NYC Syndromic Surveillance

IFH HIT Meaningful Use Workshop10/1/2010

Marlena Plagianos, MSNYCDOHMH

mplagian@health.nyc.gov

2

What is Biosurveillance?

• “Collection and integration of timely health-related information for public health action achieved through the early detection, characterization, and situation awareness of exposures and acute human health events of public health significance.”

Aaron T Fleischauer, PhD; Pamela S Diaz, MD; Daniel M Sosin MD . Biosurveillance:

A Definition, Scope and Description of Current Capability for a National Strategy.

Advances in Disease Surveillance 2008;5:175

3

Traditional Surveillance

• Case definitions• Historically low

compliance• Laboratory

confirmation can be slow

• Still important (e.g. H1N1 in NYC)

4

Laboratory Confirmation

•Making firm diagnosis commonly relies on lab result•Limited in-house testing in outpatient setting (minutes)•Commercial laboratory testing takes time (days-weeks)

5

Traditional Reporting is Labor Intensive

6

Traditional Reporting is Labor Intensive

7

Syndromic Surveillance

• Pre-diagnostic indicators of disease• Readiness scenarios: bioterrorism,

pandemics• Objectives:

– Timely, sensitive, specific surveillance– Detect outbreak before ‘astute clinician’

• Typical Process

Collect data

Process & code data

Establish baseline

Identify outbreak

Sound alarm

8

New and Exciting Data Types

Data source Level of data

Data type Setting Care phase

Medication sales Aggregate Drug category Pre-clinical Pre-diagnostic

School absences Aggregate Frequency Pre-clinical Pre-diagnostic

Nurse hotline call Individual Call type Pre-clinical Pre-diagnostic

Chief complaint / Reason for Visit

Individual Text, brief Clinical Pre-diagnostic

EMS call Individual Run type Clinical Pre-diagnostic

Temperature Individual Vital sign Clinical Pre-diagnostic

Radiology Report Individual Text, narrative Clinical Pre-diagnostic

Chest X-ray Individual CPT code Clinical Pre-diagnostic

Diagnosis code Individual ICD9 code Clinical Diagnostic

Progress Note Individual Text, narrative Clinical Diagnosis

9

EHR Syndromic Surveillance

• The Primary Care Information Project (PCIP) uses different EHR data sources to conduct & pilot its syndromic surveillance activities

• Some syndromes tracked using EHR data are:– Influenza-like Illness (ILI)– Fever– Gastrointestinal Illness (GI)

• Case definitions for these syndromes based upon text in these structured fields:– Chief Complaint– Measured Temperature– Diagnosis (ICD-9 CM Code)

10

System Screenshot

11

Aggregate Level Syndromic Data

• Only “Count” Data is collected

12

Data processing and syndrome coding

Respiratory conditions

%Macro Resp; *Respiratory; IF CC=:'COUGH' OR CC=:'COUGHING' OR CC=:'SOB' OR CC=:'DIFFICULTY BREATHING' OR CC='BREATHING PROBLEMS' OR CC=:'SHORTNESS OF BREA' OR CC=:'DIFF BREA' CC='URI' ORTHEN RESP=1; ELSE DO; RESP=

Misspelling INDEX(CC,"COUG") + INDEX(CC,"COUH") +Shortness of breath

INDEX(CC,"S.O.B") + INDEX(CC,"SOB") + INDEX(CC,"S O B") + INDEX(CC,"S O B") + INDEX(CC,"S.OB");

Difficulty breathing

INDEX(CC,"BREAT") + INDEX(CC,"BEATH") + INDEX(CC,"DIB") + INDEX(CC,"D I B") + INDEX(CC,"D.I.B") + INDEX(CC,"BRATHING") + INDEX(CC,"DIFF BR") + INDEX(CC,"DIFF, BR") +

