit about 50 females and 50 males between the ages of 18 and 64 and to represent overweight and obesity trends 62 Table 2 1 Demographic statistics used to guide sample recruitment and selection Location Gender BMI Age Clemson 46 female University students 49 female employees Pickens 50 1 female County Oconee 50 8 female County Anderson 51 7 female C
176. model for a total of 60 participants and 2 388 meals with matching data BiteCD251 was removed from this model because his average Bite Size was an outlying case 41 kcals bite Correlations among variables of interest are provided in tables 3 20 and 3 21 It can be seen that Bite Size and Bites are negatively correlated r 0 10 p lt 0 05 indicating that participants with a larger average Bite Size may take fewer Bites during meals Additionally Bite Size and Height are positively correlated r 0 28 p lt 0 05 suggesting that taller participants take larger bites Table 3 20 Within participant correlations between level 1 variables for bite size model with 60 participants Variable 1 2 3 4 5 1 Bites 2 Kilocalories 0 50 3 Energy Density 0 14 0 05 4 Location 0 06 0 10 0 03 5 Social 0 29 0 32 0 04 0 12 6 Intake Day 0 03 0 08 0 01 0 15 0 18 Note p 0 05 Location coded 0 Home Not at Home Social coded 0 Alone 1 With Others Intake Day coded 0 Weekday 1 Weekend 141 Table 3 21 Total correlations between level 1 and level 2 variables for the bite size model with 60 participants Variable 1 2 3 4 5 6 T 8 9 10 1 Bites s 2 Kilocalories 0 46 3 Energy Density 0 14 0 05 4 Location 0 08 0 090 0 08 5 Social O30 032 0 03 0 14 6 Intake Day 0 04 0 07 0 01 0 15 0 17 7 Gender 0 06 0 29 0 04 0 04
177. mparing model 12 to model 11 28312 93 28497 86 184 93 df 13 12 1 p 05 indicated that the addition of the random Energy Density Bites slope variance significantly harmed the model fit Additionally the remaining model estimates were unstable because the Hessian matrix was not positive definite Therefore the random Energy Density Bites slope variance was dropped from subsequent models Model 13 Does the Relationship between Location and Bites vary by participant In model 13 the relationship between Location and Bites was allowed to vary by participant random Location Bites slope variance The y deviance difference test comparing model 13 to model 11 28312 93 28308 11 4 82 df 13 122 1 p lt 05 indicated that the addition of the random Location Bites slope variance significantly improved the model fit However the random Location Bites slope variance 18 90 did not significantly differ by participant Due to the small increase in model fit but non 112 significant slope variation the random Location Bites slope variance was dropped from subsequent models Model 14 Does the Relationship between Social and Bites vary by participant In model 14 the relationship between Social and Bites was allowed to vary by participant random Social Bites slope variance The y deviance difference test comparing model 14 to model 11 28312 93 28305 56 7 37 df 13 12 1 p 05 indicated that the addition of the r
178. mplicity no subscripts will be used for n in this description The goal of the present study with important predictors at both levels of analysis was to maximize all three samples to provide enough power for the analysis As a rule of thumb Bosker et al 2003 suggest that n should be at least 6 and N should be at least 10 A total sample size of 60 is also suggested by Tabachnick and Fidell 2007 when only 5 or fewer parameters are being estimated Hox 2010 suggests a larger sample size of n 30 and N 30 when most interested in the fixed parameters and n 20 and N 50 when there is strong interest in cross level interactions The present study operated under both equipment and time constraints Both of these costs were considered when choosing sample size because decisions of sample size frequently involve decisions about optimal and feasible study design Hox 2010 It was assumed that participants would record three meals per day on average In order to appropriately power the analysis at both levels with samples sizes of at least 30 at each 59 level Hox 2010 and to maximize the total sample size data was collected from a minimum of 80 participants and each participant recorded bite count dietary recalls and additional measures for 2 weeks which was predicted to provide an average of 42 total meals per person To check this sample size decision against the ability to detect an expected effect size the predicted correlation
179. mproved model fit and was significantly greater than would be expected by chance as assessed by the Wald Z test of significance was retained in the model Hox 2010 Heterogeneity of variance was allowed by specifying a specific covariance type for estimates of random effects Compound Symmetry Heterogeneous Cross level interaction terms were then added to the model to examine reduction in random slope variance in addition to change in model fit and significance of cross level interaction terms 91 CHAPTER THREE RESULTS Original Data After error removal the total number of meals reported across all participants was 4 256 Of these meals 3 767 meals had bite counter data 88 5 3 976 meals had Daily Meals Questionnaire responses 93 4 and 3 882 meals had ASA24 data 91 2 3 406 meals had both bite counter and ASA24 data 80 0 3 346 meals had complete data from all three sources 78 6 MLM Analysis Data Cleaning Data for the primary variables of interest were inspected for correct values outliers normality linearity homogeneity of variance and multicollinearity First the five level 1 continuous variables Bites Meal Kilocalories Meal Duration Number of People and Meal Energy Density were inspected for appropriate means minimum values maximum values skewness kurtosis and univariate outliers within each of the 83 participants Tabachnick amp Fidell 2007 Boxplots histograms and expected normal probabilit
180. n line with full sample findings However in Model 19 the 130 direct effect of Height was significant in addition to the Height by Kilocalories cross level interaction 131 Table 3 14 Estimates of model fit and random effects for the outliers removed model Model fit Random effects j x ZEE SE HOUSE EUM ee Tm Sd eae 1 3 25288 71 559 38 15 30 154 31 28 97 2 4 24403 35 400 67 10 96 175 03 31 92 3 5 24326 24 389 67 10 66 168 02 30 68 4 6 24317 73 388 47 10 63 167 44 30 59 5 7 24259 93 380 48 10 41 161 86 29 58 6 8 24257 40 380 48 10 40 162 13 29 62 7 8 24222 20 375 23 10 27 160 50 29 32 8 10 24220 06 374 91 10 26 160 84 29 38 9 24218 62 375 22 10 27 152 06 27 84 10 24218 76 375 21 10 27 152 51 27 90 11 10 24060 94 347 34 9 61 151 43 27 87 0004 001 12 11 24210 11 372 13 10 22 160 66 29 34 a 13 11 24056 47 346 29 9 60 153 44 28 22 0004 001 4 26 4 15 14 11 24058 54 346 29 9 660 150 88 27 87 0003 001 5 17 8 09 15 12 24059 41 2347 09 9 61 154 23 28 53 0004 001 52 1 37 16 12 24056 79 347 34 9 61 143 38 26 45 0004 001 Note Model 12 failed to converge 2LL 2 log likelihood SE Standard Error e residual within participant variance 100 random intercept between participants variance 110 random slope variance Significant model improvement from previo
181. nal hours 36 4 Scott 4 Greg 3 5 3 5 amp amp o 3 _ e o 3 9 s E 2 5 Spe t gt 2 5 EN i A Slope 0 08 T 2 2 A gt 0 1 2 3 4 0 1 2 3 4 Job Status Job Status 4 X Liz 4 Ann 4 3 5 j 3 5 S Slope 0 16 X Emm lO Slope 0 08 3 3 gt E c 2 5 4 2 5 4 2 4 2 0 1 2 3 4 0 1 2 3 4 Job Status Job Status 4 Kate 3 5 3 _ 23 si 0 16 gt eo ope 0 a 9 2 _ ipa 0 1 2 3 B Job Status Figure 1 11 Student scatterplots demonstrating individual differences in the relationship between job status and GPA Q7 Can the student level variation in the relationship between job status and GPA be explained by gender Examining Figure 1 11 it can now be asked if the differences in the relationships between GPA and job status the slopes can be explained by the gender of the students It can be seen that Scott and Greg the two males have positive slopes However Liz Ann and Kate the three females have negative slopes Therefore it seems that gender 37 can explain some of the variation in the student level relationships between job status and GPA Male GPAs increase when they work more hours per day and female GPAs decrease when they work more hours per day The difference between Q3 and Q7 is that in Q3 the slopes were originally grouped by gender however in Q7 the slopes were first allowed to vary by studen
182. nd carbohydrates Yao amp Roberts 2001 Increasing the percentage of low energy density foods eaten is an eating strategy that may aid weight loss due to the increased volume of food consumed and the decreased caloric content of that food Rolls 2007 Diet diary research has found a positive relationship between the energy density of a meal and the amount of food consumed r 0 26 0 30 de Castro 2004a de Castro 2004b de Castro 2005 Reviews of studies that provided foods of varying energy density to individuals have concluded that consumption of low energy density diets is associated with reduced energy intake and comparable levels of satiety Prentice 1998 Yao amp Roberts 2001 Laboratory studies that manipulate energy density have found that increasing the energy density of a food increases the kilocalories of food consumed because individuals tend to consume a similar weight or volume of the same food across meals Bell Castellanos Pelkman Thorwart amp Rolls 1998 Bell amp Rolls 2001 The relationship between the energy density of a meal and the number of bites taken at a meal is unknown because there is no published research on the relationship between these two variables The relationship between the energy density of a meal and the number of bites taken at a meal may not follow the pattern of results that has been uncovered by the energy density and kilocalorie research That is there may not be a positive relation
183. nd the people that we eat with often reflect our social relationships Sobal amp Nelson 2003 As the number of people an individual eats with increases energy intake also increases a finding often referred to as the social facilitation of food intake Herman Roth amp Polivy 2003 This finding has been supported by 7 day diary studies by de Castro and colleagues that asked individuals to record detailed information about each meal including the number of people present de Castro and de Castro 1989 found that meals eaten alone contained about 180 fewer kilocalories than meal eaten with others Additionally the overall correlation between number of people and meal size r 0 418 indicated that 17 5 of the variance in meal size could be explained by the number of people present at the meal de Castro amp de Castro 1989 This strong positive correlation between number of people and meal size is still present after controlling for time of day meal location snacks and alcohol intake de Castro Brewer Elmore amp Orozco 1990 Analyses of over 3 800 meals have indicated that meals eaten in large groups are over 75 larger than meals eaten alone de Castro amp Brewer 1991 Interestingly it appears that social facilitation is a strong predictor of meal size but not of overall intake for an entire day de Castro 1996 The positive relationship between number of people present at a meal and energy intake has also been supported
184. ne s skeletal frame and possibly increased mouth volume Human body size measurements including height and body surface area m have been found to be associated with larger bite sizes in a laboratory setting Hill amp McCutcheon 1984 The relationship between body size and bite size in animals has also been investigated as it could have important implications for species fitness Cope Loonen Rowcliffe and Pettifor 2005 found that geese with longer bills had larger bite sizes over a range of grass heights and that bite size was proportional to body mass to the power 2 99 Wilson and Kerley 2003 found that larger animals such as the rhinoceros had larger bite sizes over a range of plants than smaller animals such as goats although differences between animals of similar size depended on the type of plant being consumed 173 To test the hypothesis that height might be a proxy for bite size bite size was added as a predictor to a model with 60 participants who had average bite sizes from the lab meal When bite size was added height was no longer a significant predictor of bites and it no longer moderated the relationship between kilocalories and bites This suggests that when controlling for bite size height does not provide any additional predictive power for the number of bites taken during a meal Therefore in the absence of a bite size measurement height is an individual difference variable that could be used to calibrate the kil
185. ng sheet to the corresponding excel spreadsheet 234 Appendix M Satiety Labeled Intensity Magnitude SLIM Scale Please rate the degree of hunger fullness that you currently feel by putting a slash mark somewhere on the line below Greatest Imaginable Fullness Extremely Full Very Full Moderately Full Slightly Full Neither Hungry nor Full Slightly Hungry Moderately Hungry Very Hungry Extremely Hungry Greatest Imaginable Hunger 235 Appendix N Labeled Affective Magnitude LAM Scale How much did you like the macaroni and cheese Please put a slash mark somewhere on the line below Greatest Imaginable Like Like Extremely Like Very Much Like Moderately Like Slightly Neither Like Nor Dislike Dislike Slightly Dislike Moderately Dislike Very Much Dislike Extremely Greatest Imaginable Dislike 236 Appendix O Data Merging and Error Screening Steps Step 1 Merge data for the meals 1 In the Dissertation Data Merged and screened data folder create a new folder named ParticipantID In the ParticipantID folder create a new Excel workbook named ParticipantID xls All of the raw meal data is imported into this file a Name this first sheet Merged Data b Name the second sheet INF c Name the third sheet Removed In the ParticipantID folder create a new Word document named ParticipantID data merging and screening history docx and save in the Merged and screened
186. ng the ASA24 dietary recall and the Daily Meals Questionnaire Only 11 7 of the participants who began the study withdrew for various reasons 3 190 complete meals across 83 participants were analyzed after outlier meal removal and 2 741 complete meals across 69 participants were analyzed after outlier participant removal This large sample size provided sufficient power for the MLM analyses conducted Hox 2010 Data collection was successful due to a combination of factors Wide advertisement to students university employees and community members attracted over 260 interested participants The 50 compensation seemed to be an adequate motivator for some participants However many participants expressed greater interest in receiving 183 their data summary with details about how many kilocalories they were eating and how many bites they were taking Future studies with ambulatory bite counters should continue to provide data summaries to participants as this seems to be a strong motivating factor Additionally the participants were given in depth instructions during a one hour orientation meeting reminded to begin using their bite counter on the start date sent daily e mails with links to the ASA24 recall website and the Daily Meals Questionnaire website and sent reminders to attend the data download meeting and the final meeting These factors held the participants accountable for their participation in the study and also made it easier fo
187. ns described below e g de Castro amp Plunkett 2002 Given the large number of parameters that need to be estimated when using a multi level design it is recommended that the model remain reasonably small Hox 2010 p 33 Therefore only those predictors that are thought to have the strongest possible relationship with bite count and that are most theoretically meaningful were examined in this study Because up to 86 of the variance in food intake is due to environmental factors many of the predictors are environmental in nature de Castro 2010 46 Meal level Predictors of Bite Count Meal level predictors are variables that could affect meal level bite count Total number of kilocalories The first meal level predictor to be examined is the total number of kilocalories consumed during the meal Arguably the relationship between kilocalories and bites is the most important relationship to understand for the bite counter project The current standard for measuring energy intake is the kilocalorie the quantity of heat necessary to raise the temperature of 1 kg 1 L of water 1 C McArdle Katch amp Katch 2005 The kilocalorie is more commonly referred to as a calorie on food packages and labeling In order for the bite counter to be understood and well accepted by the weight loss community as a measure of energy intake it should provide a reasonable estimate of the number of kilocalories consumed Within an individual mea
