Syndromic Surveillance The Municipal Public Health The Municipal - - PDF document

2/16/2011 1

Syndromic Surveillance

The Municipal Public Health The Municipal Public Health Experience

For ehealthinformation For ehealthinformation Feb 16, 2011 Cameron McDermaid MHSc Epidemiologist cameron.mcdermaid@ottawa.ca

Syndromic Surveillance

 What is surveillance?

“ongoing systematic collection, analysis and interpretation of

• utcome-specific data for use in the

planning, implementation and evaluation of public health practice”

2

p p

Centers for Disease Control and Prevention

2/16/2011 2

Syndromic Surveillance

 Syndrome:

• The sum of symptoms and signs
• The sum of symptoms and signs
• f any morbid state.
• Used in clinical medicine for

triage, investigation, and initial management

• Does not require a “confirmed”

3

q diagnosis

Syndromic Surveillance

 Syndromic Surveillance:

• Infer disease from patterns/syndromes

p y in existing data streams.

• Many possible data streams
• Case classification accomplished by

data mining health records

4

2/16/2011 3 SYNDROME

Why Do Syndromic Surveillance?

Syndromic Surveillance Notifiable Disease Reporting

umber of Cases DIAGNOSED ILLNESS S O EXPOSU RE

5

Nu

Time

Emergency Room CBRN Attack Detection by Medical Records Surveillance (ECADS)

 Funding: CRTI  PI: Richard Davies  Scientific Team

Scientific Team

• University of Ottawa Heart Institute
• Michigan State University National Food Safety and

Toxicology Center

• Carnegie Mellon University Auton Laboratory
• Grey Bruce Public Health Unit
• All Grey Bruce area Hospitals

 Federal Government Partners N ti l R h C il I tit t f M i

6

• National Research Council, Institute for Marine

Biosciences (NRC/IMB)

• National Research Council, Institute for Information

Technology (NRC/IIT)

• Public Health Agency of Canada, Laboratory Centre for

Disease Control (LCDC)

2/16/2011 4

Emergency Room CBRN Attack Detection by Medical Records Surveillance (ECADS)

 ECADS Technical Team

• AMITA Corporation
• Performance Support Services Inc

pp

• Cam Emergency Preparedness
• e-Privacy Systems Inc

 ECADS Collaborators

• CNPHI Project, PHAC
• Office for Public Health Practice, PHAC

(Centre for Surveillance Coordination)

7

• Canada Health Infoway
• MOHLTC-funded QUESST Project (Ontario

Syndromic Surveillance Pilot, KFLA Public Health Unit)

• Michigan State Department of Public Health

Background Data

 Friday May 19, 2000

• Call from pediatrician, home for aged.
• Call/Faxes to schools, hospital, PUC, Lab

 Saturday May 20

• Presumptive e. coli results

Calls/FAX to hospitals PUC

• Calls/FAX to hospitals, PUC

 Sunday, May 21

• E. coli confirmed, presumptive water samples, cultures
• btained
• Outbreak number assigned, OMT formed, boil water

advisory

 Monday, May 22

• Patients contact, food interview sheet.
• OMT Expanded, treatment protocol, hotline established

8

p , p ,

 Tuesday May 23

• LPHL advised 2 water samples collected Sunday +ve E.

coli

• Joint HU and hospital press conference
• Physician meeting
• All health units notified via Ministry of Health notice

2/16/2011 5

The redder the dot, the higher the proportion of ER visits classified as GI. May 15, 2000 The bigger the dot, the greater the number of ER visits classified as GI

9

May 16, 2000

10

2/16/2011 6

May 17, 2000

11

May 18, 2000

12

2/16/2011 7

May 19, 2000

13

May 20, 2000

14

2/16/2011 8

May 21, 2000

15

May 22, 2000

16

2/16/2011 9

May 23, 2000

17

May 24, 2000

18

2/16/2011 10

May 25, 2000

19

May 26, 2000

20

2/16/2011 11

May 27, 2000

21

May 28, 2000

22

2/16/2011 12

May 29, 2000

23

May 30, 2000

24

2/16/2011 13

May 31, 2000

25

Total ER Visits to 10 Grey Bruce Area Hospitals

All ER Visits

250 Count Count 500

26

Jan 1, 1999 Dec 31, 2001

200 400 600 800 1000 1200 DAY 00 200 400 600 800 1000 1200 DAY

2/16/2011 14

GI Syndromes vs. Total Visits 3 Year Data Set - 9 Hospitals

All ER Visits GI Syndrome Count 250 500 Count

27

Jan 1, 1999 Dec 31, 2001 DAY 200 400 600 800 1000 1200 00 200 400 600 800 1000 1200

GI Syndromes 3 Year Data Set - 9 Hospitals

150 50 75 100 125 Count

28

Jan 1, 1999 Dec 31, 2001 DAY 200 400 600 800 1000 1200 25

2/16/2011 15

All ER Visits

SBG Health Centre (Walkerton)

