Colleen Bradley, MSPH Syndromic Surveillance Conference, 2005

1
Colleen Bradley, MSPH Syndromic Surveillance
Conference, 2005
Visualizing and Monitoring Data in BioSense Using
SaTScan
2
Objective

• To describe visualization and monitoring of
  national spatio-temporal analysis results in the
  BioSense application

3
Outline

• BioSense system
• SaTScan implementation
• Application visualizations
• Experiences
• Planned enhancements
• Conclusions

4
BioSense System

• Provides early event detection and situational
  awareness capabilities
• Bioterrorism surveillance
• Public health event management
• Collects, analyzes, and visualizes near real-time
  health data
• State and local health department access to
  jurisdictional data views
• National monitoring in CDC BioIntelligence Center
  (BIC)

5
BioIntelligence Center

• Initiated June 2004
• Daily national data monitoring and tracking
• State and local monitoring support
• Data anomaly investigations
• System troubleshooting and enhancement

6
BioSense Data

• Department of Defense and Veterans Affairs
  ambulatory care diagnoses and procedures
• Laboratory Corporation of America lab test orders
• Data mapped to 11 syndrome categories
• Zip code is geographical unit of analysis

7
SaTScan

• Spatio-temporal statistical technique
• Developed by Martin Kulldorff
• To identify space-time event clusters
• To perform repeated time-periodic surveillance
  for the early detection of public health events

8
Pre-Identified Issues

• Clusters
• Cross-jurisdictional
• Large
• Numerous
• Epidemiologic and/or clinical importance
• Zip codes have highly variant populations and
  data volume
• Data lag time issue

9
Implementation

• Made available in April 2005 in the BioSense
  application
• National view for use at BIC
• Jurisdictional views for eventual state/local PH
  use
• Monitored initially at CDCs BioIntelligence
  Center (BIC)

10
SaTScan Applied in BioSense

• Algorithm applied daily to each source and
  syndrome
• Results displayed in jurisdictionally based views
  (states and metropolitan areas)
• National view available for BIC
• Nation divided into distinct geographical units
  (the grid) based on population density
• Small Area Regression and Testing (SMART) used to
  generate expected grid unit counts

11
SaTScan Parameters

• Maximums
• Number of clusters for each source and syndrome
  top 10 in nation
• Days in cluster 7
• Size of cluster 5 of US population
• Current date minus 3 days set as cluster end date

12
Visualization BioSense Home Page

• Table indicates number of SaTScan clusters by
  data source and syndrome
• Clusters highlighted if jurisdiction contributing
  data records
• Part of a jurisdiction without data records can
  be included in a cluster

13
Results Syndrome-Specific Cluster Page

• Select syndrome cluster from home page
• Navigate to syndrome-specific cluster page
• Observe jurisdictional map
• Displays clusters for that syndrome
• Indicates if a cluster extends beyond the
  jurisdiction
• Examine data source cluster table
• Lists summary cluster information

14
Visualization Syndrome-Specific Cluster
Page(Demonstration Data)
Data Source Clusters
Jurisdictional Map
Data Source Cluster Table
Summary Cluster Information
15
Summary Cluster Information

• Cluster ID
• Rank
• Central location
• Radius
• Population
• Zip codes
• Zip codes reporting in past 30 days

• Observed
• Expected
• Observed/Expected
• Recurrence interval threshold
• Start date
• End date

16
Visualization Daily Scale Table(Demonstration
Data)

• Select cluster ID to observe daily scale table
• Ratio of O/E for each day in the cluster
• Color indicates scale value
• Scale(1count)/(1expected)

17
Visualization Daily Record Count
table(Demonstration Data)

• Select one or all grid IDs to access daily record
  count table
• Record counts for patients outside the
  jurisdiction not displayed
• Option to hide zip codes without data

18
Visualization Detailed Line Listing(Demonstratio
n Data)

• Select one or all record counts to access
  detailed line listing

19
Data Monitor Issues

• Number of clusters
• Balance monitoring time with potential cluster
  importance
• Epidemiologic analysis of cluster records
• Analyze trends within cluster
• Determining clinical and/or epidemiologic
  relevance can be difficult

20
BIC Monitor Experience

• Monitors need to
• View the detailed line list for all records in
  the cluster for epidemiologic patterns
• Graph historical data for zip codes in the
  cluster for longer term temporal analysis
• Distinguish zip codes that are included in the
  cluster geographical area, but do not contribute
  record counts
• Communicate with state and local PH officials

21
Planned Developments

• Address large cluster size
• Improve data visualizations
• Increase customizability
• Enhance end user analysis capabilities
• Provide SaTScan results in jurisdictional
  visualizations for state and local public health
  monitoring
• Solicit state and local user feedback

22
Conclusions

• SaTScan implementation in BioSense involved
  unique issues due to national scope
• BIC monitor feedback was necessary in application
  development
• Enhancements will continue to improve capability
  to monitor SaTScan results in BioSense
• Communication with and feedback from state and
  local PH essential going forward

23
References

• SaTScan. Accessed at http//www.satscan.org/.
• Syndrome Definitions for Diseases Associated with
  Critical Bioterrorism-Associated Agents. Accessed
  at http//www.bt.cdc.gov/surveillance/syndromedef
  /index.asp.
• Kulldorff M. Prospective time periodic
  geographical disease surveillance using a scan
  statistic. J R Statis Soc A 200116461-72.
• Mostashari F, Kulldorff M, Hartman J, Miller J,
  and Kulasekera V. Dead Bird Clusters as an Early
  Warning System for West Nile Virus Activity.
  Emerg Infect Dis 20039(6)641-646.
• Kleinman K, Lazarus R, Platt R. A generalized
  linear mixed models approach for detecting
  incident clusters of disease in small areas, with
  an application to biological terrorism. Am J
  Epidemiol 2004159217-24.

24
Acknowledgements

• Leslie Sokolow
• Shirley Willson
• Yukun Jin
• Martin Fayomi
• Matthew Miller
• Duane Zomer

• Martin Kulldorff, Harvard
• John Loonsk, CDC
• Henry Rolka, CDC
• David Walker, CDC
• Kyumin Shim, CDC
• Steve Bloom, SAIC
• Liegu Hu, SAIC
• Nancy Grady, SAIC

BioIntelligence Center Data Monitors
25
Questions?

• Thank you!
• CBradley1_at_CDC.Gov
• 404-498-6312