188. nships between the IVs and the DV 30 Table 1 4 Data for MLM example Student Year GPA _ JobStatus Gender Scott 2007 2 5 0 Male Scott 2008 2 6 1 Male Scott 2009 2 9 4 Male Scott 2010 2 7 2 Male Scott 2011 2 8 3 Male Greg 2007 2 4 3 Male Greg 2008 2 5 4 Male Greg 2009 2 1 0 Male Greg 2010 2 2 1 Male Greg 2011 2 3 2 Male Kate 2007 3 1 0 Female Kate 2008 2 9 1 Female Kate 2009 2 3 4 Female Kate 2010 2 5 3 Female Kate 2011 2 6 2 Female Liz 2007 3 2 4 Female Liz 2008 3 8 1 Female Liz 2009 3 6 2 Female Liz 2010 3 4 3 Female Liz 2011 4 0 Female Ann 2007 3 5 0 Female Ann 2008 3 4 1 Female Ann 2009 3 3 2 Female Ann 2010 3 2 3 Female Ann 2011 3 1 4 Female QI Does job status predict GPA Figure 1 7 shows all of the GPA measurements for all students and all years with GPA on the y axis and job status on the x axis Given this plot the first question that can be asked of the data set is does job status predict GPA As seen in Figure 1 7 the overall effect of job status on GPA is slightly negative As the number of hours worked per day increases GPA decreases 3l 4 3 8 3 6 34 4 3 2 4 3 4 2 8 4 2 6 4 2 4 4 2 2 2 GPA oF 99 eo 999 e 9 4 1 2 3 Job Status Figure 1 7 Relationship between job status and GPA Q2 Does gender predict GPA Figure 1 8 shows all of the GPA measurements for all students and all years with GPA on the y axis and gender on
189. nt Kilocalories Research question investigated if kilocalories could predict bite count Kilocalories were found to explain the most variance in bite count 21 4 of within participants variance was explained for the full sample model and 28 4 of within participants variance was explained for the outliers removed model Average within participant correlations were 0 45 and 0 51 for the two models and total correlations across all meals were 0 39 and 0 46 for the two models indicating that taking more bites was associated with greater energy intake The slope between Kilocalories and Bites held reliably at 0 04 throughout model building at the meal level and the day level with the exception of a slope of 0 03 at the day level for the full sample This translated to an average of 25 kilocalories per bite across all meals Practically this could translate to using the bite counter as a calorie counter with bites multiplied by 25 to create kilocalorie feedback during meals 160 It is important to acknowledge that this relationship between kilocalories and bites was moderated by energy density height and bite size as will be discussed below A simple kilocalorie multiplier may work well when averaged across all meals for all people but it may be important to consider features of the meals and individual differences before using this kilocalorie multiplier and providing feedback at the meal level Additionally the relationship between kil
190. ntact and Online Prescreening Protocol 2 4 F Orientation Protocol uii ei e Eco eel is idit dee 218 G Written Consent POIs o dicunt Disk trees cee Aaa due ot eot oe 222 He Bite Counter Instt clIOngss acs aacsserncaud dans o eati oer Ep need olere aos 224 I ASA24 Dietary Recall and Daily Meals Survey Instructions 226 J Xppomntiment SUP sess ccarescrsoceassoansadet sacar tei Meri eoe bep obse prete geo eeaies 227 K Data Download Meeting Protocol eeeeeeeeeeeenneeene 228 L Final Meeting and Meal Protocol ete eant o I ieu aet ope RRRRARR FERRE 230 M Satiety Labeled Intensity Magnitude SLIM Scale 235 N Labeled Affective Magnitude LAM Scale sees 236 O Data Merging and Error Screening Steps sese 237 P Description of Data Quality for Each Participant esses 241 REFERENCES m 249 vi Table 1 1 1 2 1 3 1 4 1 5 1 6 1 7 1 8 2 1 2 2 3 1 3 2 3 3 3 4 3 5 3 6 3 7 LIST OF TABLES Definitions of successful weight loss weight maintenance and weight fluctuation iesaistei sk atesi isi Ea EA need eS AE SKLE EIEE 5 Factors that may impact adherence to self monitoring 19 Within and between person bite count variance examples 26 Data for MLM example
191. o model 2 Results of the 3 deviance difference test 28722 97 28617 07 105 90 df 5 4 1 p lt 05 indicated that the addition of Energy Density significantly improved model fit Next the change in within participants variance from model 2 to 105 model 3 was examined Energy Density explained an additional 3 3 442 76 428 28 442 76 100 of the within participants variance Lastly a significant negative relationship between Energy Density and Bites was observed in the Energy Density Bites slope of 4 28 Each 1 kcal gram increase in Energy Density corresponded on average to a 4 28 decrease in number of Bites Thus the bite counter recorded fewer bites when participants ate mote energy dense meals Energy Density was retained as a level 1 predictor for all subsequent models Model 4 Does Location predict Bites Location was added to the model as a fixed effect at level 1 in order to address research question 5 Does the location of a meal predict the number of bites recorded during a meal First the change in model fit was assessed by comparing model 4 to model 3 Results of the x deviance difference test 28617 07 28611 13 5 94 df 6 5 1 p 05 indicated that the addition of Location significantly improved model fit Next the change in within participants variance from model 3 to model 4 was examined Location explained an additional 0 2 428 28 A427 43 428 28 100 of the within participants variance Lastly a
192. ocalorie setting for the bite counter Shorter participants could receive a smaller kilocalorie multiplier and taller participants could receive a larger kilocalorie multiplier This suggestion should be taken with caution however noting that a bite size measurement may be a better way to calibrate the bite counter kilocalorie setting as discussed below Bite Size When bite size was entered into a model with 60 participants with bite size measurements bite size was able to explain 24 6 of the between participants variance Every additional kilocalorie per bite increase in bite size was associated with a decrease of about 1 to 2 bites taken per meal on average The interaction between kilocalories and bite size explained 35 22 of the variance in individual relationships between kilocalories and bites The simple slopes indicated that participants with smaller bite sizes ate about 19 kilocalories per bite on average participants with average bite sizes ate about 25 kilocalories per bite on average and participants with larger bite sizes ate about 174 39 kilocalories per bite on average This finding was still significant at the day level participants with smaller bite sizes ate about 20 kilocalories per bite on average participants with average bite sizes ate about 25 kilocalories per bite on average and participants with larger bite sizes ate about 33 kilocalories per bite on average Furthermore as described above the addition of
193. ocalories and bites leaves over 70 of the variance in bites within participants unexplained While additional predictors discussed below help to account for additional variance the final model for the outliers removed sample still had over 50 of the variance in bites unexplained This indicates that there may be other predictors of bites explaining significant meal level variation or there may be error in the measurements obtained by the bite counter or the ASA24 dietary recall as will be discussed in subsequent sections Energy Density Research question 2 investigated if the average energy density of a meal could predict bite count Energy density explained 3 3 of within participants variance in the full sample and 2 7 of within participants variance in the outliers removed sample indicating that it had a much smaller effect on the number of bites taken during a meal compared to kilocalories Within participant and total correlations across both samples were 0 14 indicating the increased energy density was associated with taking fewer bites The slopes between energy density and bites were 5 81 and 5 49 for the meal level models for the full sample and the outliers removed sample respectively This 161 indicated that as the average number of kilocalories per gram per meal increased by 1 participants took about 5 to 6 fewer bites per meal The slopes between energy density and bites were 27 57 and 31 81 for the day level models
194. off during meals removed Meals were very short in duration Some meals BiteCD003 35 71 8 637 Good 14 ieee underestimated removed BiteCD006 36 87 8 agi Due rounds nji Good off once Time drift BiteCD007 51 98 1 480 Sometimes did not 14 Good calibrate right away Participant thought 18 88 was an error and tried to hold down the button to BiteCD011 47 94 0 384 get past calibration 13 Good Device would turn off but participant would eventually get it to stay on BiteCD012 89 92 7 557 BI commented cua Good off twice Nutritional supplement shakes were corrected BiteCD014 59 76 6 451 Good 14 pathway of questions error One underestimated 241 meal removed ID matched matched Kilocalories meals meals Bites correlation Bite Counter problems data quality ASA24 problems data quality ASA24 completed BiteCD015 BiteCD018 BiteCD023 BiteCD025 BiteCD026 BiteCD028 BiteCD029 BiteCD030 BiteCD032 31 41 54 40 45 100 52 28 45 69 0 77 6 636 684 762 426 409 244 767 491 696 Bite counter turned off once Display problems Bite counter turned off during a few meals Time drift Good Bite counter turned off during a few meals Bite counter turned off once First bite counter turned off frequently and had a broken speaker Second bite counter was better but battery lev
195. on errors For 85 example participant BiteCD003 s meal 26 originally had a 35 minute duration one of the longest meals for this participant This meal was associated with 1 111 kcal of bread hummus potatoes chicken and coffee consumed at lunch and 112 bites were recorded at this meal However the participant reported leaving the device on for an extra 15 minutes or 43 of the recorded meal Thus 15 minutes were removed from the bite counter recording resulting in a total duration of 20 minutes and bite count was reduced by 43 for a total bite count of 64 When compared to the existing duration and bite count values this adjusted data appeared to match the data set as can be seen in Figure 2 11 Thus the decision was made to keep the data in its corrected form MeallD 44 3 7E 08 685 50 2011 at 4 19 57 36 11 25 Dinner 8 00 PM 47 3 7E 08 718 10 2011 11 5 14 36 05 11 58 Snack 2 36 PM 14 3 7E 08 729 35 2011 10 28 20 24 13 12 09 Snack 8 30PM 12 3 7E 08 785 33 2011 10 28 10 22 39 13 05 Breakfast 10 15 AM 51 3 7E 08 811 20 2011 11 6 12 47 27 13 31 Snack 12 47 PM 21 3 7E 08 897 17 2011 10 30 15 40 03 14 57 missing data 37 3 7E 08 924 22 2011 11 3 10 16 12 15 24 Breakfast 10 15 AM 64 3 7E 08 982 25 2011 11 8 19 10 46 16 22 Snack 7 10 PM 46 3 7E 08 1041 61 2011 11 5 13 42 10 17 21 Lunch 1 42 PM 58 3 7E 08 1120 80 2011 11 7 22 41 23 18 40 Dinner 10 41 PM 63 3 7E 08 1179 37 2011 11 8 15 09 2
196. onitoring programs that can increase self monitoring adherence Boutelle Kirschenbaum Baker amp Mitchell 1999 Harvey Berino et al 2002 Tate Jackvony amp Wing 2006 Finally individual differences including understanding the importance of self monitoring using one s preferred self monitoring method social support gender being male and race being Caucasian have all been linked to improved self monitoring adherence Burke Swigart Turk Derro amp Ewing 2009 Hollis et al 2008 Shay Seibert Watts Sbrocco amp Pagliara 2009 The one factor that is consistently related to a decreased self monitoring adherence is time e g Carels et al 2008 Polzien Jakicic Tate amp Otto 2007 As time in a weight loss program increases self monitoring behavior tends to decrease 18 Table 1 2 Factors that may impact adherence to self monitoring Self monitoring Program features Individual differences Barriers to self tools monitoring Simplified diaries Human counseling Understanding Time in weight better than automated importance of self loss program monitoring Using a PDA PED Support feedback and Using preferred Access to or mobile phone accountability to a method or tool acceptance of SMS counselor technology Internet technology Reminders to self monitor Food scale Pedometer Packaged meals e g Weight Watchers SlimFast Social support Gender male Race Caucasian
197. ontributed to error in bite counter recordings Furthermore bite counter algorithm development has been limited to laboratory studies under controlled and uncontrolled conditions Dong Hoover Scisco amp Muth 2012 Further bite counter algorithm improvement may be able to reduce the occurrence of false positives and to increase true detections A cafeteria study is currently underway with 300 participants which will provide an exceptionally large database of bites taken in a more naturalistic setting This future database could be used to improve device accuracy over a wide range of wrist motions resulting from eating different foods using different utensils and individual differences in bite behavior It could also be used to answer important questions relevant to the algorithm such as the average time elapsed between bites during meals In addition the present study identified behaviors that participants frequently engaged in while eating such as talking to others using a computer watching TV reading and driving These behaviors could be studied closely in laboratory and naturalistic settings to examine how they impact device accuracy Potential errors in ASA24 reporting by participants or features of the ASA24 that could lead to inaccurate kilocalorie estimates were also previously described Future 186 published work about the validity of the ASA24 for estimating energy intake should be applied to the results of the present study
198. oratory and clinical settings If an individual wanted to reduce their bite size in order to slow their eating rate that is take more bites of a meal of the same size they could use the bite counter to help them do so in their daily lives For example if a participant knows they typically take 30 bites when they eat two slices of pizza they could try taking 60 bites of the same pizza This would slow down their eating rate and 176 if their goal was to eat less as a result it would give them more time to consider feelings of hunger and satiety during the meal and perhaps even become tired or bored of the food being eaten Scisco 2009 Lab Meal Positive correlations between lab meal and real world measures provided support for using measures obtained in the lab to predict eating behavior in the real world Perhaps most relevant to the current study was the finding that bite size in the lab and real world were positively correlated and not significantly different from one another This supports the idea that bite size is consistent within individuals and demonstrates that bite size from a single laboratory meal could be a possible way to calibrate the bite counter in future research The findings that participants ate for a shorter amount of time and ate faster in the lab compared to real world meals indicated that the controlled laboratory environment may have been unnatural for many participants Participants ate alone in silence without
199. ories and Bites the negative relationship between Energy Density and Bites and the negative interaction term between Kilocalories and Energy Density were nearly identical to the previous meal level model for the outliers removed sample Therefore the relationship between Kilocalories and Bites was stronger for meals of lower energy density compared to meals of higher energy density as shown in Figure 3 6 The positive relationship between Social and Bites was also very similar and indicated that participants in this sample took 6 57 more bites on average during meals eaten with others compared to meals eaten alone New to this analysis the positive relationship between Bite Size and Bites indicated that for every 1 kilocalorie per bite increase in individual bite size the average number of bites taken during a meal decreased by about 1 34 bites The addition of Bite Size explained 24 26 of the between participants variance in Bites However there was also a significant interaction between Bite Size and Kilocalories and the addition of this interaction explained 35 22 of the random Kilocalories Bites slope variance Simple slopes were calculated in accordance with Cohen et al 2003 using the fixed effects coefficients at high 1 SD and low 1 SD values of Kilocalories These slopes were significant at low B 0 053 SE 0 003 t 19 72 p 05 moderate B 0 040 SE 0 003 t 16 44 p lt 05 and high B 0 026 SE 0 002
200. orksheet MyDropbox Dissertation Data ParticipantIDinfo a Record the participant s name e mail and phone number 2 Send the interested participant the following e mail Dear name Thank you for your interest in our research study being conducted by the Department of Psychology at Clemson University In order to determine your eligibility for the study please complete the following survey by clicking the link below or copying and pasting it into your web browser address bar https www surveymonkey com s prescreening You will be asked for a participant ID Your unique participant ID is insert 9 letter number ID here If you have any questions you may contact me by e mail at jscisco clemson edu or by phone at 864 656 1144 Sincerely Jenna Scisco Department of Psychology Clemson University 3 Download the Survey Monkey data in Advanced Spreadsheet form and save in MyDropbox Dissertation Data SurveyMonkey Prescreening a Save the ZIP file as Prescreening MonthDDYYY Time b Extract to a folder by the same name c Drag ZIP file into new folder with data d Open CSV file Sheet 1 and check for i History of an eating disorder excluded ii No daily access to an Internet connected computer excluded ili Age gender and BMI status add description to ParticipantIDinfo spreadsheet 4 If the participant is eligible and there are available bite counters schedule the first session by sending the following e mail