All ER Vi it 200 All ER Visits 100 150 Count Count

29

Jan 1, 1999 Dec 31, 2001 200 400 600 800 1000 1200 DAY 50 200 400 600 800 1000 1200 DAY

All Visits vs. GI Syndrome: SBG Health Centre (Walkerton)

All ER Vi it 200 All ER Visits GI Syndrome 100 150 Count Count

30

Jan 1, 1999 Dec 31, 2001 200 400 600 800 1000 1200 DAY 50 200 400 600 800 1000 1200 DAY

2/16/2011 16

GI Syndrome

SBG Health Centre (Walkerton)

125 50 75 100 Count

31

200 400 600 800 1000 1200 DAY 25 Jan 1, 1999 Dec 31, 2001 Boil Water Advisory

40 60 80 100 120 140 160

Culture confirmed Culture negative Not tested

Symptom onset for 1335/1346 Cases

20 40 14 17 20 23 26 29 2 5 8 11 14 17 20 23 26 29 1 4 7 10 13 16 19 22 25 29

Visits to ER by Walkerton Residents classified as GI

32

Visits to Walkerton ER classified as GI

DAY

2/16/2011 17

ECADS Retrospective Analysis

• f Walkerton Outbreak

 Syndromic Surveillance highly sensitive

to outbreak of this nature to outbreak of this nature

 Would definitely have confirmed

concerns of physician on Friday, May 19 and provided data for outbreak investigation

 Would likely have detected a GI

33

 Would likely have detected a GI

• utbreak centered in Walkerton on

Friday May 19 had physician not called

SYNDROME

Why Do Syndromic Surveillance?

Syndromic Surveillance Notifiable Disease Reporting

umber of Cases DIAGNOSED ILLNESS S O EXPOSU RE

34

Nu

Time

2/16/2011 18

Selection of Surveillance Point

e.g. Google Trends ISIC e.g. EMR Sentinel physicians

Information Seeking Self Care Primary Contact Care Emergency Rooms

e g OTC Antiviral e g

35

Telehealth e.g. OTC Antiviral e.g. Screening data Data feeds

Exploiting the Advantage

 Confidence that what you’re seeing

is relevant to you is relevant to you

 Requires a protocol that allows you

to act on what you’re seeing

36

2/16/2011 19

ASSET

 Advanced Syndromic Surveillance &

Emergency Triage (ASSET) Ver 1 g y g ( )

 Based on RODS version 3 with a

number of patches

37

ASSET Partners and Collaborators



CRTI

• Norm Yanofsky



NRC -IIT

• Janice Singer , Norm Vinson
• Joel Martin, Berry De Bruijn



Ottawa Public Health



PHAC Foodborne, Waterborne and Zoonitic Diseases

• Paul Sockett
• Victoria Edge



PHAC CNPHI Project

• Amin Kabani
• Amira Ali
• Isra Levy



Ottawa Heart Institute Team

• Susan McClinton
• Jason Morin
• Debbie Warren



Grey Bruce Public Health

• Hazel Lynn
• Alanna Leffley



Grey Bruce Area Hospitals



Ottawa Area Hospitals

• Shamir Muchti



US Partners

• Michigan State University –

Ewen Todd

• Carnegie Mellon University

Auton Lab –Daniel Neill

• Michigan Dept of Community

Health – Melinda Wilkins and Jim Collins



Consultants

• Stephen D’Silva
• Gini Bethel

38



Ottawa Area Hospitals



AMITA Corp

• Sonny Lundahl
• Monica Preston
• Anu Pinnamanini
• Gini Bethel
• Monty Montgomery
• John Boufford



Collaborators

• Kieran Moore – KFLA Health

Unit

• Greg Webster – CIHI
• Altarum Corp – Rick Keller

2/16/2011 20

Asset in Ottawa

 Four hospitals provide data to ASSET  Data are categorized and analyzed

Data are categorized and analyzed every six hours

 Data are housed at the Ottawa Hospital  Sends an email alert if the number of

cases is higher than expected S stem is monitored each da b OPH

39

 System is monitored each day by OPH

epidemiologists

40

2/16/2011 21

ASSET in OTTAWA ASSET in OTTAWA

 Data includes:

• Sequential case ID
• Admission date
• Case sex
• Case age
• 5 digit postal code
• Symptom/chief complaint

S d

41

• Syndrome

 Not all sources

• Final Diagnosis
• CTAS

Processing in ASSET

 Case is categorized using chief complaint  Tried to jump off two story house  Tried to jump off two story house  Fish hook  Fell and laid in yard for a couple of

hours

 “I have fleas”

42

 Stuck bead in nose  Squirrel bite

2/16/2011 22

Processing in ASSET

 Cases are classified using a classifier

developed by the National Research developed by the National Research Council of Canada.

