Home

DRAFT User Manual – June 2015 - KFL&A Public Health Informatics

1. sob ath 0 949 _ congas Cellulitis Ce u iti S 0 912 left Eam infection Feier panniculectomy Soyo dy ENVIRO 2013 Age 0 120 fr Eg P Environmental F G u mn 0 863 d E fest si heel bite Ww hypothermia 10 legs ons injury m EOH 2013 Age 0 120 assaulted go day avg Corr coef 0 864 mgeston intoxicated P7 PL age sl TN aid Alcohol 2 8 a IN EtOH m h d 0 864 E ain requesting alco ol injury intoxication 5 52 ACES Manual The Science of ACES Visual Representation Syndrome Words Cloud from CC Correlation NACRS v ACES GASTRO 2013 i ts 0 120 syncope diarhea rash o weakness BUSeB c days diarrhea vomiting 22 b oody Garhooa amp abdominal NACRS ILI 2013 Age 0 120 350 ILI winpterm chilla headache chest body 150 250 Mental Health mental Un ua amp om un ehaviour 9 see bizarre psychosis a sb c N E 5 at depression partum O spotting n m menstrualPOSt intact bleed Gynocological va Q nN ci vag pain bleeding 53 ACES Manual The Science of ACES 3 4 Alerts and Outbreaks According to the Center for Disease Control the aim of syndromic surveillance is to identify unusual disease clusters through the detection of syndromic early symptomatic cases allowing outbreak detection that is earlier in time than would otherwise occur with conventional repor
2. o 130 CTAS Figure 19 Line Listings AD Tools menu User Interface Guide 2 7 Map E The Map page allows you visually examine the spatial distribution of acute care hospital visits The first thing you ll see when you select the Map tab on the main navigation toolbar is a map of Ontario To move around the map simply move the cursor over the map click and drag the map reveal locations you want to view Likewise to zoom in or out you can rotate the wheel on your mouse or move the square on the scale in the top left corner beneath the small house symbol Figure 20 If you click the house symbol the map view will be returned to the default zoom and location settings At the bottom right corner the location displayed is indicated on a large scale map Remember that although the map can be moved around to various Figure 20 Map scale tool jurisdictions only data from Ontario hospitals that are actively sharing data with ACES will be displayed on the map The main features of the map be manipulated using the tabs displayed on 32 Data Layers Mapping Style Choropleth Classifiers Maximum Entropy Classification 52014 Syndromes AST Level Of Geography FSA Data Classifications Equal Interval Number Of Classifications 4 Percentage Range g 10 Date Range 04 14 2015 05 14 2015 Clear Data Request Data Figure 21 Map Features Data menu P ACES Manual the right side of the screen Each tab
3. ED 26 ACES Manual User Interface Guide 2 4 4 3 Download Chart At the bottom left corner of the page you can choose Download Chart to download the image as a Portable Network Graphics PNG file see bottom left corner Figure 13 2 5 Line Listings ED ACES gives you the ability to display emergency department ED visits as a textual list with the standard patient information we receive directly from hospitals i e Date Time Admission Type Age Gender FSALDU Hospital Syndrome Complaint CTAS EMS Arrival and Patient LHIN The main page that opens when you choose Line Listings ED will display the most recent data that you are registered to view as individual emergency department visits in descending chronological order for the past week Error Reference source not found Choose the parameters displayed using the dropdown menu from the arrow on the top right corner of the table You can choose to group according to these parameters as well see Figure 4 and use the Group By Icon 2 5 1 Line Listings ED Tools Menu To view specific line listings the Tools menu on the right of the screen will guide you in the available options to customize the emergency department visits that are displayed Figure 14 Selecting Submit at any time with generate the Line List with the details provided Press Reset to return all options to their default settings Each option is described in the following Health Unit and Hospitals Th
4. including a description of all functionalities such as mapping and alerting protocols The scientific background found in Section 3 chronicles the development of ACES important scientific concepts and technical terminology 1 1 Introduction to ACES The global nature and faster pace of the contemporary health care environment have presented formidable challenges to traditional methods of health surveillance These traditional methods including surveys regular reporting of priority diseases from sentinel primary care practices and retrospective analysis of hospital charts were put to the test in the fall of 2003 with the pandemic threat of SARS Numerous reports from this period including the Walker Report the Naylor Report and the Campbell Commission confirmed that Ontario s emergency preparedness was in need of updating in addition to the need for better health surveillance Concurrent to global events that were testing the 7 ACES Manual ACES Backgrounder limits of public health surveillance developments in machine learning were making large scale automated data analyses possible and advances in geographic information systems GIS were enabling geospatial visualizations of data at global national and regional scales At this confluence of public health need and technological advancement was the development of real time surveillance methods capable of providing essential monitoring for a growing list of health conditions such as in
5. simple hand written rules of direct word for word translation are not sufficient due to the complex unrestricted and ambiguous nature of language NLP must extract meaning from text and deal with spoken or written prose that is not grammatically correct Out of these restrictions statistical NLP methods developed with probabilistic approaches that replace numerous detailed rules with statistical frequency information The algorithms are refined or able to learn through training the program with large amounts of data with the correct answers and then testing the robustness of the system with an unknown data set and then repeating until the system performs satisfactorily The algorithms do not rely on key word searches but rather probabilistic decisions based on attaching learned weighted values to each word part of word or phrase in the CC The algorithms do not supersede hand written rules but are complementary Medical records are a particular challenge for NLP algorithms CCs for example are written to be concise descriptions of the reason for a visit to the emergency department therefore CCs are often written with abbreviations context sensitive vocabulary idiosyncratic or hospital specific nomenclature and often misspellings occur under the inherent demanding conditions in acute care settings Furthermore a single symptom may be observed for several possible diagnoses For example fever is associated with numerous conditions De
6. DROP DOWN MENU WITH GROUP BY OPTIONS The next sections will describe the different menu tabs that enable you to utilize ACES to suit your agency s needs and goals The Figure 4 Dropdown display for current alerts main navigation options are shown as tabs in the top right corner of the ACES main page Figure 5 Each tab will be discussed in sequence Epicurves Line Listings ED Line Listings AD Maps and Alerts p a ACES Epicurves Line Listings ED Line Listings AD Maps Alerts Figure 5 Main navigation options 2 4 Epicurves The Epicurves page allows you to graph acute care facility visits as a function of various user defined parameters such as geographic region gender age and date The default display is a chart of all visits for the past seven days for all hospitals gender and ages Figure 6 At the top of the chart are options to add statistically descriptive features Moving Average Standard Deviation Your Health Unit Hospital Ee Aa EE s to the graph Figure 7 None no moving average displayed is the default setting 7 seven day moving zn average is displayed as a dotted line s 14 fourteen day moving average is Visits displayed and Max the overall mean for the chosen time period is displayed Moving average is Apr 16 2015 Apr 17 2015 Apr 18 2015 Apr 19 2015 Apr 20 2015 Apr 21 2015 Apr 22 2015 Apr 23 2015 Date calculated as the mean of the last Figure 6 Defaul
7. ExNin Page VEE EE EE EE EE Ge ee 18 2 3 MamNlilanding Page VENER GEENEEN EE nnne nnne EE EE nnns 18 2 4 Seen WE TEE EE OE OE OE N rT rere Te 20 2 4 1 Epicurve Display Options Tools nnne nennen nennen nennen nnne nnne 21 2 4 1 1 ClASSIFICCIII MMIII cccccsssssscccccccscnsssssecccccssenasseascceccsscaasseeesccecseanaseeesececcenansssecseeeseoas 22 2 4 2 Epicurve Display Options Advanced ccccccesseccceesececeeseceecesececeusececseneceeseenecessuseeesseneess 23 2 4 3 Creating an Epicurve An Example sessie nn ee Ee Ge ed wee Gee Do ee dd 24 2 4 4 Additional Fest es ee EER Ee Re EE GR OG BOER EH DD tU SUI M NUS 26 EE EEN oe ad EE N OE 26 2AA DEPON DE D se ER m um 26 2443 Nr 27 2 5 ad BE dig vr 27 ii ACES Manual Table of Contents 2 5 1 Line Listings ED Tools MENU sees ss sees se ee ee ee eek ee ee ee ee RE ee RR ee ee Ee ee 27 2 5 2 Line Listings ED Advanced MENU ees sees ee ee ee ee ee ee ee ee ee ee nnns 28 2 5 3 VIEWING an Individual Line Listing oes io biet Ne ou ini N de Re Ge ee Ed Ee Ee as 29 2 6 Hs SAD ae Ee ee E Ge GE inten Ge 30 2 6 1 Line bistingss AD Tools MENU sr 31 2 6 2 Line Listings AD Advanced MENU ees sees ee ee ee ee ee ee ee ee ee ee ee RE ee RE 31 2 7 Me 32 2 7 1 Map Features Piute 32 2741 Mapping SIE eos KEENE 32 2 7 1 2 Classifiers aa ee er 33 2 7 1 3 Classification df EE ee eke sns 33 2
8. Health Protection and Promotion Act HPPA public health agencies in Ontario are required to track over 50 communicable diseases Appendix D Notification of a case of these diseases usually occurs through the testing laboratory and public health is required to notify patients of the treatment options as well as what to expect during the course of their illness and or treatment Reportable 10 ACES Manual ACES Backgrounder diseases can be monitored using ACES in several different ways Specific syndromes related to the reportable disease can be monitored key words from the chief complaints specific to the disease are used to classify potential cases Alternately epidemiologists can examine individual ED records in ACES and flag instances that require further investigation by a public health nurse infection control practitioner or health inspector In the near future this process will be automated and become an even more timely service that includes direct emailing of line list data to the appropriate person to allow for even quicker follow up than may occur via current lines of communication This will allow public health agencies to meet their obligations through the HPPA while also allowing for the quick isolation if necessary and education of the afflicted patient to mitigate the potential spread of the disease 1 2 3 Mass Gatherings 1 2 3 1 G8 G20 Summit Mass gathering events have the potential to generate significant public health
9. Hospital geographies choose Local PHU Hospitals from the Hospital dropdown menu to display local patients using the hospitals within your health units jurisdiction choose Province Wide for the hospital visits for local patients at all hospitals reporting to ACES or choose Outside of PHU Hospitals for local patients using hospitals outside of the local area When all display options have been chosen click the large green Submit button Reset will return all parameters to default settings 2 5 3 Viewing an Individual Line Listing You may be interested in the details regarding an individual record Simply place the cursor anywhere on the line describing the record and click the full record will open as a separate window Tabs at the top of the window are for Details see description below Alerts indicates if this record triggered an alert is a reportable communicable disease and the date time that the alert was reported and Metadata information specific to storing the data In the Details tab all information regarding the record is shown including the following El hospital hospital PHU and hospital LHIN El admission information date time day of week week of year El triage information CTAS arrival by emergency medical services indicated as TRUE FRI El patient demographics 5 character postal code county municipality LHIN PHU CSD FSA gender and age El chief complaint transcribed from tri
10. Jan 2014 Appendices Health Unit Key Hospitals within Health Unit ACES start Below St Joseph s Health Centre STJOE Jun 2010 St Michael s Hospital SMH Mar 2013 Sunnybrook Health Sciences Centre TBD Toronto East General Hospital Oct 2014 University Health Network General Site TGH July 2013 University Health Network Princess Margaret PMH July 2013 University Health Network Western Site TWH July 2013 William Osler Health System Etobicoke General Hospital EHC Jan 2011 WECHU H tel Dieu Grace Hospital HDGH Jun 2013 Leamington District Memorial Hospital LDMH Jun 2013 Windsor Regional Hospital WRH Jun 2013 Health Unit Abbreviations APH Algoma Public Health CKPHU Chatham Kent Public Health Unit DRHD Durham Region Health Department EOHU Eastern Ontario Health Unit GBHU Grey Bruce Health Unit HKPRDHU Haliburton Kawartha Pine Ridge District Health Unit HRHD Halton Region Health Department HPH Hamilton Public Health HPECHU Hastings and Prince Edward County Health Unit KFLA KFL amp A Public Health LAMBTON CHSD Lambton Community Health Services Dept LGLDHU Leeds Grenville and Lanark District Health Unit NRPH Niagara Region Public Health NBPH North Bay Public Health NWHU Northwestern Health Unit OPH Ottawa Public Health PEEL Peel Public Health PCCHU Peterborough County City Health Unit PHU Porcupine Health Unit SMDHU Simcoe Muskoka District Health Unit SDHU Sudbury amp Distric
11. TREE OPTH classifier predicting different syndromes with relatively low NAIVE BAYES Asthma proba bilities WINNOW2 OPTH MCME Asthma To demonstrate the use of the various classifiers consider the example of a chief complaint shown in Figure 29 The Figure 29 Results for all NLP algorithms 48 ACES Manual The Science of ACES chief complaint water infront sic of my eyes describes a symptom that may underlie several different medical conditions but is likely related to ophthalmology Using the s2006 classifier ME Other is selected With the s2014 classifier there are several different results depending upon the classifier The results for the ME classifier includes a list of possible syndromes in descending order of statistical importance the most likely syndrome is OPTH general ophthalmological condition but CV cardiovascular GMED general medical admission Other and OBS related to obstetrics all merited appreciable statistical results Using the alternate classifiers BW determines the syndrome to be CV C4 5 and Winnow2 corroborate the results of ME as the syndrome OPTH and NB and MCME would classify this chief complaint as Asthma A quick read of the chief complaint makes it obvious that the problem is likely related to an ophthalmological issue The list of syndromes for the s2006 classifier is not sufficiently comprehensive to properly classify this ED visit On the other hand the s2014 classifier appears to have classif
12. and its options are described in the following sections 2 7 1 Map Features Data The main options for determining the hospital visits to be displayed on the map are found in the Data tab Figure 21 The dark gray scroll bar can be moved to reveal the remainder of the options In the bottom right corner of the tab is an arrow to reduce or expand the menu In order for data to be displayed the various options need to be chosen and Request Data there is no default data display 2 7 1 1 Mapping Style You have two choices for how you want the data can be visualized on the map as Choropleth or as Proportional Symbols Choropleth displays the variation in the number of hospital visits for each region by variation in colour darker colours represent more visits than lighter colours as defined in the legend An example of a legend for Choropleth display is shown in Figure 22 Number of User Interface Guide Emergency Department a Visits Sample Data 0 2 5 96 EN 2 5 5 96 5 7 5 7 5 10 Figure 22 Mapping Style Chloropleth Emergency Department J Visits Sample Data O 0 25 D 25 5 og ON A 5 7 5 7 5 10 96 Figure 23 Mapping Styles Proportional Symbols 2 7 1 2 Classifiers Classifications 4 Proportional Symbols on the other hand displays the number of hospital visits as a proportionally sized circle marker on each level of geography as d
13. are expanding the scope of ACES to include mental health surveillance as well as the integration of several new syndrome specific mapping tools These non traditional approaches to syndromic surveillance will allow health care agencies to plan strategies to cope with predictable increases in emergency department volume as well as identifying possible patterns related to demographics location or timing of the cases that would justify further investigation or intervention Progress is being made to create syndromes for both emergency department visits and inpatient admissions for mental health matters such as suicidal ideations alcohol intoxication addictions sleep 14 ACES Manual ACES Backgrounder disorders and opioid intoxication addictions ACES ability to monitor mental health syndromes will help fill a gap in Ontario s surveillance capabilities as there is currently no real time surveillance system for mental health issues The real time monitoring of mental health related visits and admissions will enable the immediate provision of information to health organizations and providers on the volume of patients using emergency departments for mental health issues It will also help to identify emerging substance abuse issues and improve situational awareness around mental health issues after major events and during mass gatherings This information could help improve the understanding of local mental health needs improve service delivery and
14. data associated with your data sharing agreement Figure 27 Information included in this table includes Alert Date Time Alert Class Alert Type Syndrome Geog geography Type Geog Name Clicking on the grey arrow at the top right corner of the table prompts a dropdown menu that allows you to customize which columns are displayed as well as Group By options see Figure 18 for an example of the Group By function Click on a column to list the alerts according to those parameters To further constrain your data display there is a menu of options shown on the right of the screen These include Health Unit dropdown options available here depend 36 ACES Manual User Interface Guide on your data sharing agreement or Hospital dropdown options will include those hospitals for which you have data access choosing from these options will limit or expand the data sources displayed in the main table Data Range may be changed to accommodate the dates you would like displayed and can be entered manually dd mm yyyy or from the dropdown calendars Finally all Alert Type choices are listed in the menu check the boxes of those that you wish to display The default alerts are Extreme Trend and CuSum3 as these are the alerts of most epidemiological significance and therefore concern for the typical ACES user Fifteen commonly accessed syndromes are available in this menu as well with many more choices available in the More Syndromes menu Further explan