Upper respiratory infection

INDEX(CC,"URI ") + INDEX(CC,"URI/") + INDEX(CC,"URI;") + INDEX(CC,"U R I") + INDEX(CC,"URI,") + INDEX(CC,"U.R.I") +

13

Analysis:Calculate Baseline

Expected disease level

• Approaches: Moving average, regression, time series methods.• Length of baseline: Years, months, days

Adjustments

• Long: Seasonal, secular, environmental (e.g. heat, pollen)• Short: Day of week, weekend/weekday, holidays, reporting failures

14

Analysis:Test Observed vs. Expected

Significance tests

Predetermined number of standard deviations

Crossing statistical thresholds Signal

15

Analysis:Test Observed vs. Expected

Electronic Health Record Syndromic SurveillanceDuring 2009 Pandemic

H1N1 in NYC

18

Friday

19

Saturday

20

Sunday

21

Monday

22

Tuesday

23

Wednesday

24

Thursday

25

Friday

26

Saturday

27

Sunday

28

Monday – Memorial Day

29

Tuesday

H1N1 in New York City:

Where did patients seek treatment?

Emergency Departments or

Primary Care Clinics?

31

Objective

• To determine whether the timing of the increase in patient visits was different at emergency departments from primary care clinics during the spring 2009 H1N1 influenza outbreak across the 5 boroughs of NYC

32

Study Sites

l 58 Primary Care Providers (PCP): – 9 Institute for Family

Health (IFH) clinics– 49 practices enrolled

in the NYCDOHMH PCIP (30 visits/day)

v 50 Emergency Departments– 247 visits/day

33

Methods

• Influenza-like Illness (ILI) as a broad estimate of H1N1

• Fever + respiratory related reason for visit or diagnosisPCP

• Chief complaint of fever + a sore throat or cough, or a chief complaint of fluED

34

Methods

Compared number of days to a significant increase at EDs to PCP clinics using a log-rank test

• City-wide• By borough to see if there was a geographic difference

Two Waves:

• 4/24-5/8• 5/14-6/4

35

ED, IFH and PCIP ILI Visits

ED, IFH and PCIP ILI Visits

ED, IFH and PCIP ILI Visits

38

Facilities with a significant increase in ILI volume

39

Results, April 24-May 8

Median Days to Increase in ILI

Facilities with Increase in ILI

1-sided log rank

Borough ED PCP ED PCP p-valueAll 4 12 43/50 (86%) 36/58 (62%) <0.0001

Bronx 5 12 8/9 (88%) 10/17 (59%) 0.045

Brooklyn 3 14 12/15 (80%) 6/9 (67%) 0.025

Manhattan 4 13 13/15 (87%) 11/19 (58%) 0.008

Queens 3 7 8/8 (100%) 6/7 (86%) 0.007

Staten Island 14 10 2/3 (67%) 3/3 (100%) 0.902

40

Results, May 14-June 4

Median Days to Increase in ILI

Facilities with Increase in ILI

1-sided log rank

Borough ED PCP ED PCP p-valueAll 4 8 47/50 (94%) 50/58 (86%) <0.0001

Bronx 1 6 9/9 (100%) 16/17 (82%) 0.004

Brooklyn 4 12 13/15 (87%) 7/9 (78%) 0.039

Manhattan 4 7 14/15 (93%) 17/19 (89%) 0.016

Queens 4 8 8/8 (100%) 5/7 ( 71%) 0.091

Staten Island 5 8 3/3 (100%) 3/3 (100%) 0.012

41

Findings

• Emergency Departments experienced an increase in patients with ILI before Primary Care Providers

• PCPs were vastly under-utilized during the outbreak

• NYCDOHMH changed messaging to encourage visiting PCPs instead of EDs for mild illness

42

Future of Syndromic Surveillance

Meaningful Use

• Capability to submit syndromic data to health departments Regional Health Information Organizations (RHIOS), Hubs

Data Validation and Quality Assurance

43

Online Resources

CDC Flu Surveillancehttp://www.cdc.gov/flu/weekly/fluactivity.htm

Distribute

http://www.isdsdistribute.org/