201. os explaining how to use the bite counter to detailed one on one instructions and demonstrations with an experimenter This feedback and training could also occur in stages during a study and improvement in the relationship between bites and kilocalories could be assessed over time Future research could also examine a number of these approaches and compare them to each other and to bite counter use 189 without any training or feedback The goal of this research would be to determine what kind of training and feedback if any is necessary to improve the relationship between bites recorded by the device and kilocalories consumed during a meal Fourth future research should examine improvement in the kilocalories to bites relationship when the bite counter is calibrated based on an individual s bite size The research questions in this area are numerous The foods utensils and laboratory settings most appropriate for a calibration meal should be investigated There may be features of a meal experience that could alter bite size and these should be fully understood when designing a calibration meal Investigating the possibility of calibrating at home with an individual s own utensils and foods would have interesting applications for future calibration instructions for devices sold commercially The effect of food energy density on calibration should be investigated It would also be interesting to examine if participants trust bite counter ki
202. ounty United States Undergrad 2 3 underweight 70 77 normal and 20 30 overweight obese 29 4 obese in South Carolina 29 4 obese in South Carolina 29 4 obese in South Carolina 68 overweight includes obese Males 63 5 overweight ages 20 39 77 8 overweight ages 40 59 78 4 overweight ages 60 Females 59 5 overweight ages 20 39 66 3 overweight ages 40 59 68 6 overweight ages 60 20 mean age undergraduates 11 8 ages 20 24 13 3 ages 25 34 14 3 ages 35 44 12 4 ages 45 54 4 8 ages 55 59 4 0 ages 60 64 5 7 ages 20 24 12 8 ages 25 34 14 5 ages 35 44 14 1 ages 45 54 6 4 ages 55 59 5 7 ages 60 64 5 9 ages 20 24 13 5 ages 25 34 15 5 ages 35 44 14 0 ages 45 54 5 7 ages 55 59 4 6 ages 60 64 Note Clemson University Mini Fact Book for 2011 Huang et al 2003 Lowry et al 2000 Fishel Brown 2010 Clemson University College Portrait 2009 U S Census Bureau 2000 census Centers for Disease Control and Prevention Flegal et al 2010 63 Data collection spanned 21 consecutive weeks from October 2011 to February 2012 Ninety four participants started the study Eleven participants dropped out of the study 4 females 7 males an 11 7 drop out rate These participants were not included in any data analyses because they provided no data or because any data provided were of very low quality Reasons participants dropped out of the study were not eno
203. own that individuals tend to eat 18 20 more food on weekends than weekdays by eating larger meals de Castro 1991 Rhodes Cleveland Murayi amp Moshfegh 2007 If larger meals are eaten on weekends it follows that more bites may be detected during weekend meals than weekday meals Research Question 7 Does day of the week predict the number of bites recorded during a meal 55 An interaction between day of the week and number of people eating with is also predicted The positive relationship between number of people and meal size is larger on weekends r 0 4 than weekdays r 0 3 indicating that the social facilitation of food intake may depend on the day of the week the meal is consumed de Castro 1991 That is eating with others may not affect bite count as strongly when the social eating is part of the weekly routine de Castro 1991 Research Question 8 Does the relationship between number of people an individual eats with and bite count depend on whether it is a weekend or a weekday Individual level Predictors of Bite Count Gender On average males need to consume more calories than females due to their larger body size and greater lean body mass McArdle Katch amp Katch 2005 There are also social pressures for men to eat more than women with men desiring larger body types and females desiring a more slender figure Rolls Fedoroff amp Guthrie 1991 Laboratory studies have demonstrated th
204. pendix A Demographics Questionnaire Please enter your unique participant ID provided by the experimenter If you do not remember your participant ID please e mail jscisco clemson edu or call 864 656 1144 to receive your ID What is your age in years years What is your gender O Male Female Whatis your ethnicity optional O American Indian or Alaska Native Asian or Pacific Islander African American O Caucasian O Hispanic O Other please specify Whatlevel of education have you obtained Less than a high school diploma O High school diploma or equivalent O Some college Bachelor s degree 195 Master s degree Doctoral or professional degree PhD MD JD DPharm DPT etc What is your annual household income optional O 0 10 000 O 60 001 70 000 10 001 20 000 70 001 80 000 11 20 001 30 000 80 001 90 000 11 30 001 40 000 90 001 100 000 40 001 50 000 More than 100 000 11 50 001 60 000 How frequently do you use a computer C Never L Once per month O Once per week O A few times per week Daily Do you have DAILY access to a computer with ahigh speed Internet connection such as cable DSL or FIOS a screen size of at least 10 inches and Microsoft Silver
205. quation 1 5 Yo1 Yio Y11 and the two variance components in Table 1 7 Top 145 are the main parameters that are interpreted In Table 1 8 these five parameters and their interpretations are referenced back to questions 1 through 7 and Figures 1 7 through 1 11 42 Table 1 8 Research questions with their corresponding parameter estimates figures and interpretations Number Question Parameter Estimate Corresponding Figure Interpretation QI Does job status predict GPA Q2 Does gender predict GPA Q3 Does the relationship between job status and GPA depend on gender Q4 Does GPA when job status is average vary by student Q5 Can the student level variation in GPA when job status is average be explained by gender Q6 Does the relationship between job status and GPA vary by student Q7 Can the student level variation in the relationship between job status and GPA be explained by gender Yio slope between job status and GPA You slope between gender and GPA Vii cross level interaction term Too variance of the random intercepts Too variance of the random intercepts 119 variance of the random slopes T 4 variance of the random slopes 1 7 slope of the regression line between job status and GPA 1 8 slope of the regression line between job status and GPA 1 9 how male slope differs from female slope 1 10 variance of the student level GPA values at job status
206. r employees of Clemson University As students and employees of a university many of these participants were interested in and understood the importance of research Through conversations with these participants during meetings the experimenter learned that many of these participants were motivated to comply with instructions and provide quality data for this study Additionally almost half of the sample was motivated to change their weight during the study which could have served as a motivator to comply with the study instructions Thus this university based sample that included individuals trying to change their weight may have had higher rates of compliance and better data quality than might be expected in the general population Additionally over 80 of the sample was Caucasian Therefore results cannot be generalized to all racial and ethnic groups Future Research Directions Five key areas of future research have been identified for improving the relationship between kilocalories and bites as detected by the bite counter First as discussed above the bite counter algorithm and design should be improved to reduce false positive detection increase true bite detections and reduce user errors associated with device use This research could range from the current database of bites being developed by the cafeteria study to ongoing usability studies during device development to automatic detection of eating by the bite counter device 18
207. r example results of a one year dietary intake study indicated that individuals ate more in summer and winter months compared to the spring and ate more on weekends than during the week Basiotis Thomas Kelsay amp Mertz 1989 Between person variance Between person variance in bite count can be conceptualized as reasons why bite count would differ between individuals For example if we compared the bite counts of Jane the graduate student and Greg the professional athlete we might see large differences in bite count based on their body size bite size gender and energy needs Preliminary research in our laboratory indicates that when the energy density and portion size of a food are controlled the number of bites taken varies more between individuals than within individuals Salley Scisco Hoover amp Muth 2011 These findings are supported by the existing literature which has found large differences between people in their patterns of energy intake Tarasuk amp Beaton 1991 Therefore an important step in the bite counter project is to identify the characteristics of an individual that will predict bite counts Multi level Linear Modeling The variance structure just described is nested or hierarchical Nested data is very common is social sciences research Bickel 2007 A classic example of nested data is students nested within classrooms Hox 2010 For example a researcher may 27 have a data set with 1 0
208. r than water for about two hours before the final meeting 220 14 Describe incentives Upon completion of the study you will receive 25 If you have completed all of your recalls and used your bite counter every day you will receive an additional 25 bonus It is okay to miss one day of recalls if you are unable to complete the recall or use the bite counter one day for example can t get to a computer leave your bite counter at home etc You will also receive a data report with your bite counts and calorie counts for each meal via e mail after study completion 15 Give participant their take home folder bite counter USB cord and charger Remind them that you can be contacted by phone during normal business hours and by e mail at any time Thank them for their participation Any questions After participant leaves Enter data in Prescreening spreadsheet 2 Add participant to ASA24 using the load participants file 3 Add e mail reminders to LetterMeLater com a Bite Counter start date reminder b 14 days of dietary recall reminders 4 Identify recruitment group and add to the in progress list 221 Appendix G Written Consent Form Information Concerning Participation in a Research Study Clemson University Ambulatory Monitoring of Food Intake Description of the Research and Your Participation You are invited to participate in a research study conducted by Eric Muth The purpose of
209. r them to remember to complete the study requirements The Lettermelater com website was an invaluable resource for delivering reminder e mails at participants preferred times without placing excessive burden on the experimenter Participant Recall of Bite Counter Use An extremely important step in data collection is that participants accurately report the time that they ate their meals For example if the bite counter was turned on at 7 16AM on Monday October 1 and a meal was reported at 7 16AM on Monday October 1 in ASA24 then these meals are easily matched during the data matching process The farther apart in the time the bite counter recording and the ASA24 report become the more difficult it becomes to match the meals Thanks to pilot testing and early data collection with the bite counter Jasper Scisco Parker Hoover amp Muth 2012 it was known that this meal start time information would be crucial During the orientation meeting the fact that meal start time information would be critical for future 184 data matching was emphasized to participants and they were encouraged to use the small notebook or another tool to take notes about the time they turned the device on As a result matching meals based on time for this study was much easier than during previous data collection efforts with many participants accurately reporting their meal start time within a few minutes of the start time recorded by the bite counter Futu
210. r x kilocalories interaction p lt 05 104 Model 2 Do Kilocalories predict Bites Kilocalories was entered into the model as a fixed effect at level 1 in order to address research question 1 Do kilocalories consumed during a meal predict number of bites recorded during a meal First change in model fit was assessed by comparing model 2 to the null model Results of the y deviance difference test 29468 49 28722 97 745 52 df 4 3 1 p lt 05 indicated that the addition of Kilocalories significantly improved model fit Next the change in within participants variance from the null model to model 2 was examined Kilocalories explained 21 4 563 43 442 76 563 43 100 of the within participants variance Lastly a significant positive relationship between Kilocalories and Bites was observed in the Kilocalories Bites slope of 0 04 Each one Kilocalorie increase during a meal corresponded on average to a 0 04 Bite increase Stated in a more practically meaningful way each 25 Kilocalorie increase during a meal corresponded on average to a 1 Bite increase Kilocalories was retained as a level 1 predictor for all subsequent models Model 3 Does Energy Density predict Bites Energy Density was added to the model as a fixed effect at level 1 in order to address research question 2 Does the average energy density of a meal predict number of bites recorded during a meal First change in model fit was assessed by comparing model 3 t
211. rating individual differences in GPA when job status is average 35 Q5 Can the student level variation in GPA when job status is average be explained by gender Examining Figure 1 10 it can now be asked if the differences in GPA when working the average number of hours per day for the sample can be explained by the gender of the students It can be seen that gender may explain some of this variation Scott and Greg the top two scatterplots are the male students and their GPAs at job status 2 are 2 7 and 2 3 Liz Ann and Kate the bottom three scatterplots are the female students and their GPAs at job status 2 are 3 6 3 3 and 2 6 Overall it seems that the females may have higher GPAs than males when working the average amount of time for this student sample Q6 Does the relationship between job status and GPA vary by student Figure 1 11 shows five individual scatterplots one for each student with GPA on the y axis and job status on the x axis Linear regression lines are fit to each data set and the slopes are indicated on the scatterplots Given these plots the sixth question that can be asked of the data set is does the relationship between job status and GPA vary by student Examining the slopes of the five lines it can be seen that the relationship between job status and GPA varies by student Some students GPAs increased as they worked additional hours and some students GPAs decreased as they worked additio
212. rd J D Svetkey L P 2008 Weight loss during the intensive intervention phase of the weight loss maintenance trial American Journal of Preventive Medicine 35 2 118 126 Hoover A Muth E Dong Y 2009 Weight Control Device US Provisional patent application serial no 61 144 203 Hox J J 2010 Multilevel analysis Techniques and applications 2 4 ed New York NY Routledge Huang T T K Harris K J Lee R E Nazir N Born W amp Kaur H 2003 Assessing overweight obesity diet and physical activity in college students Journal of American College Health 52 2 83 86 Hutchings S C Bronlund J E Lentle R G Foster K D Jones J R amp Morgenstern M P 2009 Variation of bite size with different types of food bars and implications for serving methods in mastication studies Food Quality and Preference 20 456 460 doi 10 1016 j foodqual 2009 04 007 Jasper P W Scisco J L Parker V G Hoover A W amp Muth E R 2012 May Using the bite counter to measure energy intake in overweight African Americans Poster to be presented at the 59 Annual Meeting of the American College of Sports Medicine San Francisco CA Kanfer F H 1970 Self monitoring Methodological limitations and clinical applications Journal of Consulting and Clinical Psychology 35 148 152 Kanfer F H 1971 The maintenance of behavior by self generated stimuli and reinforcement
