The Distribute Project Emergency Department Surveillance for Influenza-like Illness Taha A....
-
Upload
dominic-malone -
Category
Documents
-
view
214 -
download
0
Transcript of The Distribute Project Emergency Department Surveillance for Influenza-like Illness Taha A....
The Distribute ProjectEmergency DepartmentSurveillance for Influenza-like Illness
Taha A. Kass-Hout, MD, MSDeputy Director (Acting) for Information Science (Acting),
Division of Healthcare Information (DHI), Public Health Surveillance Program Office (proposed)Office of Surveillance, Epidemiology, & Laboratory Services (OSELS)
Centers for Disease Control & Prevention (CDC)
Wednesday, April 14, 2010 Atlanta, GA
PHIN Partner Call
Overview
• Distribute: Background & Objectives• System Description• Data display & use• Comparison to ILI-Net• Community of Practice (CoP)• Assessment and “Trade-offs”• Next steps
Syndromic SurveillanceBackground
• Automated systems to monitor manifestations of illness– Health care/service use by symptom category
(syndromes), typically based on “chief complaint”– Emphasis on timeliness & daily reporting, analysis,
& visualization– Original impetus: Early detection of large-scale
bioterrorism-associated event– Evolution absent BT: Multiple uses for “situation
awareness”
Syndromic Surveillance Practice International Society for Disease Surveillance (ISDS) Survey, 2007-2008
52 HDs responded (46 states, DC, 3 local HDs, 2 terr): 83% conduct SS
Source: Advances in Disease Surveillance, 2008 http://www.isdsjournal.org/article/view/2618/2517
84%
Utility of syndromic surveillance for:Highly
UsefulSomewhat
UsefulUndecided
Not Useful
Influenza monitoring 52% 40% 7% 0%
Large area trend monitoring 47% 33% 16% 5%
Ad hoc analysis 28% 42% 23% 7%
Outbreak detection 7% 33% 16% 44%
Epidemiologists’ qualitative assessments of ED-based SS & other influenza surveillance systems (2006-2007 Season)
‘‘…when you look at them all together, you get a really good picture of what’s going on in the state. I don’t think any one system is a good stand alone surveillance system. It’s being able to take all of those pieces and put them together and then that sum total of information [is useful].’’
‘‘I think that the populations that are seen through these different systems provide a different perspective…as it starts throughout the year. I mean the ED surveillance is slightly different than sentinel flu providers and laboratory data, so it each provides a piece of the picture to create more of a whole….’’
‘‘[T]he syndromic surveillance really…was an added enhancement…because it is more timely and we are able to see and pull out those age specific groups.’’
‘‘…[W]e had…a few calls about schools that were seeing higher absence rates of influenza-like illness as well as some providers in those communities…were reporting a higher number of positive rapid tests in people coming in and we were able to get specimens collected…that…showed influenza B in particular starting at that time…. [With] syndromic surveillance, very clearly cases…were showing up in children…which we can’t really see in the traditional sentinel provider program.’’
Source: Biosecurity & Bioterrorism, 2009, http://www.liebertonline.com/doi/pdfplus/10.1089/bsp.2009.0013
BackgroundILI-Net
• ILI-Net: Influenza-like illness (ILI) monitoring:– Long-standing CDC & state partnership (all states, DC, USVI)– Part of multi-faceted mosaic of influenza surveillance methods– ILI = fever (temp ≥ 100°F [37.8°C]) and cough or sore throat absent a KNOWN
cause other than influenza. – Non-specific but useful in context of other flu tracking methods, esp. lab– >3000 providers with mix of provider types, 30 million visits per year– 2007-2008: CSTE recommended expanded use of electronic data sources,
reflecting emergence of SS – Reporting cycle: “MMWR week”– Public data presentation by HHS region
http://www.cdc.gov/flu/weekly
Background2009
• President’s Council of Advisors on Science and Technology
recommended expanded use of ED data SS data
• New CDC Director accustomed to daily use of ED SS data for influenza and other situation awareness in NYC
• CDC funded Public Health Informatics Institute (PHII) to support rapid scale-up of ISDS Distribute project
Distribute Objectives
• Improve H1N1 situation awareness/Supplement ILI-Net– Daily HD-specific ILI trends based on ED visits– Scale-up small scale “proof-of-concept” pilot to nationally