15. days The same threshold is used as with Figure 34 ED visits for Asthma additional data CuSum1 that is CuSum2 2 3 The increase in sensitivity is best described visually the data in Table 3 and Figure 32 is presented again with the addition of two subsequent days of ED counts for the Asthma syndrome Figure 34 Table 4 You can Table 4 Comparison of CuSum alerts CuSum1 t to t 7 CuSum2 ts to ts CuSum3 o Z Alert o Z Alert Alert x 05 Feb 247 2268 16 1 0 25 025 X 06 Feb 254 2276 17 0 0 56 080 X 2300 160 050 050 X 050 X 07 Feb 234 2311 197 086 0 X 2269 161 0 56 000 X 050 Xx 08 Feb 259 2366 123 082 082 X 2276 170 085 085 X 135 X 09 Feb 240 2411 140 108 0 X 2311 197 0 55 030 X 115 X 10 Feb 273 2416 139 127 127 X 2366 123 196 226 X 34 V 11 Feb 301 246 33 182 201 328 VY 2411 140 328 555 v 8 11 v 12 Feb 270 2583 228 048 2 9 X 2416 139 105 660 v 886 v 13 Feb 280 261 6 225 0 12 261 X 2463 182 085 745 v 745 v 57 ACES Manual The Science of ACES conceptualize the difference this makes by considering the values used to calculate u on 12 Feb for CuSum2 the ED counts from the two days previous to the current value are not included in the calculation The u is not therefore skewed to a higher number by the anomalously high ED visits observed on 11 Feb Thus Cusum1 does not and CuSum2 does generate an alert on 12 Feb likewise CuSum2 also generates an alert on 13 Feb 3 4 1 3 CuSum3 Ultra Sensitivity T
16. displays a table with a list of current alerts Figure 3 For an explanation of the different alerts in ACES see Section 3 4 18 ACES Manual User Interface Guide Current Syndrome Counts Ontario Wide ue All i Respiratory Gastrointestinal 1400 1200 MY 1000 800 Visits Visits Visits 600 400 200 0 0 0 04 05 2015 05 05 2015 06 04 2015 04 05 2015 05 05 2015 06 04 2015 04 05 2015 05 05 2015 06 04 2015 Date Date Date ENVIRO ILI Es Asthma Visits Visits Visits E v 0 0 04 05 2015 05 05 2015 06 04 2015 04 05 2015 05 05 2015 06 04 2015 04 05 2015 05 05 2015 06 04 2015 Date Date Date Figure 2 Main landing page current syndrome counts for Ontario Note count for current day reflect the time of day NEW SCREEN SHOT OF THE CURRENT ALERTS Figure 3 Main landing page current alerts You can customize the information display using the gray arrow on the upper right corner of the alerts table Choices for display include alert date time alert class alert type syndrome geography type geography name Click on the box before the data element to display or turn off the information or click on the Group By Icon after the data element to display the information grouped by that parameter 19 ACES Manual User Interface Guide Figure 4 ACES default setting is to display all of these parameters by date time Clicking on the gray arrow again will collapse the options display NEW SCREENSHOT FOR
17. maintained to give you the option to compare your current data to older data and syndrome sorting which may be applicable to in some circumstances The default Classification is S2014 See Sections 2 4 1 1 Classifications 3 3 Syndrome Classification using Natural Language Processing and 3 3 5 Syndrome Validation for more information 33 ACES Manual User Interface Guide 2 7 1 4 Syndromes A list of pre defined syndromes is available to choose from under Syndromes If you have chosen S2014 under Classification these represent the current syndromes that are validated for use with ACES If you choose the S2006 classification system there are only eight syndromes to choose from See Section 2 4 1 1 Classifications for more information 2 7 1 5 Level of Geography The choices for Level of Geography include FSA the first three characters of postal code County PHU or LHIN Data are displayed according to the level of geography chosen 2 7 1 6 Data Classifications Number of Classifications and Percentage Range These options work in concert to display the data in the way that you choose Data Classifications refers to the method used to display the quantities or proportions of hospital visits With the Choropleth mapping style there are three different ways that the data can be classified Equal Interval Quantile and Standard Deviation Equal Interval will map the data in an equally distributed gradient based on the Emerg
18. necessary information El Contact the relevant health care facility for additional cases and observations El Notify the hospital s physicians PHU branch offices and potentially other PHUs depending on the scope of the outbreak and if further investigation is required PHUs can raise awareness amongst the relevant hospital s regarding the infectious disease to enable more efficient diagnosis and to implement necessary and appropriate precautions such as infection control procedures Monitor ACES on a regular basis to assess real time hospital visits for the syndrome of interest using Epicurve and Map to track changes in disease patterns 2 9 Frequently Asked Questions 2 9 1 Q1 Why did my Map request crash time out When all desired parameters are entered press the Request Data button to generate the map Click Clear Data to revert back to the default settings NOTE You can choose as wide a date range as you 40 ACES Manual User Interface Guide want but it is advised to use ranges of a month or less as a request for data from larger time periods can take a long time during which the system may time out and end your session 2 9 2 Q2 When comparing the old syndromes running in EDSS RODS to the data in ACES for the same syndromes there seems to be discrepancies why is that There are a few contributing factors to why this is the case 1 There are hospitals in ACES that have not been put in RODS If patients from
19. new user and instructions on accessing the enhanced surveillance features for advanced users Each section describes how to use ACES ranging from logging in to making full use of ACES alerting capabilities Links are provided when applicable to more detailed background information in Section 3 2 1 1 Data Collection All data for ACES is collected by participating health care facilities during the registration and triage process When a patient presents at the emergency department details describing both the patient and the visit are entered into the hospital s computer system as a patient is registered or within minutes Without any additional action on the part of the hospital staff ACES captures select information from triage records and therefore has no measureable impact on the hospital staff s workload Data elements collected by the ACES system include patient demographics e g age sex first five characters of residential postal code the date and time of the visit chief complaint s Canadian Triage Acuity Score CTAS febrile respiratory illness FRI screening results admission diagnosis if recorded and available discharge diagnosis if available mode of arrival e g ambulance and admission to intensive car if applicable To ensure identity protection and privacy requirements no direct personal identifiers e g name or health insurance number are collected by ACES and the data is sent from hospitals to KFL amp A P
20. the final diagnosis it is very difficult to diagnose underlying conditions from the presentation of non specific symptoms To ensure that ACES syndromes are meaningful aggregates of CC it is important to measure the accuracy of the diagnoses that the classifier is predicting To do this we compare the ACES data to retrospective acute care gold standard data available through the National Ambulatory Care Reporting System NACRS the national 49 ACES Manual The Science of ACES repository for acute care data in Canada Data collected by NACRS is rigorously maintained and checked for accuracy It is not available in real time and therefore all validations are made with retrospective data Validation of the syndromic categorization is made by correlating the daily number of emergency department visits categorized into a specific syndrome with its corresponding time series of diagnostically defined data compiled by NACRS Standardized diagnostic codes are made by Canadian hospitals using the International Classification of Diseases 10 CA ICD 10 CA coding system ICD 10 CA codes are recorded by the attending physician or other health care professional after the patient has been examined and treated when applicable Hospital coding personnel are trained to ensure consistency in coding between hospitals and there are specific codes for disease injury causes of death as well as external causes of injury and poisoning When comparing AC
21. there are regular reviews of the ACES policies and procedures to ensure they are in alignment with PHIPA and any other applicable privacy legislation KFL amp A Public Health enters data sharing agreements with all of its health unit and hospital partners on ACES All of these policies and procedures ensure that all health information extracted stored and processed using ACES is protected 2 2 Login Page Registered users can access ACES at http aces kflaphi ca Logging in requires both username and password Users will be given access to the data according to their data sharing agreements For more information contact the ACES team at kflaphi kflapublichealth ca 2 3 Main Landing Page Upon logging in you will be directed to ACES main page This page provides an overview of recent provincial emergency department activity as well as current alerts On the top half of the page there are a series of six charts that display 1 all acute care hospital visits across Ontario for the past sixty days as well as visit counts for five syndromes of common public health interest that is 2 respiratory 3 gastrointestinal 4 environmental ENVIRO 5 influenza like illness ILI and 6 Asthma Figure 2 These graphs are not interactive and permit the user a quick overview of the whole province for situational awareness regarding acute care use in general and the relative activity for the five syndromes shown The bottom half of the page
22. your HU went to these hospitals they would not be counted in RODS 2 ACES has the ability to assign any Ontario postal code to a HU but that was not the case in RODS RODS did not have the power for us to just include all of our postal code iterations so whenever a new HU was added we used their historic data to add roughly the top 20 codes in terms of frequency so that a majority of visits would be counted but there was always the possibility that some visits would not be counted Suffice to say ACES is more accurate and is counting each and every visit and assigning it to one of our 36 health units If the patient is not from Ontario they are still counted but would be under the hospital count as a non local resident These instances you can also count in ACES but you weren t able to do that in RODS 2 9 3 Q3 If we use the new syndromes in our reporting is there an easy way to get historical data for comparison While we did provide a reporting function in EDSS RODS that allowed you to download two years of historic data for your local hospitals we are not offering that in ACES We have built ACES in a way that this should not be necessary Graphing HU data and data by hospital and then downloading a jpeg file or screen capture of the graph you created should erase the need for data dumps and having to create graphs on your own in Excel 41 ACES Manual User Interface Guide 3 THE SCIENCE OF ACES 3 1 A Short History of Syndromic Surveillanc
23. 05 13 11 gt 2015 05 14 11 4 Emergent Admission 61 4 F 34 b 2015 05 13 18 gt 2015 05 14 16 4 M 27 b 2015 05 13 18 gt 2015 05 14 9 Age Gender FSALDU Hospital Syndr Choose Columns v Date v Time co Admission Type Age N v Gender FSALDU Hospital v Syndrome Complaint CTAS EMS Arrival vi Patient LHIN Figure 18 Line Listings AD Group By options 2 6 1 Line Listings AD Tools Menu The Tools menu on the right of the screen shows the available options to customize the hospital admissions that are displayed in the line listings Figure 19 The options available under Health Unit and Hospitals depend on your data sharing agreement Data Range can be entered manually mm dd yyyy or by choosing from the dropdown calendars Choose from the options for Gender Age and CTAS and select Submit to display the selected admissions Reset will return all settings to their defaults 2 6 2 Line Listings AD Advanced Menu The Advanced tab for Line Listing AD gives you the same options that are available for with the Advanced tab for Line Listings ED Please refer to Section 2 5 2 Line Listings ED Advanced Menu for instructions 31 ACES Manual Tools Advanced Health Unit APH All v Hospitals Date Range Date From pore EX mk aM EL JG LE E Date To ms LLL brane fe 05 13 2015 05 14 2015 Gender All Male Female Age All
24. 4 for more information When you are displaying a specific syndrome the option of Normalize is available and the number of acute care visits for that syndrome will be displayed as the percentage of total visits Figure 9 Additionally several options appear below the statistical display options when you choose Normalize click on the red or Normalize Moving Average Standard Deviation Off On None rj 14 Max None Std 1 Sid 2 black circles to display the on the epicurve Figure 9 Normalize and other statistical options 2 4 2 Epicurve Display Options Advanced Tools Advanced Choosing the Advanced tab at the top of the options menu gives you Locality more choices for optimizing the display of your data Figure 10 The Local PHU Patients options include Locality which determines the geography of either the Hospitals Province Wide patients or the hospitals displayed It is important to note that you must FSA choose Local for at least one of the options Choosing Local PHU ma Patients from the Patients dropdown menu will display only patients Reset with postal codes from the PHU s local region Province Wide displays a Submit all Ontario patients Outside of PHU Patients displays all patients from Figure 10 Epicurve display options outside the PHU s region as determined by postal code Likewise ud n s 23 ACES Manual User Interface Guide patients from specific Hospital geography can be displayed choos
25. 7 1 4 Syndromes tee es df eue Eee 34 2 7 15 Levelof Geography iore ee blc ee eere e n beers 34 2 7 1 6 Data Classifications Number of Classifications and Percentage Range 34 2 7 1 7 Date Range Gender Age Group Age Range sesse see ER ee ener 35 2 1 2 Map Features Layers t ien 35 2 7 3 Map Features Map AE esee 36 2 8 Alerts d I A GEE EEUE 36 2 8 1 AlertingflliSxocols WEA V AF Es esse Gee eke ee 38 2 8 2 Suggested Procedure for Respiratory or Gastrointestinal Outbreaks in the Community 40 2 9 Frequently Asked Questions ce cccccssssseececceaneececccsessececeesaseeceessuesseceessuaseeseesauanseeeessuaaeeeessaganss 40 2 9 1 Q1 Why did my Map request crash time out essen 40 2 9 2 Q2 When comparing the old syndromes running in EDSS RODS to the data in ACES for the same syndromes there seems to be discrepancies why is that ss RE EE 41 2 9 3 Q3 If we use the new syndromes in our reporting is there an easy way to get historical data for or on TNT m 41 ele EEN EE EE EE NE O E 42 3 1 A Short History of Syndromic Surveillance ees sees see ee ee ee nennen 42 3 2 Th Development TT Ea UT 43 3 3 Syndrome Classification using Natural Language Processing ees sees ee ee ee ee ee Ee 45 3 3 1 NENNE MA N N EA OE N EE 47 3 3 2 IV Ve OT El OY JE OE MEE 47 3 3 3 Bala
26. Acute Care Enhanced Surveillance ACES User Manual O KFL amp A Public Health 2015 v01 6 04 15 Kingston Frontenac Lennox amp Addington Public Health 6 4 2015 Table of Contents 1 ACES Backeroufider sisie seder Seed vsu Gee os ode ee ed ee ee ode ee oe Gegee Eg ee Ee 7 dl ae FOUN tonto ACE EE EE 7 1 2 ACESJSDDIICAUIOPRIS ese Ee ee He T Ge GE ne Ge EE EER EE ee 9 1 2 1 Ta LEV EL vi RE EE EO OO EE OE OE 9 1 2 2 Reporiable disease de teel DI Ede ee ee n ee ee de 10 1 2 3 MENN 11 1 23 1 68 620 SUMMIT esse ei utm 11 1 2 3 2 2015 Pan Parapan American Games EA 12 1 2 4 Emergencies and Extreme Weather WB eee 12 1 2 4 1 Kingston EE EN N AT ET EE ERK 12 1 2 4 2 Midland Tornado Gill ee ee ee oe EE oe ese ee ee EE 13 1 2 5 Surveillance After Drug Policy Change see sesse ER ee ee ee RR ee ee ee RR ee RE ee ee ee 13 1 2 5 1 Methadose gm sesse ss esees sesse BO AF esse EE ee ee 13 1 2 5 2 Delisting of OxyContin SG eke dee ek GN Se ek ee ee ee 13 1 2 6 Asthma Vo GEEN EE EEUE NEE EE EN 14 1 3 Future Directions esse ss reses see EE ee Esse NE 14 2 User Interface Guider ED W ANN ee 16 21 Data Collection and System OVerVi W ees sees ee a ee ee RR ee ee RE ee ee ee ee Ee 16 2 1 1 Data Keen ee W ss se eene nnne nnn nnnne 16 2 1 2 Lers TR sm EER WERE OE nnne 16 2 1 3 Data Sec ind PRE 99 1er nnn nnnm nnns 18 2 2
27. Classifier Bucket 2014 v All B Syndrome Algorithm All v ME CTAS 1 2 3 4 5 All Figure 14 Line Listings ED Tools menu Under Classifier the dropdown menu gives the option to display syndrome data that was classified using the S2006 algorithms or the S2014 algorithms See Sections 2 4 1 1 Classifications and 3 3 Syndrome Classification using Natural Language Processing for a discussion of these options CTAS Choose what level of acuity you wish to display 2 5 2 Line Listings ED Advanced Menu The Advanced tab gives you more options regarding the Locality or geography of either the Patients or the Hospitals Figure 15 The extent of options you will have in this tab depends on your data sharing agreement It is important to note that you must choose Local for at least one of the options Choosing Local PHU Patients from the Patients dropdown menu will display only patients with postal codes from the PHU s local region whereas Province Wide displays all Ontario patients and Outside of PHU Patients displays all patients from outside the PHU s regions accessing local PHU hospitals The patient s PHU is determined using the residential 28 ACES Manual Tools Advanced Locality Patients Local PHU Patients Hospitals Province Wide FSA All Figure 15 Line Listings ED Advanced menu User Interface Guide postal code associated with the hospital record To display local patients accessing specific