213. rder to detect that the wearer has taken a bite of food or drink of liquid storing a log of time stamped bite count data It provides the capacity to detect record and store cumulative totals of bite counts over the day with little effort by the wearer Our research team has discovered that while eating the wrist of a person undergoes a characteristic rolling motion that is indicative of the person taking a bite of food Hoover Muth amp Dong 2009 The roll motion takes place about the axis extending from the elbow to the hand If for the right hand positive roll is defined as clockwise in direction as viewed from the elbow looking towards the hand and negative roll as counterclockwise motion the characteristic movement involves a cycle of roll motion that contains an interval of positive roll followed by an interval of negative roll 20 For a typical person the positive roll happens when a person is raising food from an eating surface such as a table or plate towards the mouth see Figure 1 3 The negative roll happens when the hand is being lowered or when food is being picked up by fingers or placed on a utensil The actual placing of food into the mouth usually occurs between the positive and negative rolls This characteristic roll is important because it differentiates wrist or arm motions caused by many other activities from a motion that can be directly associated with taking a bite of food or a sip or drink of a liquid
214. re research with the bite counter and dietary recall methods should continue to emphasize the importance of accurately recording meal start time An additional strength of the study was that participants reported a number of details about their bite counter use in the Daily Meals Questionnaire that aided data matching and error identification see Appendix C Without these details a researcher would not have much information to guide their error identification and decision making process However the format of this questionnaire made reporting these details tedious for some participants Future research should continue to collect these reports of bite counter use from participants but this questionnaire format should be simplified to reduce participant reporting burden Objective Measurement of Eating Behavior The bite counter is a unique device that can measure eating behaviors objectively in naturalistic real world settings Variables like bites meal duration and eating rate bites minute were measured without relying on participant self report or experimenter observations in laboratory settings This allowed for comparisons between objectively 185 measured laboratory variables and objectively measured real world variables comparisons that were previously not possible without the bite counter Study Limitations Accuracy of Bite Counter and ASA24 data As previously described technical difficulties and user errors could have c
215. re working 4 Turn on laptop computer and place on lab table a Confirm that Internet is working b Load ASA24 demo page survey monkey page and Google lab calendar 5 Puta pencil pen and both folders on lab table Now you re ready Wait patiently for the participant When participant arrives 1 Welcome the participant to the laboratory and ask them to have a seat at the conference table Put up Please Do Not Disturb signs on all 3 lab doors Ask the participant to read and sign the consent form Emphasize that participation will last for two weeks and will require about one hour of effort per day Explain the purpose of the study and general procedure ec The purpose of the study you will be participating in is to learn about the relationship between number of bites taken during a meal measured with the bite counter and a number of important variables that we are interested in studying including the number of calories in the foods you eat Today I am going take body measurements including height weight body composition waist and hip circumference After these measurements are taken I will describe the study procedures and instructions Do you have any questions before we begin Measure the participant s height and weight using the Tanita scale a b Have participants remove shoes but not socks and empty their pockets If between two height measurements e g between 1 2 inch and 3
216. ressed a single button on the device to put the device in bite counting mode At the end of each eating session the user again pressed the button to turn the device off Downloaded bite counter data provided a year month day and time stamp for each meal recorded the meal duration and the number of bites recorded at each meal The number of bites per meal recorded by the device was the main dependent variable for the present study Meal duration recorded by the bite counter served as a main independent variable Meal duration also allowed for the exploration of eating rate average bites minute or average kcal minute as a predictor of bite count 66 ASA24 Dietary Recall Dietary recalls were completed using the Automated Self Administered 24 hour Recall ASA24 National Cancer Institute 2011 ASA24 is an Internet based software tool that allows participants to complete 24 hour dietary recalls from a computer without the presence of a researcher ASA24 is based on a modified version of the interviewer administered Automated Multiple Pass Method AMPM 24 hour recall developed by the U S Department of Agriculture USDA and used in the U S National Health and Nutrition Examination Survey NHANES Food codes portion sizes and nutrient data in ASA24 originate from version 4 1 of the USDA s Food and Nutrient Database for Dietary Studies FNDDS and portion size photographs have been provided by Baylor College of Medicine Zimmerman et al
217. ributions Meal Duration also had positive skew and kurtosis Examination of bivariate scatterplots revealed almost perfect linear relationships between Bites and Meal Duration within participants Within participant correlations were examined to evaluate multicollinearity or the degree of relationship between the two variables Tabachnick amp Fidell 2007 The average within participant correlation between Bites and Meal Duration was 0 81 with one third of correlations gt 0 90 This indicated that Bites and Meal Duration may have represented the same variable and multicollinearity was present Because both Bites and Meal Duration were obtained from the Bite Counter recordings the longer the device was on the more bites were counted by the device The 93 decision was made to remove Meal Duration from the analysis because it would be likely to explain almost all of the variance in Bites leaving little opportunity for additional predictors to explain variance in Bites Number of People had extreme positive skewness and kurtosis values within participants Overall 61 2 of meals were eaten alone value 0 18 1 of meals were eaten with one other person 6 4 of meals were eaten with two people 6 4 of meals were eaten with three people and 7 9 of meals were eaten with 4 or more people values ranged from 4 to 50 Bivariate scatterplots of Bites and Number of People were non oval shaped with the majority of the data points centered on
218. rs than when eating alone The significant positive relationship between Height and Bites was qualified by a significant cross level interaction between Height and Kilocalories In order to examine the nature of the interaction simple slopes were calculated in accordance with Cohen et al 2003 using the fixed effects coefficients at high 1 SD and low 1 SD values of Kilocalories These slopes were significant at low B 0 047 SE 0 006 t 7 02 p lt 05 moderate B 0 040 SE 0 003 t 14 90 p lt 05 and high B 0 033 SE 0 006 t 2 5 32 p 05 values of Height Figure 3 6 shows that 136 the positive relationship between Kilocalories and Bites is stronger for shorter participants and weaker for taller participants 70 60 4 50 40 amp 30 E BE Low Height 20 4 Average Height 10 4 High Height 0 4 Low Kilocalories High Kilocalories Figure 3 6 The Kilocalorie x Height interaction for the outliers removed model demonstrating that the relationship between Kilocalories and Bites is strongest for shorter participants Day Level Model for the Outliers Removed Sample Data for the day level model using the sums for each day were prepared for the outliers removed sample following the same procedures as described for the day level model with the full sample The best fitting model for the outliers removed sample Model 19 was run using this day level data Tables 3 18 and 3 19 compare the meal
219. rs to determine if MLM was appropriate Heck Thomas amp Tabata 2010 Hox 2010 The amount of dependence on the individual was calculated as the intraclass correlation 89 ICC1 with values of 0 05 or greater indicating that significant nesting is present Heck Thomas amp Tabata 2010 Then the predictor variables were transformed with centering to improve interpretation of the intercept values Hox 2010 In the present MLM analysis the intercept was the expected value of bites when the predictors had a value of zero The problem with this is that zero was originally not meaningful e g the expected value for bites when kilocalories were zero Therefore the predictors were grand mean centered which resulted in the zero point for each predictor representing the mean for that predictor Hox 2010 Thus the intercept indicated the expected value of bites when the predictors were at their means for example the expected value for bites when kilocalories were at the mean Grand mean centering was also chosen for the present analysis because it allowed for comparison of parameter estimates across models with predictors at both level 1 and level 2 and it substantially reduced collinearity of interaction terms Bickel 2007 Hofmann amp Gavin 1998 Hox 2010 The research questions for the proposed study were tested with nested models using a bottom up hierarchical approach Hox 2010 That is parameters were entered into t
220. ry of weight cycling A MONET Montreal Ottawa New Emerging Team study Journal of the American Dietetic Association 109 718 724 doi 10 1016 j jada 2008 12 026 Subar A F Crafts J Zimmerman T P Wilson M Mittl B Islam N G Thompson F E 2010 Assessment of the accuracy of portion size reports using computer based food photographs aids in the development of an automated self administered 24 hour recall Journal of the American Dietetic Association 110 55 64 doi 10 1016 j jada 2009 10 007 Subar A F Thompson F E Potischman N Forsyth B H Buday R Richards D Baranowski T 2007 Formative research of a quick list for an automated self administered 24 hour dietary recall Journal of the American Dietetic Association 107 1002 1007 doi 10 1016 j jada 2007 03 007 Svensson M amp Lagerros Y T 2010 Motivational technologies to promote weight loss from Internet to gadgets Patient Education and Counseling 79 356 360 doi 10 1016 j pec 2010 03 004 Tabachnick B G amp Fidell L S 2007 Using multivariate statistics 65 ed Boston MA Allyn and Bacon Tarasuk V amp Beaton G H 1991 The nature and individuality of within subject variation in energy intake American Journal of Clinical Nutrition 54 464 470 Tate D F Jackvony E H amp Wing R R 2006 A randomized trail comparing human e mail counseling computer automated tailored counseling and no
221. s but I turned it off a few minutes after I finished eating If you turned the bite counter off a few minutes after you finished eating how many minutes elapsed between the end of your meal and when you turned the Bite Counter OFF minutes Did you turn the Bite Counter on and off multiple times during this meal You might do this for a multi course meal with break in between O Yes No If you turned the Bite Counter on and off multiple times for this meal please indicate how many times you turned the Bite Counter on and off in the box below Number of times on off Did you have any problems with the Bite Counter during this meal O Yes O No If you had problems with the Bite Counter during this meal please explain the problems below 204 12 Did you spend some or all of this meal time doing other activities For example talking reading a book watching TV using the computer working cooking etc Yes O No 13 If you spent some or all of this meal time doing other activities please list the percentage of meal time spent doing those activities and a description of the activities below Activity 1 Activity 2 Activity 3 Activity 4 Activity 5 Here are some examples Activity 1 For 50 of this meal I used my computer Activity 2 For 30 of this meal I talked to my family 14 What utensils did you use to eat your meal Check a
222. s relationship became nonsignificant Therefore when controlling for the effects of Social and the other predictors in the model the effect of Location was diminished Model 6 Does Intake Day predict Bites Intake Day was added to the model as a fixed effect at level 1 in order to address research question 7 Does day of the week predict the number of bites recorded during a meal First the change in model fit was assessed by comparing model 6 to model 5 Results of the a deviance difference test 28550 99 28543 71 7 28 df 8 7 1 p lt 05 indicated that the addition of Intake Day significantly improved model fit Next the 107 change in within participants variance from model 5 to model 6 was examined Intake Day explained an additional 0 2 419 41 418 42 419 41 100 of the within participants variance Lastly a significant negative relationship between Intake Day and Bites was observed in the Intake Day Bites slope of 2 32 On average 2 32 fewer bites were recorded for weekend meals than weekday meals Intake Day was retained as a level 1 predictor for all subsequent models Model 7 Do Kilocalories and Energy Density interact to predict Bites An interaction between Kilocalories and Energy Density was added to the model in order to address research question 3 Does the relationship between kilocalories consumed during a meal and number of bites recorded during a meal depend on the energy density of the food First th
223. s than participants with larger bites sizes Participants also took more bites when they ate meals with others and when they ate meals outside of the home although this meal location effect was not reliably produced across models Practical implications of these results for future bite counter development and research are discussed ii DEDICATION I dedicate this dissertation to my husband Gary Giumetti His love and endless support kept me moving forward during the most difficult stages of my PhD journey I also dedicate this dissertation to my family My parents Peter Scisco and Lori Scisco provided me with constant love and encouragement throughout my 20 years of formal education Thank you for teaching me that I could do anything I put my mind to and for always being there for me I would also like to congratulate and thank my sister Dr Leigh Scisco DPT for sharing the journey to doctor with me iii ACKNOWLEDGMENTS I would like to thank my advisor Dr Eric Muth for his outstanding guidance support training and advice during my graduate school career Thank you for encouraging me to grow intellectually throughout this dissertation process I would also like to thank Drs Adam Hoover Patrick Rosopa and Tom Alley for their guidance on this dissertation I would like to thank the SMART Scholarship program for providing funding for the two years during which I completed this dissertation I would like to thank all
224. se e mail jscisco clemson edu or call 864 656 1144 to receive your ID Inthe past two weeks how hungry have you felt Not hungry Somewhat hungry Moderately hungry Very hungry Extremely hungry In the past two weeks how full have you felt Not full at all Somewhat full O Moderately full Very full Extremely full In the past two weeks how often did you complete the 24 hour dietary recall For every food and beverage I consumed For most food and beverages I consumed 208 For only my main meals and the beverages consumed with those meals I forgot some meals and beverages I consumed O I forgot many meals and beverages I consumed I forgot to complete the dietary recall on one or more days Inthe past two weeks how easy or difficult did you find it to complete the 24 hour dietary recall Extremely easy Very easy O Somewhat easy Neither easy nor difficult Somewhat difficult Very difficult Extremely difficult What about the 24 hour dietary recall made it easy or difficult to complete Inthe past two weeks how much did you like or dislike completing the 24 hour dietary recall Extremely liked Liked very much O Liked somewhat 209 10 11 I2 13 14 Neither liked nor disli
225. se three lines of data were short duration recordings 25 7 and 58 seconds and the bite count values were low 2 0 and 3 bites These three meals were summed up for a total of 5 bites and duration of 1 minute 30 seconds Based on the low calorie snack description this summed up data appeared to be reasonable and was retained as corrected data ex UFC SS SM ND ATSDR west trme Did you w Di 88 3 72E 08 1507 2011 11 34 39 25 07 00 Lunch 1 03 PM Yes Yes 91 3 72E 08 347 18 2011 11 20 15 40 02 5 47 Snack 4 00 PM Yes Yes 92 3 72E 08 685 26 2011 11 20 15 46 11 11 25 93 3 72E 08 169 12 2011 12 20 17 02 51 2 49 Snack 5 00 PM Yes Yes 94 3 72E 08 1206 74 2011 11 20 18 21 37 20 06 Dinner 6 25 PM Yes Yes 95 3 72E 08 236 17 2011 11 20 21 27 08 3 56 Snack 9 34 PM Yes Yes 3 72E 08 23 04 20 0 25 Snack 11 07 PM Yes Yes 97 3 72E 08 7 0 2011 11 20 23 04 52 0 07 3 72E 08 23 05 10 99 3 72E 08 849 29 2011 11 21 8 51 36 14 09 Breakfast 8 52 AM Yes Yes 100 3 72E 08 127 4 2011 11 21 10 36 22 2 07 Snack 10 38 AM Yes Yes 101 3 72E 08 561 28 2011 11 21 12 25 10 9 21 Lunch 12 44 PM Yes Yes 102 3 72E 08 472 19 2011 11 21 13 53 25 7 52 Snack 2 02 PM Yes Yes 103 3 72F n amp 8 706 3R 2011 11 21 1602 4 11 46 Snack 4 04 PM Vas Ves c Figure 2 10 Example of a turning off bite counter data series Examining the duration of the bite counter recordings and the participants daily meals questionnaire allowed for the detection of possible meal durati
