Syndromic Surveillance for Influenza in Washington State: A Local and Regional Perspective

1
Syndromic Surveillance for Influenza in
Washington State A Local and Regional
Perspective

• Nicola Marsden-Haug, MPH
• Atar Baer, PhD
• Hilary Metcalf, MPH
• Nigel Turner, RS, MPH
• Phyllis Shoemaker
• Jeff Duchin, MD
• 2007 Syndromic Surveillance Conference

2
Washington State Background

• 3 local health departments each operate their own
  systems
• Public Health Seattle King County (1999)
• Tacoma-Pierce County Health Department (2001)
• Kitsap Regional Health District (2001)
• Multiple data sources
• ED, urgent care centers, OTC, EMS,
  absenteeism, military outpatient
• Differs by county
• Pilot project to assess statewide operation
  (2006)
• Collect ED data from each region

3
ILI Evaluation Project

• Joint project between King, Kitsap, and Pierce
  counties and Washington State DOH
• Goals
• Describe how different visualizations (e.g.,
  count vs. percent, daily vs. monthly) can
  influence your conclusions regarding ILI
  surveillance
• Assess the value of trends at the local versus
  regional level
• Determine how predictive of other traditional flu
  indicators syndromic data are
• Define new system capabilities to assist with ILI
  surveillance

4
Methods

• Emergency Department (ED) data
• King (n 19), Kitsap (n 1), Pierce (n 5)
• Visits between September 2004 April 2007
• Aggregated and by individual county
• Used fever syndrome group as proxy
• Multiple syndrome groups (fever, flu, resp)
  suggested in the literature no standard
• Key words or ICD9 codes in chief complaint or
  diagnosis
• Fev, High temp, elevated temp, hi temp, temp10,
  shiver, feeling hot, feel hot, feels hot, night
  sweat, febr, pyrexia, 780.6, 780.99

5
Methods

• Corrected for missing historic data (Pierce)
• If census lt65 of 7-day moving average
• then replace census and fever counts with an
  average of respective counts 2 days before and
  after (4-day window)
• Visual assessment of trends and correlation
  analysis to quantify similarity
• Comparison with traditional flu surveillance data
• Reported school absenteeism
• Positive influenza isolates
• Pneumonia and influenza deaths
• Collected by DOH from county reports and
  laboratories
• Collected by CDC from 121 Cities Mortality
  Reporting System

6
(No Transcript)
7
On the same scale, its difficult to see any
trend in the Kitsap data.
8
On separate scales the similarity among trends is
visible.
King Kitsap r 0.61 Kitsap Pierce r
0.53 Pierce King r 0.87
9
Counts versus Proportions Daily Data
Number of fever visits per day
Proportion of all visits per day
10
Counts versus Proportions Weekly Data
Number of fever visits per week
Proportion of all visits per week
11
Counts versus Proportions Monthly Data
Number of fever visits per month
Proportion of all visits per month
12
Regional Fever Counts by Age Group September
2004 August 2006
13
Regional Fever Proportions by Age Group
September 2004 August 2006
14
Weekly Fever Proportion Trends by Age Group
Regional Aggregate (King Pierce Counties) Shown
34 26 16 29 19 9
lt2 yrs
2-4 yrs
18 10 3
5-17 yrs

• Strong seasonal signals in 2-4 yrs, 5-17 yrs, and
  18-44 yrs
• Primarily school age and parents
• Little (if any) signal in 45-64 yrs
• No signal in lt2 yrs and gt65 yrs

4 2 0
18-44 yrs
4 2 0
45-64 yrs
2.8 1.9 1.0
gt65 yrs
15
(No Transcript)
16
Correlation of Weekly Syndromic ED Fever Visits
with School Absenteeism and Positive Influenza
Isolates
17
(No Transcript)
18
Correlation of Weekly ED Fever Visits with
Positive Influenza Isolates Regional Aggregate
by Age Group
19
(No Transcript)
20
Conclusions

• Strong correlation of ED fever visits
• across counties
• with all-cause school absenteeism (for 5-17 yo
  age group)
• with viral isolates (particularly for 18-44 yo
  age group)
• Counts better for assessing magnitude
  proportions better for comparisons between years
  and counties, and for factoring out other
  all-cause seasonal trends
• Need to see both weekly and daily scales
• Local views are consistent for counties, so
  regional aggregate gave basic trend for each
• Fills in gaps for missing data in contiguous
  counties studied

21
Additional Work

• Are trends as similar in non-contiguous counties?
• When alerting algorithms are applied for signal
  detection, does local or regional data provide
  earlier warning?
• Does this differ by counts or proportions?
• Geospatial distribution?
• Can we get a stronger signal indicator by using a
  more specific code group?

22
Implications

• System design to include ILI surveillance needs
• Versus export data for SAS, SQL, etc.
• Ideally include additional graphical display
  options
• Histograms
• Line graphs (multiple age group or location
  comparisons)
• Proportions
• Other data sources (lab, absenteeism)

23
Acknowledgements

• Judy May, RN, BSN, MPH Washington State
  Department of Health
• Jo Hofmann, MD Formerly Washington State
  Department of Health
• Yevgeniy (Eugene) Elbert, MS Johns Hopkins
  University Applied Physics Laboratory

24
More Information

• Nigel Turner
• niturner_at_tpchd.org
• 253-798-6057
• Hilary Metcalf
• metcah_at_health.co.kitsap.wa.us
• 360-337-5258
• Atar Baer
• atar.baer_at_kingcounty.gov
• 206-263-8154
• Nicola Marsden-Haug
• Nicola.Marsden-Haug_at_doh.wa.gov
• 206-418-5558

25
Age Group Distribution of Positive Isolates,
2006-07 Season
Age Group Distribution of Syndromic Fever Visits,
2004-07
26
Correlation of Weekly ED Fever Visits with
Positive Influenza Isolates Regional Aggregate
by Age Group