28. D NOTES FOR SPECIFIC FEATURES 3 4 2 2 SPC Rule 2 Bias A Bias alert is generated if the current visit count is the ninth in a row where all nine counts are on the same side of the mean that is nine daily counts in a row are either greater than the mean or less than the mean In ACES an alert will only be triggered if the nine counts are all above the mean Assuming the data is normally distributed the probability that a Bias alert will be triggered is 0 496 An example of a Bias alert is shown in Figure 36 Bias alerts are sensitive to very subtle changes in daily counts as it indicates circumstances where a slight change in counts is observed over several days even when the INSERT SCREENSHOT OF BIAS ALERT HERE Figure 36 SPC Rule 2 Bias individual counts may not be dramatic ally higher than the mean i e within one standard deviation Means and standard deviations for Bias are calculated from the previous fourteen days so it is also conservative its sensitivity in that the previously higher counts are used in the calculation of the moving average The chance of nine counts in a row being greater than the mean is very low and therefore may indicate a true trend of increasing mean emergency department visits In other words a trend of increasing visits could signify that there is a health effect of public health consequence in occurrence 60 ACES Manual The Science of ACES 3 4 2 3 SPC Rule 3 Trend A Tr
29. E SU GE ON T m HE Ne 56 Table 4 Comparison of CuSum alerts sa see se lan oe ee ee oe oe eke Ke Ke Ke Re Ge Pe ee ee eke Ee ee 57 Vi ACES Manual Table of Contents 1 ACES BACKGROUNDER cute Care Enhanced Surveillance ACES provides real time epidemiological surveillance for Ontario ACES monitors triage emergency visits and inpatient admissions to hospital records from over 70 of Ontario s acute care hospitals Records are monitored as the patients are being treated giving real time situational awareness for disease outbreak and other potential health risks Hospital visits are monitored with a sliding scale of specificity from a province wide assessment to our smallest level of geography the FSA forward sortation area the first three characters of Canadian postal codes The temporal and spatial capabilities built into ACES enable public health to be better informed on the health of the community which in turn can improve public health protection and prevention initiatives This manual has been prepared to familiarize you with ACES in the following ways 1 as an introduction to ACES capabilities 2 to provide you with practical assistance in using ACES and 3 to provide you with a detailed scientific background for the theory and technology used in ACES You can begin reading Section 1 for an overview of ACES applications and future directions In Section 2 a guide to the ACES interface is provided
30. ES data to the NACRS datasets Pearson correlation coefficients are used to describe the relative similarity between the datasets by day for and weekly bi weekly or twenty eight day moving averages In some cases where there are relatively few observed cases monthly cumulative sums must be used in order to have sufficient sample sizes for comparison Pearson correlation values between 0 and 0 25 indicate poor correlation values from 0 25 to 0 5 indicate moderate correlation values from 0 5 to 0 75 indicate good correlation and values 0 75 indicate excellent correlation between datasets These numbers roughly correspond to the percent of variation in one dataset that can be explained by variation in the second dataset 3 3 5 1 Examples of Syndrome Validation ME x fever 3 Word clouds were used in our validation process to show our users which free text words were most common in each of our O Q Od validated syndromes The word cloud shown in Figure 30 represents the relative frequency of single words abstracted from chief congestioi arache a eal 7 D C2 complaints that relate to respiratory related health conditions The words are selected Figure 30 Word cloud for Respiratory 50 ACES Manual The Science of ACES through the NLP technigues described in NACRS Section 3 3 Thus the words cough throat RESP 2014 Age 0 120 Gender All Corr coef 0 952 and sore occur more
31. Equation 6 Equation 7 Equation 8 Equation 9 where u and o are the mean and is the standard deviation for the previous fourteen days respectively SPC alerts in ACES are generally associated with only the UCL calculations as emergency department counts that are above the means are of epidemiological interest in determining possible public health threats 3 4 2 1 SPC Rule 1 Extreme An Extreme alert is the most intuitive of the SPC alerts it is generated if the current day s emergency department visit count is greater than three standard deviations from the mean i e current count UCL3 Assuming a normal distribution for the data the probability that the one data point will be more than three standard deviations than the mean is 0 1396 therefore a current count beyond three standard deviation of 59 ACES Manual ADD Example of EXTREME alert as screen shot from ACES Figure 35 SPC Rule 1 Extreme The Science of ACES the mean has only a 0 13 chance of being due to chance alone This is epidemiologically relevant as this alert would likely showcase a very sharp incline compared to historic data and require immediate investigation to contain the issue It would also be useful for hospital administration staff as this type of spike may require more staffing or diversion of patients to other hospital sites depending on the scale of the spike An example of an Extreme alert is shown in Figure 35 AD
32. FL amp A Public Health began monitoring ED visits and admissions to local hospitals that could be related to the event using ACES Specifically emergency department visits related to respiratory ailments and asthma were monitored for aberrations in normal expected volumes In particular ACES was used to monitor for increases in emergency department visits for smoke inhalation and carbon monoxide exposure Though no significant increases were detected the ACES system was able to help quickly assess the extent of the public health threat and inform hospitals of the potential degree of expected patient volume influx ACES improved the overall situational awareness and preparation throughout the duration of the emergency In addition to ACES the Public Health Information Management System PHIMS was used for enhanced situational awareness during the fire PHIMS enables the assessment of wind patterns and air quality Using ACES and PHIMS public health was able to determine the geographic locations in Kingston that were at the highest risk for exposure to the effects of the fire including determining the most vulnerable individuals in this area who should be given special consideration during evacuation efforts 12 ACES Manual ACES Backgrounder 1 2 4 2 Midland Tornado In times of unexpected disaster the ACES can assist first responders and emergency operations personnel On June 23 2010 a F2 Tornado hit Midland ON Winds in the area reached betw
33. I INF INJ INS INT 70 ACES Manual allergic reaction angioedema not bee sting asthma wheeze difficulty breathing SOB human animal bug not tick related bronchiolitis RSV burns chemical and thermal electrical shock coronary artery disease chest pain pericarditis effusion myocarditis endocarditis c difficile cellulitis non wound infection non abscess congestive heart failure carbon monoxide exposure or other gases sulphur etc concussion head injury chronic obstructive lung disease croup PIV cardiovascular excludes MI and strokes includes peripheral vascular disease dehydration dental pain infection trauma to tooth etc rash undifferentiated lesion wart diabetes and its complications electrolyte imbalance hyperkalemia hypomagnesium hyponatremia related to ears nose throat surgery non infectious tinnitus not in GNSURG heat stroke heat syncope heat exhaustion cold frost bite hypothermia alcohol and complications intoxication addiction withdrawal or end organ damage falls undifferentiated foreign body ingestion nose to anus febrile neutropenia gastroenteritis Guillain Barre syndrome flaccid paralysis GI bleed upper and lower epistaxis hemoptysis general medical admission other unconscious weakness unwell chronic diseases general surgical admission appendicitis cholecystitis bowel obstruction gynecological bleed hysterectomy PID undifferentiated headac
34. IN ME 68 Acute Care Enhanced Surveillance Balanced Winnow text classifier C4 5 Decision Tree text classifier Centre for Disease Control Complaint Coder cumulative sum dissemination area Early Aberration Reporting System Emergency department febrile respiratory illness Forward sortation area local delivery unit Hazard Identification Risk Assessment Health Protection and Promotion Act influenza like illness Kingston Frontenac Lennox amp Addington Local Health Integration Network maximum entropy ACES Manual MOHLTC MRDx NACRS NRC NLP PHAC PHIPA PHIMS PHU RODS SARS SPC SS UCL Ministry of Health and Long Term Care Most responsible diagnosis National Ambulatory Care Reporting System National Research Council Natural language processing Public Health Agency of Canada Personal Health Information Protection Act Public Health Information Management System Public Health Unit Real time Outbreak and Disease Surveillance Severe Acute Respiratory Syndrome Statistical Process Control Surveillance system Upper control limits Appendices APPENDIX B GLOSSARY alert CTAS chief complaint CuSum 1 2 and 3 EDSS epicurve FRI screening tool forward sortation area local delivery unit FSALDU ILI Mapper RODS syndrome 69 ACES Manual An aberration in data that generates an automatic real time email notification of alert to the syndromic surveillance team The typ
35. LDHU 73 ACES Manual Blind River District Health Centre BRH Lady Dunn Health Centre WAW Sault Area Hospital SAH St Joseph s General Hospital QEL Chatham Kent Health Alliance Chatham CKHA Chatham Kent Health Alliance Wallaceburg CKHAW Lakeridge Health Bowmanville LHB Lakeridge Health Oshawa LHO Lakeridge Health Port Perry LHP Rouge Valley Health System Ajax and Pickering APG Cornwall Community Hospital Glengarry Memorial Hospital HGMH Hawkesbury and District General Hospital Winchester District Memorial Hospital Grey Bruce Health Services Lion s Head LHSH Grey Bruce Health Services Markdale MDSH Grey Bruce Health Services Meaford MFSH Grey Bruce Health Services Owen Sound OSSH Grey Bruce Health Services Southampton SSH Grey Bruce Health Services Tobermory TSH Grey Bruce Health Services Wiarton WSH Hanover and District Hospital HADH South Bruce Grey Health Centre Chesley CSH South Bruce Grey Health Centre Durham DSH South Bruce Grey Health Centre Kincardine KSH South Bruce Grey Health Centre Walkerton SWH Campbellford Memorial Hospital Haliburton Highlands Health Services Northumberland Hills Hospital Ross Memorial Hospital Halton Healthcare Services Georgetown GEO Halton Healthcare Services Milton MIL Halton Healthcare Services Oakville OAK Joseph Brant Memorial Hospital JBMH H
36. Manual Perth amp Smiths Falls District Hospital Great War Memorial Site Perth GWMH Perth amp Smiths Falls District Hospital Smiths Falls Site SFH Niagara Health System Douglas Memorial Hospital Site DMH Niagara Health System Greater Niagara General Site GNG Niagara Health System Port Colborne General Site PCG Niagara Health System St Catharine s General Site SCG Niagara Health System Welland Hospital Site WHS West Nipissing General Hospital WNGH amp WNG North Bay General Hospital NBGH Mattawa General Hospital MH amp MGH Atikokan General Hospital ATGH Lake Of The Woods District Hospital LWDH Riverside Health Care Facilities Inc RHF Sioux Lookout Meno Ya Win Health Centre SLH Queensway Carleton Hospital QOQ Trillium Health Partners Credit Valley Hospital CVH Trillium Health Partners Mississauga Hospital MISS William Osler Health System Brampton Civic Hospital BCH Peterborough Regional Health Centre PRH MICs Group of Health Services Anson General Hospital AGH MICs Group of Health Services Bingham Memorial Hospital BMH MICs Group of Health Services Lady Minto Hospital LMH H pital de Smooth Rock Falls Hospital FSR H pital Notre Dame Hospital NDH Hornepayne Community Hospital HPH Sensenbrenner Hospital SBH Timmins amp District General Hospital TDH Georgian Bay General Hospital Midland Site GBGH Muskoka Algonquin Hea
37. SymbolS sesse sees esse ee dee dee ee ee ee enne 33 Equal Interval Classifications 6 Percent Range Oto 24 ees see ee ee ee RR ee Ee ee 34 Quantiles Number of Classification SA ese ee ee ee ee ee ee ee ee ee ee ee ee ee ee ee ee ee ee trenes 35 Standard Dev D E ee ee 35 Sacer GF Vs es ee ee ee ee ee 36 Example of a Trener af WA eie ee ey kx Ee Ee udes 37 po WER AA EE VERE EEUE EEN 48 Word cloud fo ir Oen nnn 50 Calculation of Pearson correlation coefficient for Respiratory rrurrrrnrnnrrrnnnnnrrnnnnnrrnnnnnerrnrnnene 51 Ep Or Asthma o EE EER EE OE N RE EE 55 TImelineWelcuSum ale ee ee EE ek EE dk nnne EG ee 56 ED visits for Asthma additional data RE ee ee AR EE ee ee ee ee EE ee ee ee ee 57 OPE Ne I Je e 59 PC RUC 2 SBI 60 PC RUC EE Me ee ee oe ee E E De E AE 61 SPC RUG Ol Edge example LS 62 PC RUle 4 0n Edge example TT 62 PU NNN 62 SPEL Rule G OAO dele 63 ACES Manual Table of Contents Figure 42 SPC Rule 7 Blissful Ignorance sees ss ee ee ee ee Re ee ee ee RR ee RE ee RE ee RR ee RR EE 63 Figure 43 SPC Rule 8 No Middle Ground ees ees ee ee ee ee Re ee ee ee RR ee ee RE ee RR ee RR EE 64 PENNEN 65 Table 1 ICD 10CA codes used to extract NACRS data for validation of Respiratory 51 Table 2 Word Cloud and validation statistics for common syndromes eee 52 T
38. U a i 2015 05 15 09 51 03 CUSUM CUSUM3 PAIN PHU time surveillance capabilities mathematical Er Re E CAS mx 2015 05 15 04 23 00 CUSUM CUSUM3 INJ PHU models are used to calculate potentially aberrant 2015 05 14 17 34 45 CUSUM CUSUM3 PAIN PHU patterns in hospital visits and these are presented ETE AED pmo kels 2015 05 14 09 01 54 CUSUM CUSUM3 PAIN PHU as alerts Alerts are then communicated to public 2015 05 14 01 57 19 CUSUM CUSUM3 INJ PHU 2015 05 13 16 36 00 CUSUM CUSUM3 PAIN PHU health officials or other health care service E Ad TE TT mem T 3 2015 05 13 08 05 00 CUSUM CUSUM3 PAIN PHU administrators to examine and confirm The 2015 05 13 07 56 08 CUSUM CUSUM3 PAIN PHU ability to detect aberrant hospital use patterns in ae me ase MEUM de 2015 05 12 16 26 23 CUSUM CUSUM3 PAIN PHU real time may for example enable the early 2015 05 12 15 12 48 CUSUM CUSUM3 PAIN PHU 2015 05 11 17 50 08 SPC Trend OTHER PHU detection of a disease outbreak Alternately an 2015 05 10 17 36 22 CUSUM CUSUM3 PAIN PHU id if bi ith th T 2015 05 09 12 30 33 CUSUM CUSUM3 INJ PHU a ert may enti y pro ems wit t e co ection 2015 05 08 13 32 56 CUSUM CUSUM3 INJ PHU transmission or storage of data Total Items 19 CSV Export Page Size 25 v 1 4 After navigating to the Alerts page you will see a KFLA 2015 isti Figure 27 Alerts main table large table listing all the alerts that have been ps erts main table recorded for the previous two weeks for the
39. additional information to aid the investigation such as identifying additional cases The Hospital Infection Control Practitioner ICP or designate Infection Control staff member is the contact for the PHU at the hospital The call Nurse on the floor may be contacted for information pertaining to admitted patients Please ensure that all relevant information e g date time of visit hospital number of cases comparison to seasonal trends lab data if available etc is ready at the time of the call to ICP or designate to facilitate efficient investigation NOTE Lines of communication between health unit and the health care system is an important part of this process It is suggested that all users have a predefined internal process to connect with these outside stakeholders to efficiently coordinate investigation and follow up If deemed appropriate contact the hospital s laboratory for acute care visits or admitted patients Public Health Laboratory for samples related to outbreak investigation and tests outlined on the Ontario Public Health Laboratory Test Directory and or private laboratories to request number of test requisitions positive results preliminary or otherwise negative results type species identification etc that may be associated with the syndrome under investigation Continue to monitor ACES for real time emergency department visits and admissions of the syndrome of interest particularly the spread of disease g
40. adone specific emergency department visits and admissions Public health professionals were able to conclude that methadone overdose numbers were stable over the weeks before and after the initiation of the new program ACES provides on going monitoring of methadone related emergency department visits 1 2 5 2 Delisting of OxyContin OxyContin is an opioid analgesic that is prescribed to treat moderate to severe pain it is safe when used as prescribed but its misuse can lead to addiction or overdose OxyContin was removed from the list of drugs funded by the province of Ontario in March 2012 and OxyNEO considered more difficult to abuse is being prescribed instead Controls on prescribing OxyNEO also became stricter around this same time These changes were intended to try to reduce the harms and deaths associated OxyContin abuse To 13 ACES Manual ACES Backgrounder help determine whether these policy changes had any impact on population health ACES was used to monitor changes in the visit and admission patterns regarding narcotics abuse Weekly reports were generated to provide evidence regarding the effectiveness of the new OxyContin policy and to identify any potential unintended consequences Analysis of ACES data is ongoing but initial results indicate that narcotic related visits and admissions remain stable from values before the new policy s inception 1 2 6 Asthma Approximately 13 of Canadian children aged 0 to 14 years hav
41. age record and Most Likely Syndromes Figure 16 see explanation below Most Likely Syndromes Each individual hospital visit is classified by ACES into a syndrome of e Saana MAXIMUM ENTROPY Other 100 medical significance This process is described in greater detail in S2014 Classifier Syndrome Sections 2 4 1 1 Classifications and 3 3 Syndrome Classification using a XB E Other 296 Natural Language Processing Briefly ACES employs several oe ce WOUND 1 classifying algorithms that sort the hospital visits into pre defined EE ails syndromes based on the occurrence of words or phrases in the chief pisk NAIVE BAYES MH complaint the results of these classifying algorithms are shown on WINNOW2 MH MCME MH this page This page should be consulted before contacting the ACES administrator if you suspect a problem with the syndrome Figure 16 Most Likely Syndromes from an individual line listing 29 ACES Manual User Interface Guide classification for a line listing For example in Figure 16 the percentage likelihood is for several different syndromes are shown for maximum entropy from the S2014 Classifier If the classification of the chief complaint not shown for this hospital visit were in question you could first check here to see if there were other likely syndromes indicated using different algorithms Remember you have the option of the algorithm used to sort your data using the Tools menu for Epicurves Section 2 4 1 1 Cla