226. sents the amount of between person variance for each variable Table 3 2 Descriptive Statistics for the Meal Level 1 and Participant Level 2 Variables Level and Variable N Mean SD ICCI Level 1 Bites 3 606 39 15 26 62 0 24 Kilocalories 3 691 479 77 359 19 0 20 Energy Density 3 691 1 18 1 00 0 14 Location 3 749 0 43 0 50 0 19 Social 3 794 0 39 0 49 0 17 Intake Day 3 749 0 27 0 44 0 00 Level 2 Gender 83 0 57 0 49 N A Body Weight 83 172 58 42 79 N A Note Location coded 0 Home 1 Not at Home Social coded 0 Alone 1 With Others Intake Day coded 0 Weekday 1 Weekend Gender coded 0 Male 1 Female Within participant correlations between level 1 study variables are presented in Table 3 3 These within participant correlations assume that these relationships are the same within each participant Snijders amp Bosker 2011 and provide some preliminary information about the relationships among the variables Interestingly Kilocalories Location and Social are positively correlated with Bites indicating that on average eating a larger number of kilocalories at a meal eating outside of the home and eating with others is related to taking a greater number of bites of food during a meal However Energy Density is negatively correlated with bites indicating that fewer bites are taken during high energy density meals 99 Table 3 3 Within participant correlations between level 1 variables Variab
227. ship between Intake Day and Bites remained significant and indicated that on average participants took 1 88 fewer bites when eating on weekends compared to weekdays The relationship between Gender and Bites became nonsignificant indicating that when controlling for the effects of the other predictors Gender was no longer a significant predictor of Bites The nonsignificant relationship between Height and Bites was qualified by a significant cross level interaction between Height and Kilocalories as seen in Figure 3 2 The relationship between Kilocalories and Bites depended on the Height of the participants with a stronger positive relationship between Kilocalories and Bites for shorter participants and a weaker positive relationship between Kilocalories and Bites for taller participants 120 Additional Two Level Model In a second analysis the level 1 variables meal level variables were aggregated to the day level In order to provide support for aggregation ICC2 an index of reliability was calculated for all level 1 variables Snijders amp Bosker 2011 ICC2 is the ratio of between participants variance within participants variance between participants variance and a recommended cut off value is 0 60 Glick 1985 Essentially the ICC2 indicates the degree to which variables aggregated up to the day level can serve as a substitute for variables at the meal level Table 3 9 shows the ICC2 values for each level 1 variable B
228. ship between energy density and the number of bites taken at a meal For example imagine an individual consumes about 500 kilocalories per day at breakfast One day the individual has 500 kilocalories of watermelon and another day the 49 individual has 500 kilocalories of breakfast sausage This individual would need to take many more bites of the low energy density food the watermelon than the high energy density food the sausage to consume the same number of kilocalories for that meal Thus an individual may take more bites during a low energy density meal than a high energy density meal Conversely it is possible that individuals will take more bites of more energy dense meals because of their rich properties and high palatability to prolong and savor their hedonic properties and fewer bites of less energy dense foods because of their lighter qualities and lower palatability although one could also find a low energy density food to have pleasing qualities as well The proposed study will be the first to explore the energy density bite relationship Research Question 2 Does the average energy density of a meal predict number of bites recorded during a meal Kilocalorie by energy density interaction An interaction between two level 1 variables total kilocalories and average energy density is predicted It is possible that the relationship between kilocalories and bites depends on the energy density of the food Following the a
229. ship for the meal level model it can be seen that the relative strength of the interaction has decreased when Bites and Kilocalories are at their totals for the day 124 180 160 140 120 100 80 60 40 20 Bites 8 Low ED Average ED 4 HighED Low Kilocalories High Kilocalories Figure 3 3 The Kilocalorie x Energy Density interaction at the day level demonstrating that the relationship between Kilocalories and Bites is strongest for days with overall lower Energy Density The relationship between Location and Bites remained nonsignificant in the day level model Therefore when controlling for the effects of the other predictors Location was not a significant predictor of Bites at the day level The relationship between Social and Bites remained significant and indicated that on average participants took 5 3 more bites per day for each additional meal eaten with others The relationship between Intake Day and Bites remained significant and indicated that on average participants took 8 15 fewer bites per day when eating on weekends compared to weekdays The relationship between Gender and Bites was nonsignificant indicating that when controlling for the effects of the other predictors Gender was not a significant predictor of Bites Finally the relationship between Height and Bites was nonsignificant meaning that the number of bites taken during a day could not be predicted by a participant s height 125
230. significant positive relationship between Location and Bites was observed in the Location Bites slope of 2 04 On average 2 04 more bites were recorded when eating outside of the home compared to eating at home Because Location significantly improved model fit and explained a percentage of the within participants variance it was retained as a level 1 predictor for all subsequent models despite its relatively small contribution 106 Model 5 Does Social predict Bites Social was added to the model as a fixed effect at level 1 in order to address research question 6 Does the number of people an individual eats with predict the number of bites recorded during a meal First the change in model fit was assessed by comparing model 5 to model 4 Results of the y deviance difference test 28611 13 28550 99 60 14 df 7 6 1 p lt 05 indicated that the addition of Social significantly improved model fit Next the change in within participants variance from model 4 to model 5 was examined Social explained an additional 1 996 427 43 419 41 427 43 100 of the within participants variance Lastly a significant positive relationship between Social and Bites was observed in the Social Bites slope of 6 77 On average 6 77 more bites were recorded when eating with others compared to eating alone Social was retained as a level 1 predictor for all subsequent models It was also noted that when Social was added to the model the Location Bite
231. so known as hierarchical linear modeling HL M random coefficients modeling multilevel regression and mixed models Tabachnick amp Fidell 2007 For purposes of consistency and clarity this analysis technique will be referred to as MLM throughout the remainder of this document MLM is considered another method of 28 regression analysis conducted under specific conditions those conditions being nested data and relationships among the measurements that are nested Bickel 2007 MLM allows the researcher to analyze nested data that violates some of the assumptions of ordinary least squares OLS regression or repeated measures ANOVA analyses Tabachnick amp Fidell 2007 Repeated measures ANOVA requires complete data for each individual at each measurement occasion equal intervals between measurements and uncorrelated errors In MLM there is no requirement for complete data for each individual or each measurement occasion there is no need for equal intervals between measurements and the sphericity assumption uncorrelated errors over time can be violated That is MLM allows for measurement occasions to be correlated In the case of repeated measures analyses measurements are correlated because they originate from the same individual e g meals are eaten by the same person over time MLM deals with these correlated measurements by estimating error separately for measurement occasions and for individuals Tabachnick amp Fidell 20
232. some individuals are more susceptible to becoming obese due to genetic characteristics the obesity epidemic is undoubtedly attributable to dietary and behavioural causes M ller Bosy Westphal amp Krawczak 2010 p 612 The sources of this energy imbalance are numerous and varied French Story amp Jeffery 2001 Broadly obesity can be attributed to an obesogenic environment that promotes energy overconsumption and under expenditure Kirk Penney amp McHugh 2010 For example at the national level excess energy intake can be traced to government subsidized commodity crops e g corn a policy that has resulted in inexpensive widely available and calorie dense food products and a shortage of fresh fruits and vegetables Wallinga 2010 Within communities reduced access to grocery stores is related to higher obesity rates Lovasi Hutson Guerra amp Neckerman 2009 Environmental factors can also impact rates of physical activity Poor neighborhood walkability limited access to facilities and greater perceived safety hazards in a community are related to higher rates of obesity Black amp Macinko 2008 Although it is clear that changes are needed at a societal level in order to reduce obesogenic factors in our environment these changes are likely to take a large amount of time money and effort Before these large scale changes are made people can try to manage their weight by changing their eating and exer
233. sting self monitoring tools Self Monitoring Self monitoring can be defined as observing oneself and one s behavior Elfhag amp Rossner 2010 p 356 In the weight loss literature self monitoring refers to the process of observing one s body weight physical activity and or food intake over time Self monitoring has been described as the single most important ingredient to successful dietary change efforts McCann amp Bovbjerg 2009 the cornerstone of the behavioral treatment of obesity Wadden amp Letizia 1992 p 395 and the single most important component of behavioral treatment for obesity Clark Pamnani amp Wadden 2010 p 301 10 Theoretical Support for Self Monitoring Self monitoring emanates from self regulation theory Self regulation is defined as the many processes by which the human psyche exercises control over its functions states and inner processes Vohs amp Baumeister 2004 p 1 any effort by a human being to alter its own responses Baumeister Heatherton amp Tice 1994 p 7 and the exercise of control over oneself especially with regard to bringing the self into line with preferred thus regular standards Vohs amp Baumeister 2004 p 2 Self regulation theory emanates from systems theory and the concept of feedback loops Baumeister et al 1994 Basic systems theory feedback loops are called TOTE loops an acronym for Test Operation Test
234. t 12 01 p lt 05 values of Bite Size 145 As can be seen in Figure 3 8 the relationship between Kilocalories and Bites is stronger for individuals with smaller bite sizes than individuals with larger bite sizes That is participants with larger bite sizes took fewer bites to eat high kilocalorie meals compared to participants with smaller bite sizes who took more bites to eat high kilocalorie meals Overall compared an intercepts only model the final meal level model explained 38 of the total variance in Bites Bites k7 RH Small Bite Size Average Bite Size 4 Large Bite Size 0 Low Kilocalories High Kilocalories Figure 3 8 The Kilocalorie x Bite Size interaction at the meal level demonstrating that the relationship between Kilocalories and Bites is strongest for participants with smaller bite sizes 146 All of these relationships remained significant and in the same direction in the day level model The interaction term between Kilocalories and Bite Size decreased slightly Simple slopes were calculated in accordance with Cohen et al 2003 using the fixed effects coefficients at high 1 SD and low 1 SD values of Kilocalories These slopes were significant at low B 0 049 SE 0 004 t 13 27 p lt 05 moderate B 0 040 SE 0 003 t 11 71 p lt 05 and high B 0 03 SE 0 002 t 15 81 p lt 05 values of Bite Size As can be seen Figure 3 9 these slopes are similar
235. t is a social eating situation may provide a cue to an individual that they should monitor their bite count more closely during those meals in order to avoid over eating An individual could do this by setting an alarm when eating with others to go off at their average number of bites per meal when eating alone Intake Day Research question 7 investigated if day of the week dichotomized as weekday vs weekend could predict bite count The within participant correlations 0 01 and total correlations 0 03 between intake day and bites were small and non significant In the meal level model with the full sample 0 2 of the variance in bites was explained by intake day which indicated that intake day was a very small effect The relationship between intake day and bites indicated that about 2 additional bites were taken during meals on weekdays than meals on weekends and 8 additional bites were taken overall for weekdays compared to weekends This translated to eating 50 additional kilocalories during weekday meals and 200 additional kilocalories during weekdays overall This result is opposite the finding in previous research that people tend to eat more on weekends than weekdays e g Rhodes et al 2007 However intake day explained 096 of the variance in bites for the outliers removed model and the relationship between intake day and bites was non significant This indicates that the finding that participants took more bites on weekdays coul
236. t place too much additional burden upon participants While one single method of recording dietary intake may not be preferred by all participants efforts should be made to provide them with tools that are easy to use and quick to complete yet accurate 181 The bite counter was perceived as much easier to use than the ASA24 Although not an equal comparison by any means 75 of participants reported that they would prefer to use the bite counter over the ASA24 mainly due to its simplicity and the minimal amount of time needed to engage with the device However device problems and user difficulties could have reduced the accuracy of bite counter recordings Some of the devices in the study would shut off during meals due to an internal battery power problem Although participants were instructed to keep turning the device on to record their meal it is possible that some bites were not recorded However steps were taken during data screening to correct these errors by adding up these turning off sequences which may have reduced bite count underestimation Additionally participants reported difficulty remembering to turn the device on and off When this was noted by participants during their recall steps were taken during data screening to correct for these errors However participants may not have remembered to report errors in device recording or participants reports of the durations for which the device was off at the beginning of
237. ta Download Meeting sheet to the participant folder Note any reported bite counter problems and ASA24 problems on the sheet Also write the scheduled final meeting date and time on the sheet Also write down how many recalls and surveys have been completed 2 Setup laptop with Bite Counter software When participant arrives 1 Record the Bite Counter number on the sheet 2 Download the Bite Counter data and save to Dropbox Dissertation Data BiteCounterRaw ParticipantID a b Name the file ParticipantID DeviceNumber MonthDayYear Check the data for errors and ask the participant about any error like data For example if there are a lot of zeros or short meals with few bites is the device turning off or are they testing the device If the problems are severe replace the bite counter with the reserve bite counter and record the new bite counter number on the sheet 228 Ask the participant about any difficulties they are experiencing with the device recall or the survey Record these on the sheet Remind the participant of their final meeting and not to eat or drink anything other than water two hours beforehand Any questions 229 Appendix L Final Meeting and Meal Protocol One day before meeting 1 Send participant a reminder e mail Dear name This is a reminder that we will be meeting tomorrow date at time in Brackett 422 This meeting will last approximately 45 minutes and you wil