representative coverage– Add measures of ILI severity (measured fever, admission) where
possible
• Foster “distributed” approach: Engage partners’ capacities & expertise– HDs: Investments in ED SS for ILI monitoring– PHII (www.phii.org): Project planning & mgmt– ISDS (www.syndromic.org): SS leaders in state/local HDs and academia– CDC (OSELS & NCIRD): Surveillance, influenza, & informatics
DistributePrinciples & Characteristics
• Aggregates counts of ILI and total ED visits from existing health dept ED-SS systems– Cross-tabulated by a limited number of variables
• Age group, 1st 3 digits of zip code, date of ED visit• Temperature and disposition (hospital admission) where possible
(proxy measures of illness severity)• Allows flexibility in use of “chief complaint” ILI Syndromic
criteria that HDs had already developed– Draws on state/local familiarity with:
• Local language/lingo and health care services• Calibration of ILI syndromic criteria to optimize alignment of
observed trends with other local influenza markers• Fosters “community of users”
Distributed Responsibilities
• Project management: ISDS, PHII, OSELS• Engage HDs: ISDS & OSELS
– Outreach & Data use agreements – Technical assistance– Receive & manage data
• Trend validation & comparison to ILI-Net: NCIRD & OSELS and vetted with ISDS
• Data display: – ISDS, public-access and contributors login-in site (U Wash & Harvard):
http://ISDSdistribute.org – CDC: Intranet site
• Analytic methods development: OSELS, NCIRD and ISDS (includes academic partners)
• Support community of users: ISDS, PHII– Distribute Advisory Board (inc. ASTHO, CSTE, NACCHO, OSELS and NCIRD)– Meetings, Webinars, Social networking
Distribute Information Pathways
Hospitals HD SS system Distribute
Hospitals HD SS system DistributeCDC BioSense
Hospitals HD SS system Distribute
Hospitals DistributeCDC BioSenseHospitals DistributeCDC BioSense *
* With HD approval
Via ISDS (UW) or CDC
Original Drawing: Taha Kass-Hout (CDC)Amended by: Bill Lober (ISDS)Date: 31 Oct 2009
Sites Reporting to ISDS
Sites Reporting to ISDS
Sites Reporting to CDC
Sites Reporting to CDC
Sites Reporting to Distribute via BioSense
Data Management by CDC
Data Analysis and Visualization by CDC
Sites Reporting to Distribute via BioSense
Data Management by CDC
Data Analysis and Visualization by CDC
Data Management by ISDSData Management by ISDS
Data Visualization by ISDSData Visualization by ISDS
Original Drawing: Taha Kass-Hout (CDC)Amended by: Bill Lober (ISDS)Date: 31 Oct 2009
Participation Status Number of HDs Cumulative Number of HDs (Not overlapping)
Prior (Aug 2009) Distribute participant 10 10
Prior (Aug 2009) HD SS system connected to BioSense 8 15
Additional HDs with commitment to participate 32 47
Data sharing with Distribute established & Data visualized on Distribute public-access web site (http://ISDSdistribute.org)
34
DistributeHealth Dept Participation as of April 14th, 2010
Distribute Data Visualized on Distribute Public Website: N=34
DistributeBioSense Contribution
• Pathway for data flow for 9/34 jurisdictions– 4 with HDs SS systems reporting to BioSense– 5 with direct reporting hospitals to BioSense
• Only source of usable data on measured temperature and patient ED disposition
• Leveraged developmental prototypes – Geocoded Interoperability Population Summary Exchange (GIPSE) data
specification– HIE projects facilitated access to ED data (NY Upstate, WA state, IN state) – Centers of Excellence in PH Informatics (Aegis Platform, Harvard CoE)
• BioSense staff supported project mgmt, data mgmt, analyses, visualization, & reporting
← Sept 2009, Percentage of ED visits due to ILI at BioSense hospitals increasing
Yellow: ED ILI patients hospitalized / ED ILI patients
→Red: ILI among patients hospitalized / Total ED patientsGreen: ILI among patients hospitalized / Total ED patients
Disposition (Admission) of ED ILI Patients
DistributeData Sources & ED Visit Coverage Estimates
• 34 Health departments: – EDs: 1,319– Urgent Care Clinics: 72 – Other : 12
• 68% of HDs (23/34): Only EDs• Number of EDs per Health Dept
– Range: 1 – 140– Median: 18
• ED visit coverage estimates (from 22 of 34 HDs)– Range: 15 - 100% – 8 HDs: ≥ 90% (MO, ND, NY, VA, NYC, Boston, Cook County, Tarrant County,
Seattle-King County, Tulsa County)
DistributeTimeliness
From April 1, 2009 thru Feb 17, 2010 Distribute recorded a total of 42,020,132 visits, averaging 130,093 visits/day.