42. al Hospital LACGH The inclusion of both Std 1 and Std 2 make this a busy graph and can be deselected if desired To return the graph to its default display settings click Reset 2 4 4 Additional Features 2 4 4 1 Focus Chart A helpful feature for data visualization and exploration is located below the Epicurve Focus Chart This chart allows the user to adjust the graph to focus on a narrower time period without having to create new graph When you move your cursor over the Focus Chart it becomes a large cross Clicking on the desired date s on the Focus Chart creates a window that can slide back and forth to either widen or narrow in on the date range Figure 13 The time range chosen will be displayed in the larger epicurve above the Focus Chart box Focus Chart 1831 Mar 01 2015 Mar 05 2015 Mar 08 2015 Mar 12 2015 Mar 15 2015 Mar 18 2015 Mar 22 2015 Mar 26 2015 Mar 31 2015 A Download Chart Figure 13 Focus Chart tool 2 4 4 2 Data Point Display Hovering the cursor over a point on any of the lines on the epicurve i e visits moving average standard deviation will open a box that displays the date and visit counts for that data point When viewing a data point from the visit counts red solid line click to open a window displaying the line listing of the emergency department visits for that day A description of the information and options available for line listing displays is found in Section 2 5 Line Listings
43. amilton Health Sciences Corporation Hamilton General Hospital and Urgent Care Centre HAH Hamilton Health Sciences Corporation Juravinski Hospital and Cancer Centre HEN Hamilton Health Sciences Corporation McMaster Children s Hospital MCM St Joseph s Healthcare Hamilton STO1 Quinte Healthcare Corporation Belleville BGH Quinte Healthcare Corporation North Hastings NHH Quinte Healthcare Corporation Prince Edward County PEC Quinte Healthcare Corporation Trenton TMH Hotel Dieu Hospital HDH Kingston General Hospital KGH Lennox amp Addington County General Hospital LACGH Bluewater Health Sarnia BWH Charlotte Eleanor Englehart Hospital of Bluewater Health Petrolia BWHP Brockville General Hospital BRGH Carleton Place and District Memorial Hospital CCC Feb 2013 Jul 2011 Sept 2014 Jul 2011 Jun 2013 Jun 2013 Sep 2013 Sep 2013 Sep 2013 Sep 2013 TBD Apr 2010 TBD TBD Nov 2011 Nov 2011 Nov 2011 Nov 2011 Nov 2011 Nov 2011 Nov 2011 Nov 2011 Nov 2011 Nov 2011 Nov 2011 Nov 2011 TBD TBD TBD TBD Oct 2011 Oct 2011 Oct 2011 Oct 2011 Sep 2011 Sep 2011 Sep 2011 Sep 2011 Sep 2005 Sep 2005 Sep 2005 Sep 2005 Sep 2005 Sep 2005 Sep 2005 Jun 2013 Jun 2013 Jul 2009 Oct 2014 Appendices Health Unit Key Hospitals within Health Unit ACES start Below NRPH NBPH NWHU OPH PEEL PCCHU PHU SMDHU SDHU TBDHU THU TPH 74 ACES
44. and is the standard deviation Consecutive daily ED counts are then compared using the CuSum formula S max fo S g sb A Equation 2 where the current CuSum S is the maximum value of zero or the sum of the previous day s CuSum 5 4 and zy if S 2 3 an alert is triggered This threshold is generally accepted as optimal but a lower value would increase sensitivity Sensitivity can also be changed by adjusting the time period used to calculate the mean and standard deviation to which daily counts are standardized CuSum alerts in ACES describe three levels of sensitivity CuSum1 mild sensitivity CuSum2 medium and CuSum3 ultra 3 4 1 1 CuSum1 Mild Sensitivity Consider the following example the daily counts for ED for visits classified as the syndrome Asthma are shown in Figure 32 These data represent the first two weeks of recorded ED counts and it is unclear if the ED counts observed on 11 Feb are high 350 in comparison to what may be expected by zm chance Using the CuSum alert method described above a threshold of CuSum 23 250 is used Mean u standard deviation o and CuSum score z are calculated beginning Feb 04 the first date with seven ED Visits previous daily ED counts Table 3 Using equations 1 and 2 z and CuSum are calculated for subsequent days Note that CuSum values approach zero when hospital visit counts generally approach the mean This method is effectively a measure
45. ant but it is advised to use ranges of a month or less as a request for data from larger time periods can take a long time during which the system may time out and end your session 2 7 2 Map Features Layers The second tab on the right side of the map screen Layers allows you to add additional information to your map For example adding Daycares will add all daycare facilities on the map using the symbol indicated Other Layers include pharmacies PHUs long term care facilities schools and traffic cameras Note that you can display multiple layers the map may become cluttered press Clear Layers to clear all layers added 35 ACES Manual User Interface Guide 2 73 Map Features Map The Map tab enables different viewing options for the base map For example the default setting is Topo or topographical which is a mapping style that describes the surface shapes and features Other choices include Street Satellite Hybrid Gray Oceans National Geographic and OSM OOpenStreetMap contributers These descriptions are generally self explanatory and are best explored while using ACES You can also change the layer and border colours in this tab Once you have finished customizing the map press Update to change the base map Reset will return the map to its default settings 2 8 Alerts Alert Date Time Alert Class Alert Type Syndrome Geog Type One of the main advantages of ACES is its real 2015 05 15 10 00 58 CUSUM CUSUM3 PAIN PH
46. ation of these alerts and the mathematical models used to calculate them can be found in Section 3 4 and syndrome classification can be found in Section 3 3Error Reference source not found For more information regarding a specific alert click anywhere on its line listing in the table a graph of the data triggering the alert will be displayed in new window Depending on the type of alert chosen various statistical information will be displayed on the graph For example the graph in Figure 28 shows Graph Data a Au Average UCL G MUCL2 UCL3 LOL LCL2 LCL3 15 1293 70 50 50 40 30 20 Visits 10 20 30 42 1544 04 1 4 2015 0449 2015 04 25 2015 05 01 2015 05 07 2015 0543 2015 0519 2015 Date Figure 28 Example of a Trend alert 37 ACES Manual User Interface Guide the data that triggered a Trend alert Briefly a Trend alert is triggered by the sixth day in a row with increasing hospital visits The legend at the top describes the various information given black circles and line for the data points the average in this case a fourteen day moving average green dotted lines showing the upper control limits UCL UCL2 and UCL3 representing the standard deviation two standard deviations and three standard deviations respectively and gray dotted lines indicating lower control limits LCL LCL2 and LCL3 representing the standard deviation two standard deviations and three standard deviations respect
47. ations reportable diseases respiratory infection non croup non bronchiolitis bacteremia SEPSIS smoke inhalation chemical gases social admission test results blood or diagnostic imaging xray US biopsy transfusion tube change thoracic pneumothorax ticks toxicology not alcohol or opioids withdrawal substance abuse chemical exposure trauma from a MVC ATV trauma from another means fall etc gunshot or stab violence assault sexual assault rape urological stones prostate UTI vomiting alone NORO like illness not secondary to chemo or with other symptoms ACES Manual Appendices APPENDIX D HPPA REPORTABLE DISEASES IN ONTARIO Acquired Immunodeficiency syndrome AIDS Acute Flaccid Paralysis Amebiasis Anthrax Botulism Brucellosis Campylobacter enteritis Chancroid Chickenpox Chlamydia trachomatis infections Cholera Clostridium difficile associate disease CDAD outbreaks in public hospitals Creutzfeldt Jakob Disease all types Cryptosporidiosis Cyclosporiasis Diphtheria Encephalitis including 1 Primary viral 2 Post infectious 3 vaccine related 4 subacute sclerosing panencephalitis 5 unspecified Food poisoning all causes Gastroenteritis institutional outbreaks Giardiasis except asymptomatic cases Gonorrhoea Group A Streptococcal disease invasive Group B Streptococcal disease neonatal Haemophilic influenza b disease invasive Hantavirus Pulmonary Syndrome Hemorr
48. ber to click on the green Submit button when all the options have been selected The result of these choices is shown below for May 12 2015 from KFL amp A Public Health Figure 11 Your display will be different but should represent measurable visit counts as you have chosen all syndromes If you like choose a specific syndrome or change the dates of the data range Remember to press Submit after making any changes To return to default display settings click Reset In the chart the red solid line displays the visit counts over the specified time period The dotted line displays the seven day moving average the dashed line is one standard deviation above the moving average and the dash dotted line displays two standard deviations above the moving average You do not need to press Submit after changing moving average or standard deviation options they should update automatically Note the 24 ACES Manual User Interface Guide seven day lag period before the moving average and standard deviations are displayed these calculations require seven days of data before they can be computed Kingston Frontenac and a Standard Deviation None 7 14 Max None Std 1 Std 2 Tools Advanced Lennox and Addington Health Unit Health Unit Hospitals KFLA v Q GAI7dayAVG AISTDDEV1 AISTDDEV2 ST am Date Range Date From z 04 21 2015 05 12 2015 Gender All Female Age All Classifications Classifier Bucket 2014 v A
49. cal acute care facilities where resources are often already overextended and reducing transference of disease Several factors are essential for an efficient syndromic surveillance system 1 routinely collected for other reasons e g administrative to reduce costs and staffing 2 recorded and accessible electronically and 3 available in near real time The described and defined syndromes need to be validated using traditional data sources e g laboratory results physicians diagnoses In general syndromic surveillance systems based on emergency department records are easily made compliant to these constraints Of all possible sources for electronic medical records acute care triage records submitted at patient registration provide the most potential for truly real time surveillance emergency departments are open twenty four hours per day every day of the year allowing a wide range of potential patients to access acute care services for an equally wide range of reasons 3 2 The Development of ACES In September of 1999 researchers at the University of Pittsburgh initiated the prototype Real time Outbreak and Disease Surveillance RODS system to monitor and characterize illness outbreaks using data from electronic emergency department triage reports The success of this system combined with the timeliness of its initiation immediately before 9 11 helped it evolve into a much larger surveillance network It was used to monitor illne
50. character generally refers to the province or territory and geographical information becomes more focussed with each character The postal code is specific to a unique geographical unit with on average 400 to 700 residents Influenza Like IlIness Activity Level Indicator This is a feature of the map function where when the user selects ILI as their syndrome a colour classification system layers over the map in the areas of the region experiencing Influenza based on number of cases detected for the hospitals in that region Stands for Real time outbreak and disease surveillance the biosurveillance system employed in by its creators at the University of Pittsburgh for the 2002 Salt Lake City Olympics For a full history of RODS as well as its current applications please view the RODS Laboratory website at www rods pitt edu A grouping of symptoms to represent a single disease e g asthma COPD similar diseases e g gastroenteritis sepsis or particular kind of adverse health outcome e g toxicology environmental Syndromes can be adjusted based on the purpose of monitoring For full list of ACES syndromes see table APPENDIX C ACES Syndromes Appendices APPENDIX C ACES SYNDROMES ACES Code Syndrome Description ALLERG AST BITE BRONCH BURN CAD CARD CDIFF CELL CHF CO CONC COPD CROUP CV DEHY DENT DERM DM ELECT ENT ENVIRO EOH FALL FBI FEB GASTRO GB GI GMED GNSURG GYN HEAD HEM HEP IL
51. cription Advantages Strengths Weaknesses Family Name no days Period x daily count chance x gt UCL2 for 2 of 3 Any 2 of the last 3 days are On Edge 14 t to t14 0 1696 consecutive counts greater than 20 above the u x gt UCL for 4 of 5 Any 4 of the last 5 days are Tendency 14 ti tO t14 0 32 consecutive counts greater than o AE 14 consecutive 14 counts in a row that Used as measure of data Not epidemiologically Oscillation 14t NA iud NA ae ete m oscillations alternate direction quality significant 15 count of 15 that are all less Used as measure of data Not epidemiologically Blissful Ignorance 14 t1 to t14 UCL lt X X 14 gt LCL 0 33 than the UCL and greater than Mad quality significant the LCL 15 count of 15 that are all Used as measure of data Not epidemiologically No Middle Ground 14 ti to tig X4 X 44 gt UCL or lt LCL 0 01 greater than the UCL or less pais quality significant than the LCL RecentMax Current count is greater than RecentMax 10T titotio x 1 5 x max x 4 x 40 1 5 times the maximum value observed in previous 10 days see Figure 33 for visual description of baseline tu and o are not calculated number indicates the number of days or the time period for the specific metrics of analysis 67 ACES Manual Quick Reference Guide for ACES Alerts APPENDIX A ABBREVIATIONS ACES BW C4 5 CDC CoCo CuSum DA EARS ED FRI FSALDU HIRA HPPA ILI KFL amp A LH
52. dropdown options or fill in manually Syndrome Algorithm All ME the default is All Classifications see details below and CTAS 1 2 3 4 5 All CTAS score the default is All NOTE When all display options have been chosen click the large miss Submit green Submit button Reset will return all parameters to default Figure 8 Epicurve display options tools settings 21 ACES Manual User Interface Guide 2 4 1 1 Classifications This category allows you to choose specific options regarding the syndromic classification you want displayed Under Classifier the dropdown menu gives the option to display either syndrome data that was classified according to the older EDSS RODS classifier 2006 EDSS Standard Syndromes v2006 or by the new ACES classifier S2014 ACES Standard Syndromes v2014 the default is S2014 If S2006 is selected only the eight with an additional category of All syndromes that were associated with the EDSS system will be displayed For a discussion of how hospital visits are classified into the various syndromes see Section 3 3 Syndrome Classification using Natural Language Processing The following syndromes are available as display options under the Syndrome dropdown menu 1 All 2 Asthma 3 Dermatological 4 Fever ILI 5 Gastrointestinal 6 Neurological 7 Other 8 Respiratory and 9 Severe Infectious Conversely the default new classifier v2014 you will notice that t
53. e Syndromic surveillance has its origins in epidemiologic surveillance the traditional systematic collection evaluation and dissemination of health information Traditional surveillance methods can be passive e g regular disease reporting to public health authorities or other administrative means or active e g gathering disease information from a population through surveys Syndromic surveillance is a distinct form of passive surveillance in that information regarding a predefined syndrome i e a particular disease condition or injury is monitored for a specific population This form of surveillance is of particular value for assessing or predicting disease outbreak such as influenza Examples include the monitoring of absenteeism logs from school or work settings over the counter drug sales internet search data such as Google Flu Trends or emergency department triage records such as ACES Syndromic surveillance drew attention and consequently increased research interest during the perceived threat of terrorism particularly bioterrorism following 9 11 in 2001 The occurrence of pandemic threats e g SARS from November 2002 to July 2003 secured the utility and value of real time syndromic surveillance beyond bioterrorism as public health agencies recognized inadequacies in emergency preparedness and the need for enhanced surveillance techniques e g the Walker Report the Naylor Report the Campbell Commission Met
54. e data and the process may out of control For ACES a Blissful Ignorance alert may not require any specific investigation by epidemiologists but Un t mon INSERT GRAPH FROM ACES ALERTS BLISSFUL IGNORANCE data quality Figure 42 shows an example of Blissful Ignorance where Figure 42 SPC Rule 7 Blissful Ignorance the highlighted data point is the fifteenth count in a row within one standard deviation or UCL lt count gt LCL 63 ACES Manual The Science of ACES 3 4 2 8 SPC Rule 8 No Middle Ground In ACES a No Middle Ground alert is triggered if the current visit count is the eighth in a row where all eight counts are greater than one standard deviation from INSERT GRAPH FROM ACES ALERTS NO MIDDLE GROUND the mean Like Oscillation and Blissful Ignorance this alert may not be of priority to epidemiologists but rather identify data with Figure 43 SPC Rule 8 No Middle Ground more variation than should occur by chance alone The probability of the occurrence of this alert assuming a normal distribution is 0 0196 Therefore this indicates that something out of the ordinary is occurring and the process may be out of control Emergency room visit counts fluctuating beyond one standard deviation is likely of no epidemiological concern No Middle Ground therefore is used to monitor data quality Figure 43 shows an example of a No Middle Ground alert the eighth point in a row that is great
55. e due to Description Advantages Strengths Disadvantages Weaknesses Family Name mo gays __Period x daily count chance Cumulative Sum CuSum Daily counts are standardized C1 gt 3 when e CuSum 1 Mild relative to mean values as Z EE daulaceline ta Second day of high counts 7 t tot C1 max 0 C1 Zy scores C is sum of z for i is obscured x included in calculate C2 angu xe uto current and previous day s calculation of u and o d counts Daily counts are standardized relative to mean values as z C2 gt 3 when ET scores C2 is sum of z for uSum 2 Medium i a 7 tztot CA max 0 C2 zj current and previous day s EE ie W oca Gu C2 COURTS X triggers alert and z bee Keno e NOTE C has different baseline than C4 C3 3 when Sum of C2 for current and 2 CuSum 3 Ultra previous days C2 is 0 if it 7 t_3 to tg C3 C2 C2 C24 C3 triggered an alert i e greater and C2 20 if gt 3 than 3 Statistical Process Control SPC Daily count is greater than 30 Extreme 14 ti to t14 x gt UCL3 0 13 above the u Xr X g gt OR 9 day of either all above the u Bias 14 t to t14 0 13 KERSE OR all below the p X 5 lt X4 lt X3 X2 6 day of six days of increasing Identify trend before x Trend NA NA 0 1496 Xx SR counts exceeds UCL3 66 ACES Manual Quick Reference Guide for ACES Alerts Probabilit Alert Baseline Rule y AM Disadvantages due to Des