238. te if they were already tracking their meals and that writing things down during the day made it easier to complete Participants liked seeing what they ate and how much explaining that it held them accountable and increased their awareness of behaviors like snacking and their overall intake patterns 153 Table 3 27 Responses to usability questions about the ASA24 dietary recall Question N of total sample Frequency of completing ASA24 For every food and beverage 18 21 7 For most foods and beverages 56 67 5 3 2 4 For main meals and beverages 3 6 Forgot some meals and beverages 2 4 Forgot one or more days 4 8 Ease or difficulty of use Extremely easy 5 6 0 Very easy 22 26 5 Somewhat easy 30 36 1 Neither easy nor difficult 13 15 7 Somewhat difficult 10 12 0 Very difficult 3 3 6 Liked or disliked Liked very much 10 12 0 Liked somewhat 26 31 3 Neither liked nor disliked 28 33 7 Disliked somewhat 19 22 9 Experienced ASA24 problems Yes 28 33 7 No 55 66 3 ASA24 resulted in eating behavior change Yes 45 54 2 No 38 45 8 Recorded dietary intake elsewhere Yes 49 59 0 No 34 41 0 Participants also described why the ASA24 was difficult to complete what they disliked about the program and problems they had with the website Some found it difficult to remember meal details such as specific foods portion sizes and the time at which the meal was eaten Many participants expressed a desire for a favorit
239. technology into psychosocial and health behavior treatments British Journal of Health Psychology 15 1 39 Hetherington M M Anderson A S Norton G N M amp Newson L 2006 Situational effects on meal intake A comparison of eating alone and eating with others Physiology amp Behavior 88 498 505 Hill S W amp McCutcheon N B 1984 Contributions of obesity gender hunger food preference and body size to bite size bite speed and rate of eating Appetite 5 73 83 Hill J O Wyatt H Phelan S amp Wing R 2005 The National Weight Control Registry Is it useful in helping deal with our obesity epidemic Journal of Nutrition and Education Behavior 37 206 210 Hill J O Wyatt H R Reed G W amp Peters J C 2003 Obesity and the environment Where do we go from here Science 299 853 855 Hirsch I B Bode B W Childs B P Close K L Fisher W A Gavin J R Verderese C A 2008 Self monitoring of blood glucose SMBG in insulin and non insulin using adults with diabetes Consensus recommendations for improving SMBG accuracy utilization and research Diabetes Technology and Therapeutics 10 6 419 439 256 Hofmann D A amp Gavin M B 1998 Centering decisions in hierarchical linear models Implications for research in organizations Journal of Management 24 623 641 Hollis J F Gullion C M Stevens V J Brantley P J Appel L J A
240. ter your unique participant ID provided by the researcher If you do not remember your participant ID please e mail jscisco clemson edu or call 864 656 1144 to receive your ID Please enter yesterday s date which is the day you are completing the ASA24 dietary recall for MM DD YYYY How many meals and snacks from yesterday will you be recalling using ASA24 0 6 1 7 E12 us 03 09 04 10 5 More than 10 Participants were asked to answer the following questions about each meal Please answer the following questions for one meal you recalled for yesterday using ASA24 1 What was this meal or snack Breakfast CI 202 Brunch Lunch O Dinner O Supper Snack Just a drink What time did you eat this meal HH MM AM PM Did you wear the Bite Counter on your wrist during this meal L Yes O No Ido not remember Did you turn the Bite Counter ON at the beginning of this meal Yes O No I do not remember Yes but I turned it on after I began eating If you turned the Bite Counter ON after you began eating how many minutes did you eat before you turned the bite counter ON minutes 203 10 11 Did you turn the bite counter OFF after you finished eating your meal L Yes No Ido not remember Ye
241. terSense IneritaCubes and a US quarter uscite her licut i eorr dapes 23 The ambulatory bite counter used in the current study 24 Relationship between job status and GPA seeee 32 Scatterplot demonstrating the average difference in GPA between the gender TR 33 Scatterplot demonstrating how the relationship between job status and GPA depends on gender iusta rst eis eisas e P qe RV Ou LER Fo AERE dE 34 Student scatterplots demonstrating individual differences in GPA when job status ISA VET AG Cisse ares ea cou es erste D dI UR eR RO ERE 35 Student scatterplots demonstrating individual differences in the relationship between job status and GPA sse 37 The two level model with meals at level 1 and individuals at level 2 46 Hypothetical interaction between kilocalories and energy density 51 Selecting a meal time location computer and or TV use and who the meal was eaten with iniecto eite petes 69 Adding foods and drinks to the Quick List for lunch 69 Meal Gap Review between lunch and dinner esses 70 List of Figures Continued Figure 2 4 2 3 2 6 2 7 2 8 2 9 2 10 2 11 2 12 2 13 2 14 3 1 3 2 3 3 Page Portion size question for salad during the Detail Pass 70 Adding milk to tea during the Detail Pass
242. tes serve as a proxy for energy intake then one may predict that individuals with larger body weights will consume more bites Alternatively individuals with larger body weights may take larger bites resulting in no relationship or even a negative relationship between body weight and bites Research Question 10 Does body weight predict bite count Additional Two Level Model Our research group has hypothesized that bite count may be a more meaningful measure when aggregated to the day level compared to the meal level because this will reduce the variation in bite count produced by bite counter errors that could originate from false detections undetected bites or device errors Therefore in a second analysis the meal level predictors were aggregated to the day level and a two level model with 57 day as level 1 and individual as level 2 were explored in addition to the two level model with meals at level 1 Bite count for the entire day served as the dependent variable 58 CHAPTER TWO METHODS Participants Sample Size Sample size determination for the statistical power of a MLM analysis must consider the multiple levels 1 the sample size at level 1 nested within level 2 n 2 the sample size at level 2 N and 3 the total sample size n x N Bosker Snijders amp Guldemond 2003 n varies from person to person e g one person may have recorded 30 meals and another person may have recorded 40 meals but for si
243. the lab under the Lab Meal heading averages for five variables from the real world meals were calculated for each participant and are listed under the Average Real World heading As can be seen in Tables 3 24 and 3 25 participants took significantly fewer bites in the lab M 22 20 SD 6 92 than during average real world meals M 39 63 SD 14 03 1 66 9 84 p lt 05 and the two were not correlated r 19 p gt 05 Kilocalories per bite a proxy for bite size did not differ significantly between the lab M 17 15 SD 4 51 and the real world M 16 52 SD 6 56 t 66 0 79 p gt 05 and the two were positively related r 37 p lt 05 Meal duration was significantly shorter in the lab M 400 45 SD 110 49 than in the real world M 783 66 SD 269 72 66 11 75 p lt 05 but the two were positively correlated r 25 p lt 05 Eating 148 rate calculated as kilocalories per minute was marginally faster in the lab M 56 99 SD 17 70 than in the real world M 52 25 SD 21 39 t 66 1 93 p 05 but the two were positively correlated r 49 p lt 05 Eating rate calculated as bites per minute was not different in the lab M 3 44 SD 1 00 compared to the real world M 3 22 SD 0 32 t 66 1 84 p gt 05 and the two were positively correlated r 33 p lt 05 149 Table 3 24 Descriptive statistics for lab meal variables and real world
244. they defined Level 1 as daily within person variation in snacking behavior and hassles and Level 2 as between person variance e g eating style gender This allowed them to examine the impact of daily hassles and individual differences on snacking behavior as well as moderators of the hassles snacking relationship As another example Fulton et al 2009 examined the within person and between person predictors of children s BMI using MLM Using a two level hierarchical structure they defined Level 1 as daily within person variation in energy intake physical activity and sedentary activity and Level 2 as between person variation e g gender race This MLM analysis allowed these researchers to examine how daily changes in 44 energy intake and activity levels impact BMI how individual differences impact BMI and how these predictors might interact The Present Bite Counter Study MLM allows for the exploration of meal level and person level variables that could predict bite count In the present study participants wore bite counters daily and recorded bite counts for each meal eaten Every 24 hours participants also completed dietary recalls for each meal and survey measures asking about features of each meal This created a rich data set that allows for the investigation of predictors of bite count The current study used a two level model In MLM the dependent variable is always at the first level of analysis Hox 2010 Thus
245. this example reflect the random effects in MLM Random effects allow the mean of the DV intercept and the relationship between the level 1 IV and the DV slope to vary by the level 2 grouping variable Q4 Does GPA when job status is average vary by student Figure 1 10 shows five individual scatterplots one for each student with GPA on the y axis and job status on the x axis A line extends from the point for each individual when job status is at its mean mean job status is 2 hours per day to the y axis Given these plots the fourth question that can be asked of the data set is does GPA when job status is average vary by student It can be seen in Figure 1 10 that all five student have different GPAs when they work 2 hours per day This provides evidence of nesting and support for using MLM 34 45 45 Scott Greg 3 5 35 GPA when job status is average 2 7 3 amp 3 GPAwhen job status d d o is average 2 3 ut 256 25 4 A A A 2 24 0 1 2 3 4 0 1 2 3 4 Job Status Job Status AX z 45 Liz X Ann 4 3 5 3 5 N 5 B O 3 x 3 4 S a o GPA when job status is o average 3 6 GPA when job status 2 5 4 25 is average 3 3 2 T T T 1 2 4 T T T 1 0 1 2 3 4 0 1 2 3 4 Job Status Job Status 4 4 Kate 3 5 4 z 9 o 2 2 5 4 o GPA when job status is e average 2 6 2 T T T 1 0 1 2 3 4 Job Status Figure 1 10 Student scatterplots demonst
246. tial part of the individual behavior change process Self monitoring has been used to successfully help individuals manage their health behaviors For example self monitoring of blood glucose SMBG helps individuals to manage type 2 diabetes Hirsch et al 2008 Poolsup Suksomboon amp Rattanasookchit 2009 Self monitoring has also been linked to successful smoking cessation Fisher Lichtenstein Haire Joshu Morgan amp Rehberg 1993 A substantial body of literature has focused on examining the features of self monitoring that are associated with successful weight loss and weight maintenance Specifically self monitoring of body weight physical activity and food intake have been primary topics of investigation Self Monitoring of Body Weight Self monitoring of body weight or self weighing is associated with weight loss and weight maintenance success In a review of 12 studies that examined the relationship between self weighing and body weight 11 studies demonstrated that self weighing weekly or daily was associated with greater weight loss or more successful weight maintenance when compared to less frequent or no self weighing VanWormer French Pereira amp Welsh 2008 In some of these studies self weighing frequency was self reported and retrospective Butryn Phelan Hill amp Wing 2007 Linde Jeffery French Pronk amp Boyle 2005 Welsh Sherwood VanWormer Hotop amp Jeffery 2009 Wing 15 Tate Gorin
247. ticipants were removed from the data set and analyses were conducted on the data for the 69 remaining participants For the new data set 3 474 meals remained Of those meals 2 783 80 1 had bite counter and ASA24 data and 2 741 78 9 had data from all three sources Removing data from 14 participants 16 996 of the 83 original participants resulted in 449 meals removed from the data set 14 196 of 3 190 original meals Within participant correlations between level 1 variables are presented in Table 3 12 Overall these correlations were very similar to the correlations in the original model see Table 3 3 The Bites and Kilocalories correlation increased by 0 06 and remained significant r 0 51 p 0 05 Table 3 12 Within participant correlations between level 1 variables for outliers removed model Variable 1 2 3 4 5 1 Bites 2 Kilocalories 0 51 3 Energy Density 0 14 0 04 4 Location 0 06 0 09 0 03 5 Social 0 28 0 32 0 04 0 11 6 Intake Day 0 03 0 08 0 01 0 16 0 18 Note p lt 0 05 Location coded 0 Home 1 Not at Home Social coded 0 Alone 1 With Others Intake Day coded 0 Weekday 1 Weekend 127 Total correlations among all variables of interest are presented in Table 3 13 With outlier removal the correlation between Bites and Kilocalories increased by 0 07 and remained significant r 0 46 p lt 05 The negative correlation between Gender and Bites
248. tion were dropped from the model to create a more parsimonious model with significant predictors of Bites This final model was also run at the day level for the 60 participants The results of the final meal level and day level models including Bite Size are presented in Tables 3 22 and 3 23 143 Table 3 22 Random effects for the meal level and the day level bite size models for 60 participants 101 Model ei SE t00 SE Kcalories Meal level 331 26 9 83 89 68 18 34 000217 001 Day level 1426 39 77 05 1347 67 269 80 000181 001 Note SE Standard Error eij residual within participant variance 100 random intercept between participants variance t10 random slope variance p lt 05 Table 3 23 Fixed effects for the meal level and the day level bite size models for 60 participants Model oo y10 y20 y40 y120 y05 y15 SE SE SE SE SE SE SE Meal level 38 87 0 04 5 29 6 57 01 34 003 1 29 002 52 90 002 29 0005 121 11 0 04 28 58 4 42 01 4 40 002 Day level 4 99 004 5 19 1 53 006 1 11 0008 Note 00 grand mean of bites y10 kilocalories bites slope y20 energy density bites slope y40 social bites slope y120 kilocalories x energy density interaction y05 bite size bites slope y15 bite size x kilocalories interaction p lt 05 144 At the meal level the positive relationship between Kilocal
249. tlying meals of very high energy density with very few bites again indicating high energy density snacks Outliers for Meal Energy Density were removed within participant if the standardized value z score of the data point was greater than approximately 3 29 and if the data point was clearly separated from the rest of the distribution for the participant Tabachnick amp Fidell 2007 Sixty eight Meal Energy Density outliers were removed 1 8 of the meals with Meal Energy Density data Re inspection of the skewness kurtosis and plots of Meal Energy Density within participants revealed reduced positive skewness and kurtosis values and relatively normal distributions Then the dichotomous level 1 variables Location Home vs Not at Home Intake Day Weekday vs Weekend and the new variable Social Alone vs With Others were examined to see if the split between categories was 90 10 or greater within participants which would indicate reduced variability Tabachnick amp Fidell 2007 For Location 2 participants ate over 90 of meals at home 2 participants ate over 90 of their meals not at home and 1 participant ate all of their meals at home Across all meals for all participants 56 9 of meals were eaten at home and 43 1 of meals were eaten outside of the home For Intake Day 1 participant had 90 of reported meals that occurred on weekdays Across all meals for all participants 73 196 of meals were eaten on weekdays 95 and 26 9
250. to 18 Are you currently following a specific diet or way of eating Yes O No If yes please describe your diet 199 Appendix B Three Factor Eating Questionnaire R 18 TFEQ R 18 When I smell a sizzling steak or juicy piece of meat I find it very difficult to keep from eating even if I have just finished a meal 1 2 3 4 Definitely false Mostly false Mostly true Definitely true I deliberately take small helpings as a means of controlling my weight 1 2 3 4 Definitely false Mostly false Mostly true Definitely true When I feel anxious I find myself eating 1 2 3 4 Definitely false Mostly false Mostly true Definitely true Sometimes when I start eating I just can t seem to stop 1 2 3 4 Definitely false Mostly false Mostly true Definitely true Being with someone who is eating often makes me hungry enough to eat also 1 2 3 4 Definitely false Mostly false Mostly true Definitely true When I feel blue I often overeat b 2 3 4 Definitely false Mostly false Mostly true Definitely true When I see a real delicacy I often get so hungry that I have to eat it right away 1 2 3 4 Definitely false Mostly false Mostly true Definitely true I get so hungry that my stomach often seems like a bottomless pit 2 3 4 Definitely false Mostly false Mostly true Definitely true I am always hungry so it is hard for me to stop eating before I finish the food on my plate 1 2