 Expert classifies symptoms using sample

data from the Ottawa hospitals.

 The classifier can be retrained or learn new

classifications

43

classifications.

 Allows a very regional approach

The Standard Syndromes

 Gastrointestinal  Respiratory

A th

 Asthma  ILI  Constitutional  Hemorrhagic  Botulinic

44

 Botulinic  Neurological  Rash  Other

2/16/2011 23

Alert Algorithms

 Recursive Least Squares (RLS)

based on past 120 days based on past 120 days

 CUSUM based on last 14 days  Alerts = excess of 3 SD of

predicted

 No alert will occur if case count <5

45

 12:15 AM, 6:15 AM, 12:15 PM and

6:15 PM

• 1. Describe

Alert ASSET Alert CONCERN

• Large case excess
• Severe symptoms
• High CTAS
• Final diagnosis of public health

importance

• Multiple hospitals involved
• Demographic clustering
• Low case excess
• Single hospital involved
• 2. Contextualize

Alert

• 3. Case based

Analysis

• Consistent with expected seasonal

pattern or historic trends

• Alert is new
• Few new cases with sustained alert
• Alerts maintained
• Ongoing case count

excess

• Geographic clustering
• Multi alert clustering
• Differences from similar historical

alerts

• Multiple syndrome patterns

4 Internal

• Cases validated from benign cause
• No evidence of clustering or common

source

46

NO

• 4. Internal

Consultation

• Concurrent events
• Validated from ER

Document Next Steps YES Action Required

2/16/2011 24

What is the response?

• 1. An alert is triggered based on the

applied algorithm applied algorithm

• 2. Email is received by OPH

epidemiologist

• 3. System access via 2 step security portal

47

An alert is triggered

48

2/16/2011 25

Describe the Alert

Total Syndrome counts in O Respiratory Counts Ottawa Normalized Counts

49

Describe the Alert

 Software used is STATA

50

2/16/2011 26

Describe the Alert

51

Describe the Alert

52

2/16/2011 27

Describe the Alert

53

Context for Alert

 Direct communications with ER

S h l hild LTCF tb k

 School, childcare, LTCF outbreaks  FRI screening in hospital settings  Telehealth alerts  Adjacent health units

54

2/16/2011 28

Context of Alerts - Telehealth

55

Context of Alerts - Telehealth

56

2/16/2011 29

Dissemination

 Still a work in progress  Internal reporting  Internal reporting

• Planning

 External reporting

• Communication to partners
• Communication to the public e.g. PSA

57

GI complaints in early 2010

 January 5th alert for GI

Al t t i d

 Alert was sustained  Norovirus was known to be in the

community

 2 local hospitals had increased GI

amongst admitted patients w/ SRV

58

amongst admitted patients w/ SRV confirmed at one.

 PSA and childcare notifications sent

2/16/2011 30

ASSET in Ottawa

59

Ongoing ECADS Surveillance in Grey Bruce

• Monitoring for GI illness following water

filtration plant problem Jan 2006 Ch t i ti f R i t O tb k F b

• Characterization of Respiratory Outbreak Feb.

2006

• Detection of Cryptosporidium outbreak July

2006

• Characterization and tracking of simultaneous

norovirus and influenza outbreaks Dec 2006 - Jan 2007

60

Jan 2007

• Monitoring for GI illness following e. coli

contaminated strawberries Summer 2007

• Detection of Scarlet Fever Outbreak Oct-Dec

2007

2/16/2011 31

Next Steps

 Testing of Response Protocols

O i C lt ti ith P t

 Ongoing Consultation with Partners

• Would it help?
• What would it look like?

 Mapping based on 5 digit postal

code

61

code

 Analysis for concurrent validity  Better parsing of symptoms

62

2/16/2011 32

Thank you for your attention!

• Google Trends: http://www.google.com/trends
• ISIC: http://sites.google.com/site/diseasesurveillance/Home
• KFLA live maps: http://www.kflainfectionwatch.com/KFLAMapping/livemaps.aspx
• RODS: http://openrods.sourceforge.net/index.php?page=download
• Fluwatch http://origin.phac-aspc.gc.ca/fluwatch/

Answers to additional questions:

• How much data is in ASSET? We have ASSET data from June 01, 2007
• Why was RODS selected as the basis of the system? At the time of inception, only

RODS did not require manual data capture and was available to Canadian public health units.

• What is the accuracy of the classifier? The classifier accuracy was tested against

63

• What is the accuracy of the classifier? The classifier accuracy was tested against

ILI classification by experts (gold standard). Results were precision 97.4%, recall 97.5%, and 99.7% specificity. This work was presented and is available here: http://www.syndromic.org/conference/2009/presentations/Thursday330/vinson.pdf Please don’t hesitate to send me any other questions about this topic that you may have.