56. e Local PHU Hospitals from the Hospital dropdown menu to display patients using the hospitals within the health units jurisdiction choose Province Wide and Outside of PHU Hospitals for all Ontario hospitals and just those outside of the local area respectively Keep in mind that only data from participating hospitals are included and the extent of specific patient and hospital data to which you have access depends on your account credentials Hospital users will only have the option to view their own hospital data and health unit staff will be able to view all participating hospital s within their health unit When all display options have been chosen click the large green Submit button Reset will return all parameters to default settings 2 4 3 Creating an Epicurve An Example The following is an example of the steps needed to create an epicurve in ACES After logging in select the Epicurves tab at the top of the page Figure 5 At the top of the graph choose a moving average and standard deviation optional These options should now appear in the legend depending on your choices and corresponding lines on the graph Now turn your attention to the Tools menu and select the following V Health Unit your health unit Hospitals All Data Range from today s date to three weeks previous to today Gender All Age All Classifications Classifier S2014 Bucket All Syndrome All and Algorithm ME CTAS All Remem
57. e asthma and it is the leading cause of hospitalization for children and a significant cause of absenteeism from school and the workplace Asthma also presents a significant financial burden on the health care system see the MOHLTC s Taking Action on Asthma Each year shortly after children return to school in September there is a recurring increase in ED visits hospital admissions and unscheduled physician consultations for childhood asthma across North America Rhinovirus infections allergens and decreased use of asthma medications during the summer are all thought to be contributors to this problem ACES is used to monitor asthma related emergency department visits by children in the KFL amp A region as they return to school each fall and as expected a significant rise in emergency department visits is observed each year Analysis of this information enables health professionals to assess the annual epidemic and understand the extent of the issue as well as assist the preparation of prevention efforts to reduce the impact of the problem by determining the local regions with the highest rates of asthma and therefore in greatest need of intervention strategies 1 3 Future Directions The expansion and upgrades made to ACES have gone a long way to improve public health awareness across the province and its wide acceptance inspires our team to maximize the potential of ACES In addition to the applications mentioned in the sections above we
58. e first option allows you to choose the details of the health unit and or hospitals to be displayed The availability of these options depends on your data sharing agreement Hospital users will only have the option to view their own hospitals data Users from a regional health authority however may be able to data from individual hospitals in their region in addition to the aggregated hospital data 27 ACES Manual User Interface Guide Date Range Enter the dates of interest as numbers mm dd yyyy format or choose Date From and Date To from the dropdown calendars Gender Choose Male Female or All Age Clicking on the dropdown menu in this section gives six age group options All Child ages 0 to 17 School Child ages 5 to 17 Adult ages 18 to 64 Senior ages 64 to 130 and Adult Senior ages 18 to 130 Alternatively you can manually input the desired age range v Classifications Selecting a specific Syndrome from the dropdown menu will list the visit details for patients whose chief complaint was categorized into that syndrome see more details in Sections 2 4 1 1 Classifications and 3 3 Syndrome Classification using Natural Language Processing You can also choose to display all syndromes which will list visit information for all emergency department visits Tools Advanced Health Unit Hospitals KFLA All Date Range m 04 21 2015 05 12 2015 Gender All Male Female Age All 0 130 Classifications
59. e of alert is dependent on the algorithm that generates it The Canadian Triage Acuity Score is a tool that provides a measure of the relative acuity or severity and enables EDs to prioritize their patient care requirements and assess ED workload and resource allocation relative to case mix Scores are defined as Level 1 most acute Resuscitation Level 2 Emergent Level 3 Urgent Level 4 Semi Urgent and Level 5 Non Urgent Chief complaint refers to a short phrase entered by a triage nurse or admissions clerk describing the reason for a patient s visit to an emergency department CCs are meant to be as short as possible and therefore will include abbreviations and often idiosyncratic p Emergency Department Syndromic Surveillance is the original name for the ACES system A graphical display of the number of cases of a particular syndrome over a period of time ACES uses the distribution of cases to compare normative and aberrant values in order to generate alerts The FRI screening tool was introduced during the SARS outbreak in 2003 Patients presenting to participating hospitals were screened for respiratory illness using a standard series of questions to identify new or worsening cough shortness of breath fever shakes or chills in the last 24 hours The first three characters of a Canadian postal code are the forward sortation area FSA and the last three characters are the local delivery unit LDU The first
60. ease make note of any discrepancies and contact ACES at kflaphi kflapublichealth ca 2 Plot the cases using the Epicurve function to examine the trend causing the alert Use the normalize function and the standard deviation and moving average options in ACES to enhance your graph Note that graphs and cases specific to an alert can also be viewed when clicking on the alert under investigation 38 ACES Manual User Interface Guide 3 4 5 6 7 39 Map the syndrome under review using the Maps function to observe the geographic dispersion regarding the cases Be sure to use different geography levels eg FSA County etc to see if syndrome is specific to a small or large area If clustering is evident the investigator may examine schools daycares long term care facilities and other sources of data to help guide next steps If possible determine the population that is at risk e g children the elderly individuals residing in an area defined by a postal code or FSA etc using steps 2 and 3 If an outbreak is suspected notify the hospital s physicians PHU branch offices and potentially other PHUs depending on the scope and if further investigation is required PHUs can raise awareness amongst the relevant hospital s regarding the syndrome in question to enable more efficient diagnosis and to implement necessary and appropriate precautions such as infection control procedures The hospital s may also provide
61. een 180 and 240 kilometres an hour as the tornado cut a strip through the centre of the town causing severe damage and leading the town to declare a state of emergency In the minutes after the tornado struck ACES was employed to show in real time the ED registrations related to the tornado at the local hospital ACES was used to monitor visits for trauma in order to assess the extent of the injuries caused by the tornado as well as monitor the potential emergency department surge relative to local emergency department capacity and led to enhanced situational awareness during this emergency 1 2 5 Surveillance After Drug Policy Changes 1 2 5 1 Methadose Methadone is a synthetic opioid used in the maintenance treatment of patients who are addicted to opioids such as heroin and morphine On June 26 2014 the delivery of methadone maintenance therapy was changed in Ontario the program began a transition from prescribing a compounded methadone solution to a more concentrated formulation of an oral solution of Methadose leading to potential dosing errors and accidental overdose On July 17 2014 KFL amp A Public Health initiated a surveillance program using ACES to assess recent ED visits for methadone overdose or accidental opiate overdose An increase in emergency department admissions may indicate a negative effect of the new program The pre existing syndrome used to capture opiate related emergency department visits was modified to monitor meth
62. efined in the legend see example in Figure 23 Note the small arrow in the top right corner of the legend that can be used to reduce or expand the size of the legend For both mapping styles moving your cursor over a specific region will give more details regarding that region in a pop up information box The pop up box contains information specific to that region the name of the level of geography requested i e the FSA for the region the county name the PHU or the LHIN the number of visits for the syndrome requested the total number of hospital visits and the percentage of total visits for the selected syndrome of total visits Using Choropleth the box will appear in any section of the map with Proportional Symbols you need to place the cursor over the marker The Classifiers drop down menu offers five options for the algorithm used to sort the hospital visits into pre defined syndromes Theses classifiers are derived from machine learning applications more information regarding this process and the choices included can be found in Sections 2 4 1 1 Classifications and 3 3 Syndrome Classification using Natural Language Processing The default classifier is Maximum Entropy 2 7 1 3 Classification The Classification option allows you to choose to display the data sorted into syndromes using the most recent syndrome lists and algorithms S2014 or using the previous EDSS RODS system S2006 The older system has been
63. ements are amenable to graphical display as a function of time SPC alerts identify aberrations in measures of quality control or process stability that are unlikely to be caused by chance alone Measurements of the parameter are compared to their means and standard deviations over defined time periods and an alert is triggered according to differences in the current measurement from the mean and within defined levels of standard deviation depending on the type of SPC alert SPC 58 ACES Manual The Science of ACES alert methods have been used to monitor and improve hospital performance and are used in disease surveillance to detect large increases in disease reports for the National Notifiable Diseases Surveillance System of the Centres for Disease Control In ACES eight different SPC alert types are used Extreme Bias Trend On Edge Tendency Oscillation Blissful Ignorance and No Middle Ground Means and standard deviations are typically calculated from the preceding two weeks of ED visit counts for a particular syndrome It may be useful to change the sensitivity of an SPC alert by changing the time periods used for calculating means and standard deviations For every data point measured three upper control limits UCL UCL2 and UCL3 and three lower control limits LCL LCL2 and LCL3 are calculated as follows UCL u 0 UCL2 u 20 UCL3 u 30 LCL u o LCL2 u 2o LCL3 u 30 Equation 4 Equation 5
64. ency Department Visits normalized values or percentages for the selected syndrome and geography Fr The number of gradients is determined by the Number of Classifications i 8 12 96 chosen For example if you choose Data Classifications Equal Interval Number of Classifications 6 and Percentage Range between 0 and 24 the 12 16 legend will have 6 equal gradients of 4 each When using percentage range 16 20 be aware that for each syndrome percentages can vary a good knowledge of L REN Figure 24 Equal Interval Classifications 6 Percent syndrome RESP in the summer months numbers are generally low and Percent Range 0 to 24 the syndrome s behaviour is necessary For example when mapping the Range between 0 and 20 should be sufficient to display all the data In the winter months the upper extreme needs to be increased as RESP visits generally increase proportionally versus total hospital visits at this time and may be greater than 20 of all visits and would therefore not be represented on the map 34 ACES Manual User Interface Guide F Quantile is a more simplistic approach Choose the Number of Emergency Department Visits Ouantiles z Classifications 4 5 6 or 10 and the same number of equally sized data Sample Data Q1 Least amount of eek subsets will be displayed quartiles quintiles sextiles and deciles s respectively For example if 4 is chosen
65. end alert is a unigue SPC alert in that it is not dependent on means or standard deviation calculations It is triggered in ACES if the current visit count is the sixth in a row where INSERT TREND ALERT FROM ACES HERE all six counts are increasing Six counts in a row of decreasing values should also technically Figure 37 SPC Rule 3 Trend trigger a Trend alert but ACES is primarily concerned with monitoring for increasing trends or counts that are anomalously high The Trend algorithm therefore follows these conditions count tj lt count t lt count ts lt count t count t lt count tg An example of this alert is shown in Figure 37 the highlighted data point is the sixth in a row where all six points are increasing A Trend alert is epidemiologically relevant as it may reveal an increasing trend in emergency department visit counts before those counts exceed the UCL such as is described in an Extreme alert Six points in a row that are steadily increasing is statistically very unlikely and would represent a rise that is not due to chance alone This alerts is of epidemiologically relevance therefore as it may determine subtle changes in emergency department trends that would not be picked up by other surveillance methods that rely on the current day s values in relation to means and standard deviation 3 4 2 4 SPC Rule 4 On Edge In ACES an On Edge alert is generated if the current visit count is the second
66. eospatial analysis changes in patient demographics and disease severity CTAS levels Contact hospital s as required ACES Manual User Interface Guide 2 8 2 Suggested Procedure for Respiratory or Gastrointestinal Outbreaks in the Community 1 Asa first step Public Health staff will be notified via telephone call fax and or mail from a laboratory an institution such as long term care of daycare physician etc of an infectious disease event reduiring further investigation In most cases this notification is concerning a reportable disease which must be reported by law to Public Health Public Health investigation is warranted for most reportable diseases unusually high incidence of disease particularly in institutions new or emerging diseases or any other infectious disease event deemed to be a threat to the health of the public Notification may come to any of the appropriate CD or EH team members during regular business hours Monday to Friday 8 30 am to 4 30 pm or to the Medical Officer of Health or assigned back up during evenings and weekends 2 ACES may be utilized in addition to standard investigative processes as follows Following notification to Public Health access ACES to examine the syndrome of interest time period demographics and location of interest as described above One of more of these features may be explored individually to characterize the potential outbreak El Use Epicurve and or Map to display the
67. er than UCL is highlighted As this alert is used for monitoring data quality an alert is also triggered when the counts are less than LCL 3 4 3 RecentMax NOTE THIS PREVIOUSLY CALLED NRC A final alert used by ACES to detect aberrations in daily emergency department visit counts is an experimental method designed by the National Research Council NRC RecentMax uses a ten day baseline period to determine a deviation from expected counts Simply defined an alert is generated is the current day s count exceeds twice the largest count for the previous ten days as follows X gt 2 x max xy 4 X1 49 Equation 10 where x are visit counts for time t This algorithm although parsimonious has been shown to perform well at detecting public health threats while simultaneously minimizing the rates of false positives according to initial research conducted by the NRC For use in ACES testing of this algorithm with the multiplicative factor of two resulted in the least sensitivity of all the alerts Sensitivity has been increased by using the multiplicative factor of 1 5 ADD EXAMPLE FROM ACES AND DISCUSS 64 ACES Manual The Science of ACES INSERT GRAPH FROM ACES ALERTS RecentMax Figure 44 NRC Algorithm Figure 44 3 4 4 Quick Reference Guide for ACES Alerts The following chart summarizes the alerts and uses Please print this out for quick reference 65 ACES Manual The Science of ACES Probability Alert Baseline Rul
68. erestingly this study spanned the increased ED volumes observed during the H1N1 pandemic in the fall of 2009 The specific syndromes were defined from key words found in recorded CCs for respiratory illness cough sore throat upper respiratory tract infection and sinus infection were used for influenza like illnesses fever and influenza related symptoms were used The results were aggregated weekly from the EDSS and compared to lab results Significant correlations were measured for selected respiratory viruses and emergency department visits pH1N1 was the most closely associated with hospital visits and a general lag time of about two weeks was found between the acute care data and test results In 2014 the EDSS system was overhauled to reflect both the expansion of the participating hospitals across Ontario and the development of additional and more specific syndrome classification At the same time numerous technological improvements were made To reflect these changes the system was renamed Acute Care Enhanced Surveillance ACES The number of hospitals participating in ACES has increased to more than 8196 of Ontario hospitals see Appendix EAPPENDIX E ACES Participating Hospitals and we anticipate that over 8596 of all Ontario EDs will be participating in ACES by the end of 2015 The list of syndromes has been expanded to a comprehensive list of over 80 health conditions APPENDIX C ACES Syndromes The technological cha