251. to steady d Press zero Wait a few seconds for the scale to read 0 0g e Press PRINT to begin sending data to the computer f Remove plate and pull table cloth back over the table g Center the empty plate on the scale Again make sure it is not touching any wood h From the desktop open the WinWedge document JennaDissrtn SW3 and the excel document scale xls i Confirm that data is being sent from the scale to the excel file ii Close the excel file Set the table with a fork napkin and flowers Put the chair without arm rests at the table Turn on the laptop at the conference table Open the usability questionnaire on Survey Monkey Add the following to the participant folder and label with participant number date and time a Start SLIM scale b End SLIM scale c End LAM scale d Final meeting sheet i Add age and weight to the sheet as well as any problems from the last week Check ASA24 and survey monkey for the total number of completed recalls and surveys Obtain the participant payment from the safe and the participant compensation sheet Put with the participant folder on the conference table When the participant arrives at the laboratory 1 Welcome the participant to the laboratory and ask them to have a seat at the conference table Record the returned bite counter number on the final meeting sheet Download the Bite Counter data and save to Dropbox Dissertation Data BiteCounterRaw Part
252. to the slopes in Figure 3 8 and indicate that the relationship between Kilocalories and Bites is stronger for individuals with smaller bite sizes than individuals with larger bite sizes at the day level 200 180 160 140 120 100 amp a0 v 60 40 Average Bite Size RE Small Bite Size 20 4 Large Bite Size Low Kilocalories High Kilocalories Figure 3 9 The Kilocalorie x Bite Size interaction at the day level demonstrating that the relationship between Kilocalories and Bites is strongest for participants with smaller bite sizes 147 Lab Meal At the end of the study 75 participants ate a meal in the laboratory Eight participants declined to eat the macaroni and cheese either because they did not like the food or because it did not fit into their diet 1 e it was not low sodium or low fat Of those who ate the meal two participants had missing data on variables of interest and seven participants had outlying values across variables of interest z scores gt 3 29 separate from the rest of the data set when examining histograms that could have overly influenced relationships among variables e g ate for a very long time in the lab or ate very fast in the lab After dropping these nine participants 66 participants remained in the lab meal data set for analysis Descriptive statistics and correlations among variables are provided in Tables 3 24 and 3 25 In addition to the variables measured in
253. trical plates on the palms of the hands McArdle et al 2005 The current passes more quickly through hydrated fat free body tissue and extracellular water than fat or bone tissues McArdle et al 2005 Impedance is entered into an equation with height weight age and sex and body fat percentage is estimated Gibson Heyward amp Mermier 2000 The Omrom Body Logic Body Fat Analyzer provides an accurate estimate of body fat percentage 3 5 for approximately 7 out of every 10 men and 2 out of every 3 women when compared to hydrostatic weighing Gibson et al 2000 Additionally the Omron Body Logic Fat Analyzer is a noninvasive and economical way to measure body fat percentage MyoTape Tape Measure The MyoTape Accu Measure Greenwood Village CO was used to measure waist and hip circumference To measure waist circumference the tape measure was wrapped around the smallest circumference around the abdomen The tape measure was adjusted snugly without causing compressions on the skin To measure hip circumference the tape measure was wrapped around the biggest circumference around the buttocks Questionnaires All questionnaires were administered electronically using Survey Monkey Survey Monkey Palo Alto CA 75 Demographics A demographics questionnaire Appendix A asked participants to report a number of variables including age gender ethnicity handedness education level eating disorder history and frequency o
254. triction recommendations for the Dietary Guidelines for Americans help individuals lose weight Eating Behaviors 9 328 335 Carver C S 1979 A cybernetic model of self attention processes Journal of Personality and Social Psychology 37 8 1251 1281 Carver C S amp Scheier M F 1990 Origins and functions of positive and negative affect A control process view Psychological Review 97 1 19 35 Clark V L Pamnani D amp Wadden T A 2010 Behavioral treatment of obesity In P G Kopelman I D Caterson amp W H Dietz Eds Clinical obesity in adults and children 33 ed pp 301 312 West Sussex UK Wiley Blackwell Clemson University College Portrait 2009 Retrieved from http www collegeportraits org SC CU characteristics Clemson University Mini Fact Book 2011 Retrieved from http www clemson edu oirweb 1 FB factbook minifactbook cgi Cohen J 1992 A power primer Psychological Bulletin 112 1 155 159 Cohen J Cohen P West S G amp Aiken L S 2003 Applied multiple regression correlation analysis for the behavioral sciences Mahwah NJ Lawrence Erlbaum Associates Condrasky M Ledikwe J H Flood J E amp Rolls B J 2007 Chefs opinions of restaurant portion sizes Obesity 15 8 2086 2094 doi 10 1038 oby 2007 248 Conway J M Ingwersen L A amp Moshfegh A J 2004 Accuracy of dietary recall using the USDA five step multiple pass method in m
255. ts and then they were grouped by gender a level 2 variable that could explain this level 1 slope variation MLM Estimation Method Before demonstrating the MLM equations it is important to acknowledge that MLM uses a different estimation method compared to OLS regression based repeated measures ANOVA The OLS estimation method estimates intercept and slopes by seeking to make the sum of the squared differences between the observed value and the predicted value of the dependent variable across all observations as small as possible Cohen et al 2003 This is an analytic solution meaning the values can be derived directly from a set of equations Cohen et al 2003 The most common estimation method for MLM analyses is maximum likelihood ML Bickel 2007 Hox 2010 ML estimation provides values for the intercepts and slopes by seeking the values that have the greatest likelihood of resulting in the observed data Bickel 2007 That is ML estimation uses the values of the predictors and the dependent variable to find the intercepts and slopes that make the sample as likely or as typical as possible Cohen et al 2003 This is an iterative process Initial intercept and slope values are generated the likelihood of the estimates given the predictor and dependent variable values is 38 calculated and this guides the next iteration which tries to increase the likelihood of the sample values Cohen et al 2003 Hox 2010 The proc
256. udies However they may provide a useful starting point for the future bite counter With these two simple steps a bite size calibration before using the device and an indication of meal energy density before eating the bite counter could become an exciting new tool for self monitoring kilocalorie intake in real time during meals Conclusion The present study was motivated by the obesity epidemic that affects millions of individuals worldwide Although changes to the food and physical activity environments are necessary to reverse obesity trends those who are already obese can use tools to help them self monitor their energy intake The bite counter is a tool that has the potential to help individuals self monitor a number of different eating behaviors in real time including the number of bites taken meal duration bite rate and perhaps even the 192 number of kilocalories consumed The present study identified meal energy density and individual bite size as two important factors to consider for future bite counter development Once the relationship between kilocalories and bites has been improved through a possible combination of device calibration to the individual and to the meal type participants who receive device feedback and appropriate training may be able to use the device to reduce their energy intake This reduction of energy intake could lead to successful weight loss and weight maintenance 193 APPENDICES 194 Ap
257. ugh time in daily schedule to participate 3 illness 2 non compliance 2 losing a bite counter 1 getting bite counters wet 1 unable to use ASA24 on computer 1 and not wanting to wear and use the bite counter 1 Eighty three participants completed the two week study 43 females 40 males mean M age 33 73 standard deviation SD 13 02 Demographic characteristics of the sample are provided in Table 2 2 64 Table 2 2 Demographic characteristics of the 83 study participants Characteristic N of total sample Gender Male 40 48 2 Female 43 51 8 BMI category Underweight BMI 18 5 2 24 Normal weight BMI 18 5 24 9 38 45 8 Overweight BMI 25 0 29 9 23 214 Obese BMI gt 30 0 20 24 1 Ethnicity American Indian or Alaska Native 1 1 2 Asian or Pacific Islander 5 6 0 African American 5 6 0 Caucasian 67 80 7 Hispanic 2 2 4 Other 3 3 6 Education level High school diploma or equivalent 3 3 6 Some college 17 20 5 Bachelor s degree 31 37 3 Master s degree 22 ZLT Doctoral or professional degree 9 10 8 Household income 0 30 000 36 43 4 30 001 60 000 11 13 6 60 001 100 000 19 22 9 More than 100 000 15 18 1 Handedness Right hand 78 94 0 Left hand 5 6 0 Trying to lose weight 35 42 4 Trying to gain weight 3 3 6 Following a certain diet or way of eating 23 27 1 Note BMI calculated from orientation measured height and weight Other ethnicities reported were Persian African Black and
258. ulated meals Physiology and Behavior 47 569 76 264 Wilson S L amp Kerley G I H 2003 Bite diameter selection by thicket browsers The effect of body size and plant morphology on forage intake and quality Forest Ecology and Management 181 1 2 51 65 Wing R R amp Hill J O 2001 Successful weight loss maintenance Annual Review of Nutrition 21 323 341 Wing R R amp Phelan S 2005 Long term weight loss maintenance American Journal of Clinical Nutrition 82 2228 2258 Wing R R Tate D F Gorin A A Raynor H A amp Fava J L 2006 A self regulation program for maintenance of weight loss The New England Journal of Medicine 355 15 1563 1571 World Health Organization 2011 Obesity and overweight fact sheet Retrieved from http www who int mediacentre factsheets fs3 1 1 en index html Yao M amp Roberts S B 2001 Dietary energy density and weight regulation Nutrition Reviews 59 8 247 258 Yon B A Johnson R K Harvey Berino J Gold B C amp Howard A B 2007 Personal digital assistants are comparable to traditional diaries for dietary self monitoring during a weight loss program Journal of Behavioral Medicine 30 165 175 doi 10 1007 s10865 006 9092 1 Zick C D amp Stevens R B 2011 Time spent eating and its implications for Americans energy balance Social Indicators Research 101 267 273 doi 10 1007 s11205 010 9646 z Z
259. us significant model using the Chi square deviance difference test p 05 132 Table 3 15 Estimates of fixed effects for level 1 and level 2 predictors for the outliers removed model 00 10 20 30 40 50 120 560 01 02 T uc M i a SE SD GE GE GE GE GE G GE GE 39 757 l 1 57 3809 04 z 4 1 64 C001 2 3810 04 3 69 161 00D 42 r 3811 04 3 73 249 161 00D 42 C85 3821 04 3 51 L94 6 90 1 58 001 4D C85 090 3195 04 3 50 168 7 17 140 159 001 4D 80 92 C88 3832 40 5 45 L80 628 01 1 58 000 C52 C8 90 002 i 5 38 20 40 540 164 629 93 O1 144 159 000 5 80 95 89 00 175 gt 3861 04 547 178 624 01 5 90 154 001 52 84 90 002 3 09 H 38 38 04 547 178 627 01 06 154 001 52 84 90 002 03 3 i 30 60 04 5 50 L84 580 01 0 54 00D 51 81 87 002 3950 04 5 53 154 5J6 01 1 55 003 5D C85 87 002 E 3954 04 547 183 5 67 01 1 53 002 C5D 8 92 002 3943 04 5 47 167 5 96 89 01 1 56 003 51 C83 89 85 0020 a 3998 04 5 51 L86 5 75 01 5 61 01 L5 000 5D C8 87 002 3 01 005 Note Model 12 estimates were unstable and thus were not included 00 grand mean of bites y10 kilocalories bites slope y20 energy density bites slope y30 location bites slope y40 social bites slop
260. using a pedometer to estimate distance the user can calibrate it by running or walking a set distance e g mile on the inside of a track The number of steps that it takes the user to travel this distance is then used to calculate future distances For example if it took someone 1 000 steps to travel 1 2 mile then their pedometer would tell them that they went 1 mile when 2 000 steps were recorded A similar calibration step could be imagined for the bite counter A standard food with known calorie content and energy density could be eaten by a new bite counter user For example 500 kilocalories of low energy density food like pasta with an energy density of 1 5 kcals g could be eaten by a new user If the bite counter detected 20 bites for this meal 25 kcals bite would serve as the user s calibrated bite size This could then be held constant across all meals or for improved accuracy it could be adjusted based on the energy density of the foods being eaten with a decrease in kcals bite for lower energy density foods and an increase in kcals bite for higher energy density foods Manipulating bite size also has some applicability to the bite counter When bite size is manipulated taking smaller bites is associated with less energy intake Walden Martin Ortego Ryan amp Williamson 2004 Zijlstra de Wijk Mars Stafleu amp de Graaf 2009 or no change in energy intake Spiegel Kaplan Tomassini amp Stellar 1993 in controlled lab
261. uter The first day of data collection with the bite counter typically the day after the orientation meeting was scheduled The participant was asked for their preferred e mail address for daily reminders and their preferred e mail delivery time The data download 78 meeting and the final meeting and meal were scheduled and an appointment sheet was provided with dates times and meeting instructions see Appendix J Data Collection During the two week data collection period the participant was instructed to wear the bite counter for the entire waking day except when exercising swimming or showering They were instructed to record bites using the bite counter for every meal and snack they consumed during the day that consisted of foods and or beverages excluding meals for which an ending time would be far in the future greater than one hour and difficult to define Participants completed dietary recalls and surveys the day after a midnight to midnight period For example a participant completed a dietary recall on Wednesday October 26 anytime from 12 00am 11 59pm for the food and beverages consumed on Tuesday October 25 Participants received an automated e mail message at their preferred time reminding them to complete the recall and the survey This reminder included links to the ASA24 recall system and the Survey Monkey survey The participant was encouraged to contact the researcher via e mail or telephone anytime they exper
262. variables Mean Variable Min Max Mean SD t difference Lab Meal Kilocalories 142 410 359 91 76 31 Water ml 0 500 320 91 135 32 Bites 8 45 22 20 6 92 Kcals bite 6 26 17 15 4 51 Duration sec 242 698 400 45 110 49 Rate kcal min 26 46 96 85 56 99 17 70 Rate bites min 1 47 5 78 3 44 1 00 SLIM Before 13 68 33 77 10 96 SLIM After 24 90 67 48 13 85 LAM 34 87 65 89 13 12 Average Real World Lab Real world Bites 20 58 80 29 39 63 14 03 9 84 17 43 Kcals bite 6 82 34 46 16 52 6 56 0 79 0 63 Duration sec 367 62 1418 58 783 66 269 72 11 75 383 20 Rate kcal min 19 33 113 64 52 25 21 39 1 93 4 14 Rate bites min 2 48 4 02 3 22 0 32 1 84 0 21 Note SLIM scores below 50 indicate hunger and above 50 indicate fullness LAM scores below 50 indicate disliking and above 50 indicate liking All t test df 65 p lt 05 p 05 150 Table 3 25 Correlations between lab meal variables and real world variables Variable 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Lab Meal 1 Kilocalories 2 Water ml BD 3 Bites 37 01 4 Kcals bite 35 17 68 5 Duration sec 23 7 43 26 6 Rate kcal min 57 06 09 49 64 y Rate bites min 24 12 62 44 41 53 8 SLIM Before 49 28 05 38 06 36 02 9 SLIM After 21 09 02 18 16 31 14 12 10 LAM 07 09 02 3 04 09 02 15 7 Average Real World 11 Bites 06 18 J9 17 30 21 06 15 07 3 12 Kcals bite 2 07