DistributeTimeliness
• Available data from most recent day(s) may be unstable due to incomplete reporting from some hospitals within HD SS systems• Led to development of censoring procedure that was
applied in CDC intranet visualizations to exclude data from dates with incomplete reporting
DistributeRatio of Average Monthly ED Visits
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
1: CT, ME, MA, NH, RI,
VT
2: NJ, NY 3: DE, DC, MD, PA, VA,
WV
4: AL, FL, GA, KY, MS, NC, SC, TN
5: IL, IN, MI, MN, OH, WI
6: AR, LA, NM, OK, TX
7: IA, KS, MO, NE
8: CO, MT, ND, SD, UT,
WY
9: AZ, CA, HI,
NV
10: AK, ID, OR,
WA
USA
Ra
tio
: Dis
trib
ute
/ T
ota
l ED
vis
its
HHS Region (States Only)*Interval is more recent for health departments where historical data for 2007-2008 is not available
• ED source data– Scan text of ED chief complaint: 33 HDs – ED discharge diagnosis (ICD-9 code): 1 HD
• Mix of syndrome definitions used– Many HD-specific– “Broad” vs. “Narrow”
• Narrow: attempts to replicate ILI-Net definition, may exclude many with influenza due to brevity of CC recording
• Broad: less restrictive, yields parallel, higher amplitude signal
DistributeILI Syndrome Definitions
• All Sites (N=34) provide ILI-Broad
DistributeIndicators
Distribute Temperature, Disposition or Age Reporting
DistributeZip-3 Reporting
Public Site
Restricted (Contributor’s) Site
DistributeOnline Data
http://ISDSDistribute.org
DistributePublic Site
http://ISDSdistribute.org
http://ISDSdistribute.org
DistributePublic Site
DistributeComparison to ILINet
• Calculated % ILI by week by jurisdiction for both systems
• Compared systems using Pearson Correlation Coefficients and visually with time series graphs
DistributeComparison to ILINet
• State-based jurisdictions– Correlations ranged from 0.64 to 0.96 with mean and
median of 0.83 and 0.83, respectively
• Local-based jurisdictions– Correlations ranged from 0.38 to 0.91 with mean and
median of 0.76 and 0.81, respectively– Visually, major peaks in % ILI in the 2 systems tracked well
together
Distribute Community of Practice
• Approximately 90 state and local epidemiologists
• Representing 43 health departments• Wide range of expertise in syndromic
surveillance
Distribute Community of Practice
Pre-meeting survey results, December 2009 CoP meeting
Distribute Community of Practice
Pre-meeting survey results, December 2009 CoP meeting
Distribute Trade-Offs
• Timeliness– Possible to collect and display daily, HD-specific ILI
data (2-3 day lag for most HDs, including censoring for dates with incomplete reporting)
– Instability of daily data: 2-3 day “trends” not consistently born out by subsequent observations
– Reporting lag of 1 - 3+ days, includes censoring of data for most recent date(s) if hospital reporting is incomplete from a HD (applied for CDC Intranet visualizations)
Distribute Trade-Offs
• Flexibility in ILI syndrome criteria– Allowed rapid participation—did not require HDs to
change syndrome classification procedures. – Allowed use of criteria “validated” by state/local flu
surveillance experience– Allowed cross-site comparison of timing of up- and
down-swings in ILI trends, but…– Variability in amplitude of signal precluded
comparisons of H1N1 impact or summary estimate of H1N1 ED visits
Distribute Next Steps
• Identification of EDs participating in Distribute AND ILINet to prevent duplication
• Assignment of Distribute point of contact at health departments and collaboration with influenza coordinators
• Method to appropriately analyze variation in signals (censoring data)• Co-visualize with ILINet (CSTE Recommendation)
– CSTE recommended continuing to display Distribute data separate from ILI-Net
• Increase coverage (ongoing)• Address variability in ILI criteria
– Statistical: Apply methods from ILI-Net to define HD-specific baselines and standard assessments of relative departures from baseline (ongoing with Influenza Division & Johns Hopkins U)
– “Harmonize” ILI criteria (ongoing with Distribute Community of Practice)
DistributeBaseline Analysis
Multi-Site Comparison
Distribute Longer-term
• Role and utility– Aggregate vs. individual-level data?– HD-specific vs. regional data: who controls access? – How timely is timely?
• Needs for PH decisions/value of additional data source• Expectations & perceived credibility, e.g., media, policy makers
• Value of collaboration & partnership– Influenza surveillance– Other conditions, e.g., seasonal gastroenteritis– Ongoing development of automated surveillance methods
DistributeAcknowledgements
• ISDS Staff and Volunteers• Public Health Informatics Institute (PHII)• Project liaisons from NACCHO/CSTE/ASTHO• Support to ISDS
– Tufts Health Care Institute (THCI)– Markle Foundation
• CDC:– NCIRD, Influenza Division– OSELS & former NCPHI – H1N1 response team– OPHPR
Taha A. Kass-Hout, MD, MSDeputy Director (Acting) for Information Science (Acting),
Division of Healthcare Information (DHI), Public Health Surveillance Program Office (proposed)Office of Surveillance, Epidemiology, & Laboratory Services (OSELS)
Centers for Disease Control & Prevention (CDC)
Wednesday, April 14, 2010 Atlanta, GA