69. fecting otherwise healthy and young people at alarming rates The first cases of H1N1 in Ontario were confirmed in late April 2009 When a second wave of the H1N1 outbreak occurred in the fall of 2009 ACES then called EDSS was used to monitor the outbreak in real time within the KFL amp A region Analyses of ACES respiratory and fever ILI syndromes were able to detect and describe the increase in emergency department visits for respiratory complaints leading up to and during the H1N1 outbreak Surveillance maps generated using ACES were able to visually identify regions with high numbers of ILI related emergency department visits Additionally the momentum demonstrated in the ACES generated epicurves indicated that there would likely be continuously increasing numbers of patients visiting KFL amp A s emergency departments threatening an overload of each emergency department s capacity Based on this information assessment centres were established to relieve the pressure on the emergency departments resulting in a decline in emergency department visits over the next two weeks The use of ACES throughout the H1N1 pandemic demonstrated its multiple capabilities in assisting with a disease outbreak not only was the system able to help detect describe and track the outbreak earlier but it also helped to warn health professionals of increased patient number to enable a concerted health system response 1 2 2 Reportable disease detection Through the
70. fluenza outbreaks the effects of climate change and extreme weather events specific causes for injury and asthma and clusters of health effects related to illicit drug use ACES was started as a two year pilot project called the Emergency Department Syndromic Surveillance EDSS system with just two Kingston emergency departments in 2004 The Emergency Department Syndromic Surveillance system was both developed and funded by a collaboration between Kingston Frontenac and Lennox amp Addington KFL amp A Public Health the Public Health Division of the Ministry of Health and Long Term Care MOHLTC Queen s University the Public Health Agency of Canada PHAC Kingston General Hospital KGH and Hotel Dieu Hospital HDH EDSS was based on an open source software package from the University of Pittsburgh s Real time Outbreak and Disease Surveillance RODS system The RODS system was modified for the needs of an Ontario based population such as the customized geospatial mapping and the optimization of alerts and syndrome classification The EDSS system collated specific data elements from triage records such as chief complaint as free text date and time of visit hospital name patient age patient sex and patient postal code to 5 characters Classifying algorithms derived from machine learning applications were used to categorize each emergency department visit into pre defined and medically significant syndromes according to the words part
71. hagic fever including 1 Ebola virus disease 2 Marburg virus disease 3 Other viral causes Hepatitis viral 1 hepatitis A 2 hepatitis B 3 hepatitis C Influenza Lassa Fever Legionellosis Leprosy Listeriosis Lyme Disease Malaria Measles Meningitis acute 1 bacterial 2 viral 3 other Meningococcal disease invasive Mumps Opthalmia neonatorum 72 ACES Manual Paralytic Shellfish Poisoning Paratyphoid Fever Pertussis Plague Pneumococcal disease invasive Poliomyelitis acute Psittacosis Ornithosis Q Fever Rabies Respiratory infection outbreaks in institutions Rubella Rubella congenital syndrome Salmonellosis Severe Acute Respiratory Syndrome SARS Shigellosis Smallpox Syphilis Tetanus Trichinosis Tuberculosis Tularaemia Typhoid Fever Vertoxin producing E Coli infection including Haemolytic Uraemic Syndrome HUS West Nile Virus including 1 West Nile Fever 2 West Nile neurological manifestations Yellow Fever Yersiniosis Bolded items and Influenza in institutions should be reported immediately to the Medical Officer of Health by telephone Other diseases can be reported by the next working day by fax phone or mail Appendices APPENDIX E ACES PARTICIPATING HOSPITALS Health Unit Key Hospitals within Health Unit ACES start Below APH CKPHU DRHD EOHU GBHU HKPRDHU HRHD HPH HPECHU KFLA LAMBTON CHSD LG
72. he non traumatic hematological condition anemia thrombocytopenia not oncological hepatitis undifferentiated and A B C fever myalgia undifferentiated flu non specific infections potential interest to public health epiglotitis tonsil abscess sprain strain laceration dislocation bruise swelling insomnia sleep disorder intussecption Appendices ACES Code Syndrome Description LAC MEDREN MEDSE MEN MH MHS MIGR NEC NEURO NEUS NEWB OBS ONC OPI OPTH ORTHF ORTHH ORTHO OTHER PAIN PE PHYSC PN PO REN REPORT RESP SEP SI SOC TEST THOR TICKS TOX TRMVC TRO TRW TRS URO VOM 71 lacerations medication renewal request medication side effect not OD meningitis and encephalitis mental health suicidal ideation attempt or overdose migraine necrotizing fasciitis severe cellulitis gangrene dementia Alzheimer s stroke seizure vertigo syncope fainting neurosurgery aneurysm bleed etc subdural SAH newborn related to obstetrics oncology opioid intoxication addiction overdose withdrawal general ophthalmological condition fracture non hip fracture of the femur or hip orthopedic elective surgery cast change or assessment null missing other pain undifferentiated non cancer radiculopathy back pain sciatica pulmonary embolism DHT VTE physician consultation pneumonia post op infection or complication renal failure dialysis renal disease and its complic
73. he most sensitive CuSum alert is CuSum3 The calculation of this statistic depends on CuSum2 values it is the sum of CuSum2 from the current day and the previous two days but only if the previous two days did not generate alerts If the previous days CuSum2 did generate alerts their values are set to zero as in the following formula CuSum3 St Set St Equation 3 where 5 is the CuSum2 value for the current day and Sr and Sr are the CuSum2 values for the previous two days If St gt 3 then 0 and if St 3 then 7 O ACES alert settings will trigger a CuSum 2 alert if CuSum2 gt 3 Again referring to the data shown in Figure 34 and Table 4 the relationship between CuSum alerts can be observed for this data set CuSum1 gives one alert on 11 Feb CuSum2 triggers three alerts on consecutive days 11 Feb to 13 Feb and CuSum3 has the same results as CuSum2 except that an additional alert is calculated for 10 Feb The graph Figure 34 reveals that there appears to be the start of an elevation in ED counts on 10 Feb and CuSum3 gives a statistical foundation for concern 3 4 2 Statistical Process Control SPC The Statistical Process Control SPC family of alerts were developed and used by the manufacturing industry to improve product quality by reducing product variability During a manufacturing process quality control and or the stability of the process are monitored by measuring a defined parameter generally these measur
74. here are many more syndromes available for display reflecting the number of syndromes currently being monitored The syndromes are shown in the dropdown menu in alphabetical order with All as the first option The abbreviations and descriptions of v2014 syndromes are given in APPENDIX C ACES Syndromes A discussion of their meaning and validation against diagnostic codes is found in Section 3 3 5 Syndrome Validation In addition to the syndromes available using the v2014 the Bucket dropdown menu allows you to customize your epicurve using pre defined groupings of certain syndromes of interest For example ENVHE Environmental Health Exposures displays the results from Asthma CHF congestive heart failure COPD chronic obstructive pulmonary disease DEHY dehydration ELECT electrolyte imbalance hyperkalemia hypomagnesium hyponatremia ENVIRO heat stroke heat syncope heat exhaustion cold frost bite hypothermia and MIGR migraine syndromes Bucket categories and their definitions are an on going and dynamic project and may potentially be customizable in future versions of ACES Finally the Algorithm dropdown menu provides five options for choosing the classifying algorithm used to categorize the emergency department visits into the selected syndrome The classifying algorithms used in ACES are mathematical models used to sort ED visits into syndromes by statistically categorizing the phrases words and parts of words found in the c
75. hief complaints of the triage records For example if the word fever was the only word used in a chief complaint the hospital visit would very likely be sorted into a syndrome for which fever is a common symptom such as ILI fever myalgia 22 ACES Manual User Interface Guide undifferentiated flu and not very likely to be sorted into DERM rash undifferentiated lesion wart The algorithm options are BW Balanced Winnow C4 5 C4 5 Decision Tree ME Maximum Entropy NB Naive Bayes and W Winnow2 The default setting is ME The option for changing the sorting algorithm is included in ACES to provide corroborative evidence for the sorting processes for example if you think an emergency department visit has been misclassified it may be helpful to observe the results for different algorithms before seeking further technical advice For further descriptions of the classifying algorithms used in ACES see Section 3 3 Syndrome Classification using Natural Language Processing NOTE when all display options have been chosen click the large green Submit button Reset will return all parameters to default settings When a specific syndrome is displayed you can choose from several different statistical display options As described earlier for A visits 7 or 14 day moving averages can be displayed or Max to display the average for the displayed data range as well as the corresponding standard deviation Std 1 or Std 2 see Section 2
76. hods for the early detection of intentional poisoning of water or food for example were needed in real time to ensure rapid treatment and allocation of resources These same methods would be useful for monitoring a growing list of possible threats to both individual and public health such as influenza outbreaks the health effects of climate change e g increased extreme weather specific causes for injury and asthma and clusters of health effects related to illicit drug use Electronic data collection has led the way for modern syndromic surveillance systems to provide efficient sensitive and near real time capabilities to detect and statistically analyze aberrations from historical trends Geospatial analytics allow for visualization of these trends at regional national and global levels Assuming that aberrations are detected in a timely manner reasonable public health measures can potentially minimize adverse health outcomes the addition of geographical information allows for population specific interventions Interventions might include but are not limited to the reallocation of health care resources to high risk populations public health reassurances and health care 42 ACES Manual The Science of ACES recommendations For example if a syndromic surveillance system indicates that there is heightened influenza risk within a specified region dedicated clinics can be opened to treat patients possibly reducing patient influx into lo
77. ied the chief complaint correctly into OPTH The inclusion of the results for the alternate classifiers adds power to this classification the highest ranking syndrome using ME is OPTH at 4296 and although the other rankings have much lower statistics of likelihood the choice of OPTH is supported by two other algorithms The inclusion of the results of all algorithms also allows for the examination of classification results on an individual hospital visit basis if necessary 3 3 5 Syndrome Validation The ACES system uses NLP to sort each hospital record into a specific and medically relevant syndrome some syndromes are used to monitor communicable diseases and other syndromes reflect hospital visits for medical conditions reflecting social or environment factors that may affect the health of the community The default algorithm for syndrome categorization in ACES is ME but you can choose to sort your data by several other algorithms i e BW C4 5 MCME and Winnow 2 Classification proceeds using the free text of the CC registered by triage nurses when the patient first arrives in an emergency department Using CCs to make real time assessments of population health leads to numerous complications chief complaints tend to be very brief and include jargon idiosyncratic short forms and spelling mistakes Furthermore CCs are the triage nurse s interpretation of a patient s reason for visiting the emergency department and do not necessarily reflect
78. in a CuSum1 alert which by definition is considered an alert of mild sensitivity Cusum1 Current The means p and standard deviations CuSum2 and CuSum3 Event c used to calculate CuSum1 are based on daily ED visit counts for the Figure 33 TImeline for CuSum alerts seven days previous to the current day that is from t to t Figure 33 The threshold limit is defined as three in ACES and therefore an alert is flagged when CuSum1 gt 3 The sensitivity of CuSum1 can be increased or decreased by changing this threshold It is important to note that because the current daily count is compared mean and standard deviation of the previous seven days if an aberrant value is detected on the current day t the CuSum1 calculated for the next day t will be less likely to surpass the threshold as the previous day s higher than average value will increase the values of both the mean and standard deviation 56 ACES Manual The Science of ACES 3 4 1 2 CuSum2 Moderate Sensitivity CuSum2 is calculated using the same N UT methods as CuSum1 but with a shifted baseline for calculating the mean u and N O O the standard deviations o see Figure 33 N ES N gt 150 O LLJ Changing the timeline for these calculations may increase the sensitivity of this method the additional two day lag period between the current count and the values used to calculate the mean may increase sensitivity on subsequent
79. itiated in September 2004 as a two year collaborative pilot project between the Public Health Division of the Ministry of Health and Long Term Care KFL amp A Public Health Queen s University the Public Health Agency of Canada PHAC Kingston General Hospital and Hotel Dieu Hospital The EDSS system used open sourced RODS v3 0 software to categorize ED triage CCs from the participating hospitals collected in near real time Major modifications to the RODS software included the addition of geospatial mapping alert optimization syndrome classification and other changes to the user interface such as CTAS and FRI screening results As in the RODS system key words phrases and parts of words phrases found in CCs are used to classify each ED visit into pre defined syndromes The initial syndromes were 1 gastroenteritis 2 fever influenza like illness ILI 3 respiratory 4 dermatologic infectious 5 severe infectious 6 asthma 7 neurological infectious and 0 other included all CCs not otherwise classified and accounted for the majority of ED visits Results were aggregated and displayed using tables graphs and maps and statistical aberrations alerts were detected as results above expected historical thresholds During the initial two year pilot project an outbreak of Salmonella enteritidis in November and December of 2005 provided a pragmatic illustration of the effectiveness and value of the EDSS system At the time the EDSS
80. ively The last six data points that caused the alert are shown in blue and red on the day the alert is triggered Although there are no points greater than the UCL3 an alert is triggered because the rule has been satisfied for a Trend alert It is important to note however that although this statistical information is included in the graph a Trend alert is based solely on the last six data points showing an increasing trend and therefore each data point s relation to statistical parameters is unimportant Choosing the Data tab at the top right of the graph will show the same data in tabular format 2 8 1 Alerting Protocols For a detailed description of all the alerts used with ACES see Section 3 4Error Reference source not found As a first step ACES users will investigate the alert and when warranted pass on the alert via email to the Communicable Diseases CD and or the Environmental Health EH teams of the associated PHU or similar line of notification as per each HU s policies The specific steps in this investigative process can include users can also use the Risk Assessment of ACES Alarms tool Appendix X need to update 1 Confirm the syndrome classification The individual ED visits that are causing the alert are viewed individually to ensure that the classification corresponds with the chief complaint and or diagnosis See Section 2 4 Epicurves Note If the syndrome classification is incorrect of inconsistent pl
81. lculation methods used to determine probability Winnow 2 is one of many variations of BW in which slightly different methods are used to determine defined iterative values 3 3 4 Algorithm use in ACES Several NLP algorithms are used to classify the ED visits into syndrome in ACES ME is used as the default algorithm due to its robustness with similarly complicated applications As previously stated RODS used NB and the EDSS system used ME to provide continuity and comparability between the various iterations of ACES the technical support necessary to use the different algorithms is maintained in ACES To that end the results of the additional algorithms can be E T ief Complaint displayed as well i e BW C4 5 MCME and Winnow2 WATER INFRONT OF MY EYES Provision of results by several different NLP algorithms Most Likely Syndromes enables an assessment of the strength of the classification i 2006 Classifier Syndrome we display the most likely syndrome using multiple MAXIMUM ENTROPY Other 100 classifiers so that end users may better understand chief 2014 Classifier Syndrome complaints that classify easily and for which we have high MAXIMUM ENTROPY OPTH 42 CV 1096 confidence The user can consider the difference between ss ee Other 4 all six classifiers unanimously predicting the same most OBS 3 BALANCED WINNOW CV likely syndrome with high probabilities versus all six C4 5 DECISION
82. ll Syndrome Algorithm 0 Apr 21 2015 Apr 24 2015 Apr 26 2015 Apr 29 2015 May 01 2015 May 04 2015 May 06 2015 May 09 2015 May 12 2015 All Focus Chart CTAS 11213 als an Apr21 2015 Apr 24 2015 Apr 26 2015 Apr 29 2015 May 01 2015 May 04 2015 Mav 06 2015 May 09 2015 Mav 12 2015 Em uomi xk Download Chart KFLA 2015 Figure 11 An epicurve example HDH KGH LACGH e HOHSTDDEV1 HDHSTDDEV2 Qe KGH STD DEV1 KGHSTDDEV2 LACGHSTDDEV1 LACGHSTDDEV2 AI STD DEV 1 e l STD DEV 2 HOH 7 day AVG KGH7 day AVG LACGH 7 day AVG QAI 7 day AVG 0 Apr 21 2015 Apr 24 2015 Apr 26 2015 Apr 29 2015 May 01 2015 May 04 2015 May 06 2015 May 09 2015 May 12 2015 Figure 12 An epicurve example Local PHU Patients and Hospitals 25 ACES Manual User Interface Guide In the Advanced options tab the data can be further constrained For this example the data is displayed by hospital by choosing Local PHU Patients and Local PHU Hospitals this sorts the data to display to just the patients that reside within the PHU region that visited each hospital within the PHU region In Figure 12 the epicurve is now broken down by hospital in the KFL amp A Public Health region The black line represents data from all hospitals combined The red line represents data from Hotel Dieu Hospital HDH the sreen lines represent data from Kingston General Hospital KGH and the blue lines represent data from Lennox and Addington County Gener