263. vel demonstrating that the relationship between Kilocalories and Bites is strongest for days with overall lower Energy Density 125 xi List of Figures Continued Figure 3 4 3 5 3 6 3 7 3 8 3 9 Page Within participant correlations between Kilocalories and Bites for the original 83 participants iet cnt neo et enu Po aee canes 126 The Kilocalorie x Energy Density interaction for the outliers removed model demonstrating that the relationship between Kilocalories and Bites is strongest for meals with lower Energy Density 136 The Kilocalorie x Height interaction for the outliers removed model demonstrating that the relationship between Kilocalories and Bites is strongest for shorter participants 137 The Kilocalorie x Height interaction for the outliers removed model at the day level demonstrating that the relationship between Kilocalories and Bites is strongest for shorter participants 140 The Kilocalorie x Bite Size interaction at the meal level demonstrating that the relationship between Kilocalories and Bites is strongest for participants with smaller bite sizes sene 146 The Kilocalorie x Bite Size interaction at the day level demonstrating that the relationship between Kilocalories and Bites is strongest for participants with smaller bite sizes see 147 xil CHAPTER ONE INTRODUCTION Purpose The purpose of this study
264. vely related in order to generate predictor variables 25 Some examples of within and between person variance are described in Table 1 3 and in the text that follows Table 1 3 Within and between person bite count variance examples Within person variance in bite count Between person variance in bite count Energy of food kilocalories Body size e g body weight BMI Energy density kilocalories gram Body fat percentage How food is eaten e g utensils used Waist to hip ratio WHR Location of the meal Age Day of the week Gender Number of people at the meal Energy needs energy expenditure Meal duration Dietary restraint Bite size Within person variance Within person variance in bite count can be conceptualized as reasons why bite count would change for a given individual For instance if Jane the graduate student is wearing a bite counter and tracking her bite count during meals there are many possible reasons why her bite counts might vary There may be differences between her meals such as the caloric content or energy density of the foods the utensils used to eat and other activities engaged in while eating There may be differences between the days that she tracks bite count For example she may be 26 on vacation and eating all of her meals at Las Vegas buffets one day and she may be at work and eating at her regular meal times another day There may even be differences between weeks and seasons Fo
265. views 67 7 379 390 doi 10 1111 j 1753 4887 2009 00204 x Palmer M A Capra S amp Baines S K 2011 To snack or not to snack What should we advise for weight management Nutrition amp Dietetics 68 60 64 doi 10 1111 j 1747 0080 2010 01497 x Patrick K Raab F Adams M A Dillon L Zabinski M Rock C L Norman G J 2009 A text message based intervention for weight loss Randomized controlled trial Journal of Medical Internet Research 11 1 el Periwal V amp Chow C C 2006 Patterns in food intake correlate with body mass index American Journal of Physiology Endocrinology and Metabolism 291 E929 936 Phelan S Wyatt H R Hill J O amp Wing R R 2006 Are the eating and exercise habits of successful weight losers changing Obesity 14 4 710 716 Pliner P Bell R Hirsch E S amp Kinchla M 2006 Meal duration mediates the effect of social facilitation on eating in humans Appetite 46 189 198 doi 10 1016 j appet 2005 12 003 Polzien K M Jakicic J M Tate D F amp Otto A D 2007 The efficacy of a technology based system in a short term behavioral weight loss intervention Obesity 15 4 825 830 260 Poolsup N Suksomboon N amp Rattanasookchit S 2009 Meta analysis of the benefits of self monitoring of blood glucose on glycemic control in type 2 diabetes patients An update Diabetes Technology and Therapeutics 11 12
266. w Create a New Window in Excel and view the Merged Data and INF sheets side by side Using the date and time from the bite counter data and the Daily meals questionnaire match the data Copy and PASTE VALUES from the INF sheet into the Merged Data sheet as appropriate If you do not paste values the MealKCAL and MealFoodAmt will not transfer correctly Make note of any missing or incomplete ASA24 data on the Merged Data sheet Create a new column named MealED and calculate Meal Energy Density as MealKCAL MealFoodA mt On the Merged Data sheet create a new first column named MealID Number all meals sequentially regardless of missing or incomplete data This will help with sorting and identification of errors and outliers by number 238 Step 2 Identify data errors Note Figures 2 10 and 2 11 describe the decision making process for how to deal with the flagged data described below i e potential data errors 1 Daily meals questionnaire data a Was the bite counter turned on and off multiple times If yes flag data and sum up rows Record which meals were summed in ParticipantID data merging and screening history docx Move the deleted meals to the Removed sheet in ParticipantID xls b Were bite counter problems reported If yes determine if problem may have negatively affected the data For example participant reported the device turning off and there are 10 rows of data where the participant tried to get the d
267. we can to protect your privacy Your identity will not be revealed in any publication that might result from this study 222 In rare cases a research study will be evaluated by an oversight agency such as the Clemson University Institutional Review Board or the federal Office for Human Research Protections that would require that we share the information we collect from you If this happens the information would only be used to determine if we conducted this study properly and adequately protected your rights as a participant Voluntary Participation Your participation in this research study is voluntary You may choose not to participate and you may withdraw your consent to participate at any time You will not be penalized in any way should you decide not to participate or to withdraw from this study Contact Information If you have any questions or concerns about this study or if any problems arise please contact Eric Muth at Clemson University at 864 656 6741 If you have any questions or concerns about your rights as a research participant please contact the Clemson University Office of Research Compliance ORC at 864 656 6460 or irb clemson edu If you are outside of the Upstate South Carolina area please use the ORC s toll free number 866 297 3071 Consent I have read this consent form and have been given the opportunity to ask questions I give my consent to participate in this study Participant s signature Date
268. weight and have been successful at maintaining that weight loss In their first report from the NWCR Klem Wing McGuire Seagle and Hill 1997 surveyed 629 women and 155 men who had lost at least 30 kg and kept it off for at least one year They found that a wide variety of weight loss strategies were used including restricting intake of certain types or classes of food 87 6 of the sample eating all types of food but limiting the quantity 44 2 counting calories 43 7 and limiting the percentage of daily intake from fat 33 1 Once the weight had been lost weight loss was successfully maintained by limiting intake of certain foods 92 limiting quantities of foods eaten 49 2 limiting the percentage of daily energy from fat 38 1 counting calories 35 5 and counting fat grams 30 Almost all of the registry members also exercised and weighed themselves regularly The NWCR researchers also investigated if losing weight using different strategies and approaches resulted in different weight maintenance behaviors McGuire Wing Klem Seagle and Hill 1998 examined three groups in the registry those who had lost weight on their own those who had lost weight using a program e g Weight Watchers or Jenny Craig and those who had lost weight using liquid formulas e g Slim Fast Despite using different methods to lose weight all groups maintained their weight loss by consuming low calorie low fat diets and performing
269. weight gain Elfhag amp R ssner 2010 Varying definitions of weight fluctuation from the literature are provided in Table 1 1 Table 1 1 Definitions of successful weight loss weight maintenance and weight fluctuation Successful weight loss Weight Maintenance Weight fluctuation 5 10 weight loss 10 weight loss maintained Repeated gains and significantly improved for 1 year National Weight losses of weight over obesity related metabolic risk Control Registry NWCR time Diaz Mainous factors Goldstein 1992 Wing amp Hill 2001 amp Everett 2005 p 153 5 weight loss Crawford 5 weight loss maintained Losing and regaining Jeffery amp French 2000 for 2 years Crawford between 5 and 20 Jeffery amp French 2000 pounds at least once Bishop 2002 Losing more than 2 BMI Maintaining weight loss forat The number of times a points Cuntz Leibbrand least 6 months Elfhag amp diet has resulted in a Ehrig Shaw amp Fichter 2001 R ssner 2010 weight loss of 10 kg or more Strychar et al 2009 Successful weight loss maintainers provide important insights into behaviors that promote successful weight loss maintenance In a qualitative study Haeffele 2008 identified a four stage process of weight loss maintenance shown in Figure 1 1 First an individual has an ahah or epiphany moment when they decide that they are going to lose weight These moments have been described as triggering events
270. wer bites per day on average Additionally height explained 9 8 of individual differences in the relationships between kilocalories and bites The interaction between kilocalories and bites was significant for the full sample at the meal level and for the outliers removed sample at the meal level and the day level Simple slopes were consistent across these three models and indicated that participants of average height about 5 7 in both samples ate about 25 kilocalories per bite taller participants about 5 10 5 in both samples ate about 30 172 kilocalories per bite and shorter participants about 5 3 5 in both samples ate about 21 kilocalories per bite This leads to the possibility that taller individuals take larger bites and height could possibly serve as an individual difference variable approximating bite size To explore this idea the total correlation between bite size in the lab kilocalories per bite and height was calculated for the 60 participants in the bite size model and a significant positive correlation of 0 28 indicated that bite size and height are somewhat related However bite size and body weight were also significantly positively correlated 0 26 and body weight was not a significant moderator of the kilocalories bites relationship This suggests that there may be something unique about height that potentially allows it to be related to bite size in the real world such as the overall size of o
271. with 60 participants eerte e tereti tete etre e nene Re 141 Total correlations between level 1 and level 2 variables for the bite size model with 60 participants sseeseeeeeeeeeeeeeneeennennns 142 viii List of Tables Continued Table 3 22 3 23 3 24 3 25 3 26 3 27 3 28 Random effects for the meal level and the day level bite size models fot GO participants iseset in n eMe ue a E a PR ERR Ee c ae PNE Fixed effects for the meal level and the day level bite size models for OE Dart cipantso er oae ds Fou e EEE E Descriptive statistics for lab meal variables and real world variables Correlations between lab meal variables and real world variables Body measurements from self report pre study and post study Responses to usability questions about the ASA24 dietary recall Responses to usability questions about the bite counter 1X Figure 1 1 1 2 1 3 1 4 1 5 1 6 1 7 1 8 1 9 1 10 22 2 3 LIST OF FIGURES Page The four stage process of weight maintenance described by Haeffel 2008 b RO vm ee erated 6 A basic TOTE feedback loop example for weight loss 12 Positive wrist roll when taking a bite eese 21 The tethered InertiaCube3 attached to an athletic wristband 22 The smaller MEMS sensor center compared to the In
272. y Quarterly Journal of Medicine 99 565 579 doi 10 1093 qjmed hcl085 McArdle W D Katch F I amp Katch V L 2005 Sports and exercise nutrition gn ed Baltimore MD Lippincott Williams amp Wilkins McCann B S amp Bovbjerg V E 2009 Promoting dietary change In S A Shumaker J K Ockene amp K A Riekert Eds The handbook of health behavior change New York Springer Publishing Company McGuire M T Wing R R Klem M L amp Hill J O 1999 The behavioral characteristics of individuals who lose weight unintentionally Obesity Research 7 485 490 McGuire M T Wing R R Klem M L Seagle H M amp Hill J O 1998 Long term maintenance of weight loss Do people who lose weight through various weight loss methods use different behaviors to maintain their weight International Journal of Obesity 22 572 577 Medicis S W amp Hiiemaie K M 1998 Natural bite sizes for common foods Journal of Dental Research 77 295 Mesas A E Munoz Pareja M Lopez Garcia E amp Rodriguez Artalejo F 2012 Selected eating behaviours and excess body weight A systematic review Obesity Reviews 13 106 135 doi 10 1111 j 1467 789X 2011 00936 x Micco N Gold B Buzzell P Leonard H Pintauro S amp Harvey Berino J 2007 Minimal in person support as an adjunct to Internet obesity treatment Annals of Behavioral Medicine 33 1 49 56 Mishra A Mishra
273. y amp Behavior 50 4 729 738 de Castro J M 1996 How can eating behavior be regulated in the complex environments of free living humans Neuroscience and Biobehavioral Reviews 21 119 131 de Castro J M 2004a The time of day of food intake influences overall intake in humans Human Nutrition and Metabolism 134 104 111 de Castro J M 2004b Dietary energy density is associated with increased intake in free living humans Human Nutrition and Metabolism 134 335 341 de Castro J M 2005 Stomach filling may mediate the influence of dietary energy density on the food intake of free living humans Physiology amp Behavior 86 32 45 de Castro J M 2010 The control of food intake of free living humans Putting the pieces back together Physiology amp Behavior 100 446 453 252 de Castro J M amp Brewer E M 1991 The amount eaten in meals by humans is a power function of the number of people present Physiology amp Behavior 51 1 121 125 de Castro J M Brewer E M Elmore D K amp Orozco S 1990 Social facilitation of the spontaneous meal size of humans occurs regardless of time place alcohol or snacks Appetite 15 89 101 de Castro J M amp de Castro E S 1989 Spontaneous meal patterns of humans Influence of the presence of other people American Journal of Clinical Nutrition 50 237 247 de Castro J M amp Plunkett S 2002 A general model of int
274. y Reviews 9 624 630 doi 10 1111 j 1467 789X 2008 00516 x Salley J N Scisco J L Hoover A W amp Muth E R 2011 Variability in bite count and calories per bite across identical meals Poster presented at the 29th Annual Meeting of the Obesity Society Orlando FL Scisco J L 2009 The bite detector A device for the behavioral treatment of overweight and obesity Master s thesis Retrieved from ProQuest Dissertations and Theses database UMI No 1473362 261 Scisco J L Blades C Zielinski M amp Muth E R under review More pieces lead to larger portion size estimates of JELL O squares Scisco J L Muth E R Dong Y amp Hoover A W 2011 Slowing bite rate reduces energy intake An application of the bite counter device Journal of the American Dietetic Association 111 1231 1235 Sharma A M amp Padwal R 2010 Obesity is a sign over eating is a symptom An aetiological framework for the assessment and management of obesity Obesity Reviews 11 362 370 doi 10 1111 j 1467 789X 2009 00689 x Shay L E Seibert D Watts D Sbrocco T amp Pagliara C 2009 Adherence and weight loss outcomes associated with food exercise diary preference in a military weight management program Eating Behaviors 10 220 227 Schutz H G amp Cardello A V 2001 A labeled affective magnitude LAM scale for assessing food liking disliking Journal of Sensory Studies 1
275. y plots q q plots were evaluated in addition to skewness and kurtosis values Bites and Meal Kilocalories had positive skew and positive kurtosis values within participants Inspection of within participant histograms boxplots and q q plots indicated that the positive skew and kurtosis values were most likely the result of outliers on the positive end of the distributions In order to determine if transformation of these 92 variables was appropriate and to examine linearity bivariate scatterplots of Bites and Meal Kilocalories were examined within participants The pattern of data was mostly linear and oval shaped indicating that the positive skewness and kurtosis were not contributing to nonlinearity Therefore transformation was not appropriate for Bites and Meal Kilocalories Tabachnick amp Fidell 2007 Outliers for Bites and Meal Kilocalories were removed within participant if the standardized value z score of the data point was greater than approximately 3 29 and if the data point was clearly separated from the rest of the distribution for the participant Tabachnick amp Fidell 2007 Fifty five Bites outliers were removed 1 4 of the meals with Bites data and 45 Meal Kilocalorie outliers were removed 1 2 of the meals with Meal Kilocalorie data Re inspection of the skewness kurtosis and plots of Bites and Meal Kilocalories within participants revealed reduced positive skewness and kurtosis values and relatively normal dist

Sources of Variance in Bite Count

Contents

Download Pdf Manuals

Related Search

Related Contents