83. lthcare Huntsville amp Bracebridge MAH North Simcoe Hospital Alliance Huronia District Hospital Orillia Soldiers Memorial Hospital OSMH Royal Victoria Hospital RVH Stevenson Memorial Hospital Chapleau General Hospital CHS Espanola Regional Hospital and Health Centre EGH Manitoulin Health Centre MHC Sudbury Regional Hospital SRH Dryden Regional Health Centre DRH Geraldton District Hospital GDHO Manitouwadge General Hospital MGHO McCausland Hospital MCCA Nipigon District Memorial Hospital NDMH Thunder Bay Regional Hlth Sciences Centre TBRH Wilson Memorial General Hospital WMGH Englehart And District Hospital EDH Kirkland And District Hospital KDH Temiskaming Hospital TEM The Hospital for Sick Children Mount Sinai Hospital MSH North York General Hospital Rouge Valley Health System Centenary RVC The Scarborough Hospital General Site TSH The Scarborough Hospital Birchmount Site TSHB Nov 2008 Jul 2008 Feb 2010 Feb 2010 Feb 2010 Feb 2010 Feb 2010 Apr 2010 Apr 2010 Apr 2010 Sep 2010 Aug 2010 Sep 2010 Sep 2010 Oct 2014 Nov 2011 Feb 2013 Jan 2011 Apr 2009 Jul 2011 Jul 2011 Jul 2011 Jul 2011 Feb 2013 Feb 2013 Feb 2013 Jul 2011 Feb 2010 Feb 2010 TBD Feb 2010 May 2010 TBD Jul 2011 Feb 2013 Feb 2013 Feb 2011 Sep 2010 Sep 2010 Sep 2010 Sep 2010 Sep 2010 May 2010 Sep 2010 Jul 2011 Jul 2011 Jul 2011 TBD Oct 2013 TBD Sept 2013 Jan 2014
84. nced Winnow C4 5 Decision Tree Monte Carlo Maximum Entropy and Winnow2 48 3 3 4 PEO ICRI USE N ACE es RENT 48 3 3 5 TE NNN 49 ACES Manual Table of Contents 3 3 5 1 Examples of Syndrome Validation ees see ee ee ee RR ee ee RE nennen 50 3 4 Alerts and Outbreaks rrrrnnnnnnnrrnnnrrnnnnnnnnsrnnnnrnnnnnnsssrnnnsnnnnnnnsssnnnnsnnnnnnnsssnnnnsnnnnnnnsssnnnnsnnnnnnnsssnneenn 54 3 4 1 The Cumulative Sum CuSum Family of Alert S esse sesse ss see ee ee ee ee ee ee ee RE ee 54 ALL NNN se sd GE ER oe ee Ge oe Ee Ee ve ded 55 3 4 1 2 CuSum 2 Moderate SensitiVitV sees see ee ee ee ER ER ee ee ee ee ee RE ee RE ee 57 3 4 1 3 CuSum3 Ultra Sensitivity ee ee RE ee RR ee RR ee RE ee RR ee ee ee RE ee 58 3 4 2 Statistical Process Control IS PC Lee 58 34424 SPC Rule BON ier cote ttes callo id Everett 59 24522 SPC RUC e Bi caccia er de M NN etr EG EIER rta ee 60 3 4 2 3 SPC Rule 3 Trend BE WE ooo seek sss 61 3 4 2 4 SPCRule4 On Edge 4 Buen 61 3 4 2 5 SPCRule5 Tendency 4BBB ee Eee 62 3 4 2 6 SPC Rule 6 Oscillation Wi DB 62 3 4 2 7 SPC Rule 7 Blissful Ignorance B 009 V 63 3 4 2 8 SPC Rule 8 No Middle Ground rrorrrrnnnnnnnvnnnnvnnnnnnnnsrrnnnvnnnnnnnsrsnnnnvnnnnnsssessnnnnrnnnnnnsssnn
85. ncing health system preparedness 11 ACES Manual ACES Backgrounder 1 2 3 2 2015 Pan Parapan American Games In the summer of 2015 Ontario s Greater Golden Horseshoe area will be hosting the Pan ParaPan American Games The event will include an estimated 10 000 athletes and 250 000 visitors In preparation for the Games a provincial Surveillance Work Group was created and several surveillance objectives were outlined including the recommendation that ACES be used throughout the event for public health surveillance With over 100 hospitals in twenty six health units participating in the system across Ontario ACES extends through all regions that may be affected by the Games and includes nearly all of the hospitals within close proximity to event locations Due to its flexibility adaptability and its strong monitoring and analysis capabilities ACES will be extensively used by stakeholders during this time 1 2 4 Emergencies and Extreme Weather 1 2 4 1 Kingston Fire On December 1772013 a large fire broke out in a construction site near downtown Kingston ON The size of the fire and the wind conditions at the time led to concerns regarding the spread of fire to nearby buildings as well as the potential production of a substantial amount of smoke The fire therefore posed a significant threat local air quality in addition to the health risks associated with exposure to fire smoke and carbon monoxide Upon being notified of the fire K
86. ne nnns 23 iv ACES Manual Table of Contents Figure 10 Figure 11 Figure 12 Figure 13 Figure 14 Figure 15 Figure 16 Figure 17 Figure 18 Figure 19 Figure 20 Figure 21 Figure 22 Figure 23 Figure 24 Figure 25 Figure 26 Figure 27 Figure 28 Figure 29 Figure 30 Figure 31 Figure 32 Figure 33 Figure 34 Figure 35 Figure 36 Figure 37 Figure 38 Figure 39 Figure 40 Figure 41 Epicurve display options advanced iese ee eek ee ee ee Re ee ee ee RR ee RR ee RE ee 23 Mt de SS 1001 9 AE OE LR OR N ER OE N 25 An epicurve example Local PHU Patients and Hospitals sesse sees ss see ee se ee ee ee ee ee ee ee 25 Tok 26 Line Listings ED Tools MENU esse ies Ee Dee Ge ER Ge See Ge Re EG ee ee ve ge ee ede Ge ed Ee 28 Line Listings ED Advanced menu esse sees see ee ee ee ee ee ee Re ee ee ee ee ee RE ee 28 Most Likely Syndromes from an individual line listing issie sees ee ee ee ee RR ee RE ee 29 Example of Line Listings AD ees sees ee ee ee RR Ee ee ee Re ee ee ee nnns nennen nnns 30 Line Listings AD Group By options ees ee RR ER ER EA EA EA ee ee EA ee RE ee EE ee RE ee rennes 31 Line Listings AD Tools menu esse dan nnne eaa oe ee ee GE eek dek GE erase eek GE ee 31 Map scale tool etre EAE eee eek 32 Map Features Data menu B ooo soes e soe e ees ese sesse see NE 32 Mapping Style Chloropleth B 289 Aes 33 Mapping Styles Proportional
87. nges to ACES include improved and more epidemiological assessment tools such as Y improvements in algorithms for more accurate syndrome classification Y more intuitive graphing capabilities Y built in calculations of standard deviations and moving averages Y monitoring capabilities have been updated to reflect real time values instead of arbitrarily fixed intervals and Y visit and or admission volumes based on either a hospital location or b patient address 3 3 Syndrome Classification using Natural Language Processing To understand the methods used to classify the chief complaint CC text into specific syndromes some concepts of machine learning need to be explored Machine learning deals specifically with the construction of classification algorithms that can learn from data ACES uses natural language processing NLP algorithms which are those algorithms designed to enable a computer system to understand 45 ACES Manual The Science of ACES human text NLP algorithms are used in daily life from the spam detector of your email program to the automated online assistant that pops up in your browser when you access certain web pages NLP began in the 1950s at the intersection of artificial intelligence and linguistics Initially it was considered a guest towards automatic translation with much emphasis on translating Russian into English reflecting its roots in the Cold War The task proved much more difficult than expected
88. nnnn 64 3 4 3 RecentMax M EIS EE EE EE EE EE EE N 64 3 4 4 Quick Reference Guide for ACES Alerts seeeeeeesseseeeeeee nnne nee 65 Appendix A Abbreviatiopeffie 9 2289 90993 68 Appendix B Glossary das AF 99er 69 APPENDIX C ACES SyndWEes 9898 MBs mmn rennen 70 APPENDIX D HPPA Reportable Diseases in OntariO sesse sesse se ee see ee ee ee ee ee ee ee nnnm 72 APPENDIXE ACES Participating Hospitals ararotlonererroranveranserennnvernnnnnsnnennnserevnnnenennnnsnnenunserennnvennennnsenene 73 Figures amp Tables Figure 1 ACES data C diB ction and Mv cycles rernerenorneronnanernanerenrenennerennunervanenenneneunerennunernansnennaneunerennune 17 Figure 2 Main landing page current syndrome counts for Ontario rrrurrrnnnnnrrnnnnnnrvnnnnnrrnnnnnernnnnnesennnneee 19 Figure 3 Main landing page Wrrent alerts ccce rete SE ee ee XY EG de SR Ee ee ed ee Ee un 19 Figure 4 Dropdown display for current alertS sees sees ee ee ee ee ee ee ee ee ee ee ee ee RR ee ee Ee ee 20 Figure 5 Main NAVIE ALON OP ONS ee A 20 Figure 6 Default Foldede 20 Figure 7s Data kes ee od oe ee EE Ee EE da eni tion bass Mte edu enis p Lem uM ELE ELEME 21 Figure 8 Epicurve display options tools ees see ee ee ee ee nennen nnne nnn nnns nnns 21 Figure 9 Normalize and other statistical OptiONS ees sees seek ee ee ee Re ee nnne nn
89. of ACES allows for unprecedented situational awareness both within specific public health jurisdictions and across Ontario The enhanced capabilities of ACES include improved and more interactive epidemiological assessment tools such as more intuitive graphing built in calculations of standard deviations and moving averages and monitoring capabilities that have been updated to reflect real time values instead of arbitrarily fixed intervals Furthermore ACES gives you access to report visit and or admission volumes based on either a hospital location or b patient address Like EDSS ACES continues to provide real time health surveillance but with a wider reach to more Ontarians and with a greater capacity for extracting meaningful information for public health 1 2 ACES Applications The flexibility adaptability monitoring and analysis capabilities of ACES enable situational awareness for a variety of common emerging or unexpected public health issues ACES syndromic surveillance capabilities are useful in a variety of situations including routine monitoring of seasonal influenza reportable disease detection public health emergencies surveillance of mass gatherings public health emergencies such as fires and extreme weather events monitoring of asthma surveillance after drug policy and NN NS WN NN S mental health surveillance 1 2 1 Influenza ACES provides invaluable information for health care services thr
90. of a value s proximity to the Figure 32 ED visits for Asthma 55 ACES Manual The Science of ACES mean Using this method the daily emergency Table 3 CuSum1 alert department count for Feb 11 would trigger an alert ACES would notify the administrator of an E 5 o YW ar Q x that hospital that the counts for asthma need 28 Jan 239 2323 to be examined in further detail 20 Jan 242 2333 30 Jan 229 229 6 One of the main benefits of Cusum methods for 31 Jan 196 227 6 aberration detection is the short period of O1 Feb 227 223 6 02 Feb 237 222 9 background data required to calculate alerts 03 Feb 240 225 4 10 Feb 273 241 6 13 9 1 27 1 27 11 Feb 301 246 3 182 2 01 3 28 seven days previous to 11 Feb and changes in Again sensitivity to changes in the daily ED i OA Feb 217 230 0 16 0 1 81 0 x counts can be varied according to need and the 05 Feb 247 2268 16 1 0 25 0 25 x x normal aberrations observed for that data In EN c wA 9 on O7 Feb 234 231 1 19 7 0 86 0 x the above CuSum CuSum1 calculations the 08 Feb 259 2366 123 08 0 82 x mean daily ED visits are determined from the 09 Feb 240 241 1 14 0 1 08 0 x x Y sensitivity can be achieved by changing the baseline of seven days from ts to ts see Figure 33 as will be discussed for CuSum1 and Cusum2 The benefits of each CuSum approach are discussed in the following sections The above example describes the calculations involved
91. of alerts comprise twelve alert types using diverse alerts encompasses as wide a range as possible of potential statistical aberrations and ensures that ACES maintains the highest possible sensitivity 3 4 1 The Cumulative Sum CuSum Family of Alerts The Cumulative Sums CuSum family of alerts is based on algorithms developed by the Center for Disease Control s Early Aberration Reporting System EARS to detect bioterrorism threats and subsequently implemented by a wide range of health departments across the United States and Canada EARS was designed to provide enhanced surveillance for a short duration around a discrete event e g Olympic Games political convention for which very little background data existed and therefore alert algorithms use just seven 24 hour periods of data The set of three EARS algorithms used by ACES are adapted to incorporate a cumulative approach to evaluating the data and are based on whether daily syndrome or total ED counts exceed the expected value by a certain threshold with varying sensitivity Aberrations in daily counts of ED visits that are above what is expected are assessed according to one 54 ACES Manual The Science of ACES sided CuSum calculations Daily ED counts are described as z value that standardize the daily count to the mean and standard deviation for the previous seven days as follows z ld Equation 1 Oo where x is the current daily ED count at time t and u is the mean
92. of any two of the last three counts greater than UCL2 The probability of this occurring by chance alone assuming a normal distribution is 0 1696 The highlighted data point in Figure 38 is the second point out of the previous three that is greater than two standard deviations from the mean Likewise Figure 39 is a second example of a circumstance where an On Edge alert would be triggered but the two points greater than UCL2 are not sequential DESCRIBE THE GRAPH SPECIFIC DATA POINTS POINT OUT THE RELATIVE SENSITIVITY TO OTHER SPC ALERTS 61 ACES Manual The Science of ACES Insert ACES screenshot of on edge alert 2 examples side by side Figure 38 SPC Rule 4 On Edge example 1 Figure 39 SPC Rule 4 On Edge example 2 3 4 2 5 SPC Rule 5 Tendency A Tendency alert is generated if the current visit count is the fourth count in the previous five counts that is greater than The probability of a Tendency alert being triggered assuming a normal distribution is 0 3296 In Figure 40 the highlighted data point would trigger an alert as it is the fourth point of five that falls beyond one standard deviation from the mean or is greater than the UCL DESCRIBE THE GRAPH AND DATA POINTS 3 4 2 6 SPC Rule 6 Oscillation An Oscillation alert is generated if the current visit count is the 14th in a row where all fourteen counts alternate in direction sequential increase then decrease Figure SENSE ORATIE TENDEN 41 describes a da
93. often than trave urti in our Respiratory syndrome classifications Figure 31 shows the correlation between the and emergency department visits that are 400 500 600 700 800 classified into the Respiratory syndrome and NACRS the retrospective NACRS data for the same Figure 31 Calculation of Pearson correlation coefficient for Respiratory time period Diagnostic codes used to extract data from NACRS are shown in Table 1 Excellent correlation statistics 0 952 are observed for all ages and all genders Thus we can say with certainty the trends observed in Respiratory reflect the extent of diagnoses for the medical conditions that Respiratory represents The validation statistics for the most commonly used syndromes are shown Table 2 Validation statistics will be re run regularly for the syndromes available in ACES Table 1 ICD 10CA codes used to extract NACRS data for validation of Respiratory Medical Diagnosis JOO JO1 J02 J03 J04 R05 51 ACES Manual Acute nasopharyngitis common cold Acute sinusitis Acute pharyngitis Acute tonsillitis Acute laryngitis and tracheitis Cough The Science of ACES Table 2 Word Cloud and validation statistics for common syndromes Visual Representation Syndrome Words Cloud from CC Correlation NACRS v ACES cough difficulty exacerbati AST 2013 Age 2 50 breathing asthma N 7 dav avg Corr coef 0 949 z d oi hon dyspnea
94. oughout Ontario for seasonal influenza Most notably the ILI Mapper was developed from ACES a feature that displays provincial respiratory and influenza activity in both map and graph forms to assist public health professionals to track influenza activity and various other respiratory illnesses throughout the annual influenza season The 9 ACES Manual ACES Backgrounder interactive and freely accessible on line ILI Mapper detects monitors and describes the annual influenza outbreak in your region Information about surge capacity virus progression and infection rate enables public health agencies and health care services to plan and implement alternative strategies for protecting the public during influenza season Furthermore the ILI Mapper is publically available so that all Ontarians can access real time information regarding the overall influenza status of the province as well as the trends observed at increasingly specific levels of geography This increases the awareness of influenza risk and enables individuals to be better informed to protect themselves from transmission of the virus The value of ACES for monitoring potential influenza pandemics was shown during the H1N1 epidemic in 2009 Although influenza viruses circulate each year the H1N1 pandemic was of particular concern as it had a higher than normal infection rate and it did not infect older adults at disproportionate rates as is normally expected In other words H1N1 was in
95. provide evidence for relevant policy changes The flexibility of the ACES system will enable efficient expansion for a range of mental health indicators allowing for surveillance and improved public health responses to these important health issues As mentioned previously ACES is currently being leveraged via the ILI Mapper to help local and provincial stakeholders along with the public be better informed on the progression of the annual influenza season There are currently plans to develop similar applications with other ACES syndromes including 1 an Asthma Mapper to detect both spatially and temporally the annual rise of asthma related emergency department visits for children as they return to school from the summer break 2 a GI Mapper to track gastrointestinal related visits to hospitals throughout the year for the early detection of potential food related outbreaks and 3 a Respiratory Syncytial Virus RSV Mapper to detect specific respiratory complaints such as bronchiolitis in children under the age of 5 to ensure they are differentiated from the overall respiratory data for focussed interventions Mappers for any disease or condition could potentially be created with appropriate statistical validation stakeholder interest and potential benefit 15 ACES Manual ACES Backgrounder 2 USER INTERFACE GUIDE 2 1 Data Collection and System Overview In Section 2 the normal usage of ACES is described providing guidance for the
96. ributions of categories and the terms cannot be assumed to be independent as is the case for the categorization of the words and partial words in CC The ME algorithm computes many different probabilities for the association of the terms and chooses the class selection that has the highest associated entropy or the largest probability distribution ME has a wide range of applications such as sentiment analysis language detection and topic classification Although ME takes more training time than NB it generally provides robust results and is competitive in terms of resource use to NB 47 ACES Manual The Science of ACES 3 3 3 Balanced Winnow C4 5 Decision Tree Monte Carlo Maximum Entropy and Winnow2 Although the default classifier employed by ACES is ME the data can be sorted using several additional NLP algorithms Balanced Winnow BW is a simple NLP algorithm that learns from training data i e labelled examples and uses a multiplicative scheme for probabilities that lends itself well to accounting for many dimensions of information that is irrelevant BW scales well to high dimensional data C4 5 Decision Tree C4 5 is a statistical classifier that is a sequence of branching comparisons at each intersection the probability of either branch is calculated and a decision is made towards classifying the ED visit into a single syndrome Monte Carlo Maximum Entropy MCME is a variation on ME so named for the particulars regarding the ca
97. risks Appropriate public health preparations are required before the event as well as sufficient surveillance during the event to attenuate potential health emergencies From June 25 to 30 2010 the G8 and G20 Summits were held in Huntsville ON and Toronto ON respectively A Hazard Identification Risk Assessment HIRA was completed in preparation for the event due to the large crowds expected to attend the events The HIRA identified several potential risks including infectious and contagious diseases food related hazards environmental severe weather emergencies and injury health and safety hazards The ACES system was used as a surveillance tool during the G8 and G20 Summits to monitor syndromes that coincided with the identified risks From June 17 to June 30 the system monitored specific health syndromes every hour looking for increases in hospital ED visits and admissions that may indicate an emerging public health threats related to the event Throughout the monitoring timeframe three indicators on the system were higher than expected values 1 total hospital admissions 2 fever ILI emergency department visits and 3 dermatological infectious emergency department visits It was noted however that these values were not significant or sustained The use of ACES during the G8 and G20 Summits allowed for real time monitoring of identified public health risks enabling situational awareness of potential emerging issues and enha
98. s of words and phrases found in the chief complaints Eight syndromes were defined for the EDSS system including for example gastrointestinal to capture all potentially infectious gastrointestinal conditions and fever influenza like illness ILI to capture all potentially infectious ILI conditions Since its inception in 2004 ACES has grown from its original two emergency departments at KGH and HDH to more than 115 acute care hospitals in 28 health units across Ontario with the long term goal of including all acute care facilities in Ontario In early 2015 the EDSS was renamed Acute Care Enhanced Surveillance ACES to reflect the broadening of the scope of the acute care partners participating in the system across Ontario as well as the system s enhanced capabilities The system continues to be maintained by KFL amp A Public Health with funding from the MOHLTC From the more than 115 acute care hospitals contributing triage and admission records to ACES approximately 12 000 visits and 3 000 hospital admissions are logged per 8 ACES Manual ACES Backgrounder day ACES like the EDSS continues to allow users to 1 monitor acute care hospital volume admissions and surge capacity to help prepare for high volumes of patients particularly in the event of an influenza pandemic 2 monitor trends and or changes in the incidence of endemic disease and 3 detect new or emerging public health threats The additional acute care coverage
99. spite these limitations in diagnostic precision CCs are of critical value to syndromic surveillance CCs are available in real time whereas vetted diagnostic codes e g most reasonable diagnoses codes MRDx using ICD 10CA coding are often not available for several days to weeks for an individual record The original EDSS system based on the early RODS system used only one type of NLP algorithm Naive Bayes NB to categorize the emergency department visits into its eight syndromes RODS NB text classifier was called the Complaint Coder CoCo and was found to be problematic in two key ways 1 the algorithm classifies using just single words not phrases and 2 the NB system assumes statistical independence between words in the CC In 2006 a new classifier was developed for use with the EDSS the new classifier was based on a Maximum Entropy ME model and analyzes the occurrence of character sequences rather than whole words In 2014 the EDSS was rebranded as ACES and now 46 ACES Manual The Science of ACES computes and displays the results of several text classifiers in addition to its default algorithm ME These are Balanced Winnow BW C4 5 Decision Tree C4 5 NB Winnow 2 and Monte Carlo ME MCME These algorithms are based on the MALLET Machine Learning Toolkit open source software developed at the University of Massachusetts Amherst In the ACES system the resulting classification for each algorithm can be displayed b
100. ss and possible outbreaks during the 2002 Olympic Games in Salt Lake City and has been deployed by several cities and US states to monitor both disease outbreak and pandemic In August of 2003 RODS was made freely available as open source software RODS categorizes emergency department visits using free text input from chief complaints included in electronic triage reports information that is regularly collected at virtually all emergency departments for administrative purposes Probabilistic algorithms categorize free text words or groupings of words into pre defined syndromic classes The earliest RODS syndromes reflect its origins in monitoring for bioterrorism 1 gastrointestinal nausea or vomiting diarrhea abdominal pain cramps swelling 2 constitutional non localized systemic complaints including fever faintness lethargy 3 respiratory congestion cough sore throat asthma pneumonia 4 rash any description of a rash 5 hemorrhagic bleeding from any site 6 botulinic ocular abnormalities difficulty speaking and swallowing 7 neurological non psychiatric complaints and 8 other pain or process in system area 43 ACES Manual The Science of ACES not monitored by RODS descriptions from Chapman et al Art Int Med 2005 33 1 31 doi 10 1016 j artmed 2004 04 001 The current ACES system has been developed from RODS and was first called the Emergency Department Syndromic Surveillance EDSS system It was in
101. ssifications Line Listings ED Section 2 5 1 Line Listings ED Tools Menu and Line Listings AD Section 2 6 1Line Listings AD Tools Menu 2 6 Line Listings AD In the same way that ACES enables you to view a textual list of individual emergency department visits hospital admissions can also be displayed according to the standard patient information associated with these records Specifically this information includes Date Time Admission Type Age Gender FSALDU Hospital Complaint and Patient LHIN Similar to Line Listings ED Line Listings AD will display the most recent data that you are registered to view as individual hospital admission in descending chronological order for the past week See Figure 17 for an example of admissions line listings Not that Admissions Type includes Elective Admissions and Emergency Admissions Admissions are categorized into syndromes in the same manner as emergency department visits see Section 3 3 Grouping the line listings by Admissions Type shows these two categories for admissions grouping can be achieved using the dropdown menu from the arrow at the top right corner of the table Figure 18 EXAMPLE OF LINE LISTINGS AD ADD WHEN SYNDROME CLASSIFICATION COMPLETE Figure 17 Example of Line Listings AD 30 ACES Manual User Interface Guide Date 7 Time Admission Type 4 Elective Admission 36 4 M 14 b 2015 05 13 8 gt 2015 05 14 6 4 F 22 gt 2015
102. t Health Unit TBDHU Thunder Bay District Health Unit THU Timiskaming Health Unit TPH Toronto Public Health 75 ACES Manual Appendices
103. t epicurve display seven or fourteen days including the 20 ACES Manual User Interface Guide current day and is often used in time Moving Average SEE berg so series data to smooth out short term None 7 14 Max None Std 1 Std 2 fluctuations When considered in Figure 7 Data display options concert with the standard deviation also calculated from the same seven or fourteen day period the moving average provides a simple smoothing technique for your data You can choose to display None one standard deviation Std 1 or two standard deviations Std 2 2 4 1 Epicurve Display Options Tools There are many options available to customize and optimize the Tools Advanced epicurve Be sure to carefully choose from the options available Health Unit Hospitals for each parameter as there are many options available to TOR All optimize data display and retrieval The Tools menu shown in Date Range Figure 8 has various options for sorting data Date From TN BE 04 07 2015 05 12 2015 V Health Unit choices available here are dependent upon Gender your administrative credentials All Male Female VI Hospital choose from the dropdown list the default ing is All Age setting is All B All 7 Date Range enter dates manually or click on date to ER choose from dropdown calendar the default setting displays the previous week ER Classifier Bucket Gender the default is All s2014 All lt Age choose from
104. ta series that would results in an Oscillation alert The emergency Figure 40 SPC Rule 5 Tendency department visit count on ti triggers alert since this count is the fourteenth point in a row that alternates direction or oscillates This test was introduced with the intent of 62 ACES Manual The Science of ACES describing a process that is likely out of control fourteen consecutive points that oscillate is statistically very unlikely and thus the data is not behaving as expected This INSERT GRAPH FROM ACES ALERTS OSCILLATION alert is therefore of little epidemiological concern for disease surveillance as it would Figure 41 SPC Rule 6 Oscillation not represent any particular kind of outbreak It is nevertheless of interest for ACES to monitor as it could indicate problems associated with data collection and synthesis 3 4 2 7 SPC Rule 7 Blissful Ignorance A Blissful Ignorance alert is generated when the current day s emergency department visit count is the fifteenth in a row where all fifteen counts are less than the UCL and greater than the LCL This provides a measure of data duality Under the assumptions of normal distribution it is very unlikely that fifteen data points in a row would not exceed one standard deviation difference from the mean the probability of this occurrence is 0 33 For the purposes of a manufacturing process such a pattern may indicate that there is insufficient variation in th
105. the values of each level of B _ Q4 Most amount of visits values Q2 the next 25 etc Figure 25 shows a legend for a data display of Figure 25 Quantiles Number of Classification 4 hospital visits as quartiles geography are divided into fourths Q1 representing the lowest 2596 of the Standard Deviation is used when you prefer to display the data on the map in x Emergency Department Visits Standard Deviation reference to the average for the whole province for the selected syndrome or Sample Data in comparison to the chosen level of geography For example the legend shown in Figure 26 displays the gradients between greater than three standard EH deviations above orange or less than three standard deviations below royal 04 1 blue the mean ADDITONAL OPTIONS NEED TO BE DISCUSSED 0 1 EET 2 7 1 7 Date Range Gender Age Group Age Range B Scrolling down the tab with the dark gray slider will reveal several more Figure 26 Standard choices for data display including Date Range Gender Age Group and Age Deviation Range Date Range can be entered manually mm dd yyyy or using the dropdown calendars Age Group include some pre defined age groups that may be helpful for your analyses When all desired parameters are entered press the Request Data button to generate the map Click Clear Data to revert back to the default settings Please note that you can choose as wide a date range as you w
106. ting methods CDC 2004 Standard disease surveillance techniques include modelling disease incidence and prevalence as well as the geographical distribution and the spread of disease With the ability to automate disease and even syndrome specific data the current interest is to collect data in real time and at regular intervals in order to quickly detect important changes in the data series and facilitate timely public health interventions ACES fulfills both of these constraints 1 automated and validated classification into medically relevant syndromes and 2 aberration detection using time series focused algorithms whereby data is synthesized in real time to detect aberrations as they occur An aberration is defined as when a time series exceeds a certain value or behaves in a way that is not likely to have occurred by chance alone ACES employs four families of syndromic surveillance algorithms to detect trends in emergency department visits from across Ontario in real time which are out of the ordinary and possible public health threats Aberrations may indicate an outbreak of public health consequence the beginning of a seasonal trends or problems with data transfer or submission The four different families of alert algorithms employed by the ACES system are 1 cumulative sum CuSum alerts 2 Statistical Process Control SPC alerts and 3 an exploratory alert method developed by the National Research Council NRC All three families
107. ublic Health s data centre over the secure Ontario e Health network The data collection process is shown in Figure 1 2 1 2 Data Processing All data collected from each participating emergency department is stored at KFL amp A Public Health s secure data centre Each visit from our acute care partners are classified into one of ACES pre defined syndromes medically significant categories based on the chief complaint or admission diagnosis see APPENDIX C ACES Syndromes The classification process uses algorithms derived from natural language processing NLP The various phrases words or parts of words found in the chief complaint are used to classify each visit into the syndrome it most likely belongs At this stage anomaly detection 16 ACES Manual User Interface Guide Data Flow Cycle for Acute Care Enhanced Surveillance ACES Step 8 Application and reporting is Step T delivered to clients over the Anomaly detection algorithms analyze the public internet using Transport data for trends or unusual events and Level Security encrypted via 256 Firewall Rout Public Internet KFL amp A Firewall Router Hospital Realm Data is routed point to point via FTP SFTP or direct HL7 Socket technologies available to each hospital Figure 1 ACES data collection and flow cycles methods are used to discern any potential trends or daily visit increases that may be statistically distincti
108. ut the default algorithm is ME Distinguishing features of each algorithm will be described in the following sections 3 3 1 Naive Bayes The original text classifier used by EDSS and developed by RODS was based on the Naive Bayes NB algorithm In comparison to other text classifiers NB classifiers use the simplest mathematics require relatively less resources in terms of computational power i e CPU and memory or otherwise i e training time size of training dataset and are amenable to coding in all standard programming languages They are based on the conceptual framework of Bayesian statistics in contrast to the inductive reasoning of frequentist statistics Bayesian statistics approach probabilities as independent hypotheses in a deductive manner For example a NB text classifier treats each term i e word word portion or phrase within the phrase being classified as independent variables NB classifiers tend to perform very well in many complex real world problems despite this oversimplification and are used in a variety of familiar text classification techniques including sentiment and spam detection email sorting and the detection of sexually explicit content 3 3 2 Maximum Entropy Maximum Entropy ME is the default text classifier used by ACES Unlike NB ME text classifiers do not assume the terms are independent of each other The ME classifier is particularly useful when nothing is known regarding the prior dist
109. ve suggesting at worst an outbreak see Section 3 6 Alerts and Outbreaks In the event that an abnormal number of visits for a syndrome are detected alerts generated by the ACES system are immediately posted to the website http aces kflaphi ca Epidemiologists and other health professionals can then use the secure web based interface to monitor the collated information and assess the emergency department visits that caused the alert They can determine if there are any patterns related to demographics location or timing of the cases that would justify further investigation by public health staff The data is collected in real time and is based on disease symptoms rather than diagnosis therefore ACES improves opportunities for both early detection and response to public health threats 17 ACES Manual User Interface Guide 2 1 3 Data Security and Privacy Although all data collected for ACES from participating hospitals are stripped of key identifiers i e name health card number residential address full postal code the data are nonetheless treated as personal health information under the Personal Health Information Protection Act PHIPA and steps are taken to protect the security and confidentiality of all information The ACES Privacy and Confidentiality Charter outlines the policies principles and procedures for ACES that are necessary to meet the intent of PHIPA A Privacy Impact Assessment has been conducted for ACES and
110. was operational with only the hospitals within the KFL amp A Public Health catchment area Use of the EDSS allowed for real time monitoring and identification of patients faster than would have occurred previously as well as a faster linkage of front line health workers to the suspected cases for identification of the outbreak vector The EDSS system clearly facilitated faster response from and better communication between both public health and acute care workers For a more detailed discussion of these results see CJEM 2008 10 2 114 9 EDSS researchers presented results verifying and validating the EDSS system for respiratory illnesses with the both the National Ambulatory Care Reporting System NACRS and Telehealth Ontario Emerg Inf Dis 2009 15 5 799 Very strong Spearman s correlations between the ED visits categorized as respiratory illness with both Telehealth results r 0 91 and NACRS r 0 98 confirming that the EDSS provides timely data collection and surveillance of respiratory illnesses Expanding upon these results the efficiency of the EDSS system over primary care sentinel surveillance for example is described in a paper from 2013 Can J Inf Dis Med Microbiol 2013 24 3 150 The authors report on using ED visits 44 ACES Manual The Science of ACES categorized by the EDSS as either ILI or respiratory as a proxy for the detection and prediction of actual lab confirmed cases of respiratory viral diseases and ILI Most int

DRAFT User Manual – June 2015 - KFL&A Public Health Informatics

Contents

Download Pdf Manuals

Related Search

Related Contents

DRAFT User Manual &ndash; June 2015 - KFL&amp;A Public Health Informatics

Contents

Download Pdf Manuals

Related Search

Related Contents

DRAFT User Manual – June 2015 - KFL&A Public Health Informatics