Overview of Syndromic Surveillance presented as background to
Overview of ‘Syndromic Surveillance’ presented as background to Multiple Data Source Issue for DIMACS Working Group on Adverse Event/Disease Reporting, Surveillance, and Analysis II Henry R. Rolka, R. N. , M. P. S. , M. S. Centers for Disease Control and Prevention February 19, 2004
New data types and functional objectives have largely expanded the scope of public health surveillance
New surveillance challenges and opportunities are growing in complexity
Outline of Presentation • Background and context for appreciation of new complexities. • Major themes and issues. • Focus for this meeting • Summary and discussion.
Public Health Surveillance “Ongoing systematic collection, analysis, and interpretation of outcome-specific data for use in the planning, implementation, and evaluation of public health practice. ” *Stephen Thacker, CDC
Surveillance System Data Collection Analysis Dissemination
Surveillance System Components Population of interest which generates events Measurement and recording Transactional data Data Management • Quality checks • Editing Public health response Interpretation for associations, trends, unusual patterns, signals Analytical applications Data preprocessing for a specific purpose (‘views’, ‘data marts’)
Conceptual Taxonomy Public Health Surveillance Medical Utilization and Adverse Events Drug Vaccine Other Products/Services Disease Traditional Infectious Disease Birth defect ‘Syndromic’ Other Injuries Etc.
NETSS • Weekly data regarding cases of nationally notifiable diseases. • Core surveillance data: date, county, age, sex, and race/ethnicity. • Some disease-specific epidemiological information. • Transmitted electronically by the states and territories to CDC each week.
Figure 1 published weekly in the MMWR
Syndromic Surveillance “Monitoring frequency of illnesses with a specified set of clinical features in a given population, without regard to the diagnoses. ” Arthur Reingold, UC Berkeley
Surveillance System Components Data collection and preprocessing A Epidemiological decisions Data View Reporting or recording anomaly Application of statistical algorithms ‘Something unusual’ noted in data Data processing error Statistical aberration due to natural variability etc. B True increase in disease Requires information from other data sources Naturally occurring outbreak Deliberate exposure event C
Non-traditional Data Types for Public Health Surveillance • Pre-diagnostic/chief complaint (text data) • Over-the-counter sales transactions – Drug store – Grocery store • • • 911 -emergency calls Ambulance dispatch data Absenteeism data ED discharge summaries Managed care patient encounter data Prescription/pharmaceuticals
Potential Syndromic Surveillance Data Sources • • • Day 1 - feels fine Day 2 - headaches, Pharmaceutical Sales Day 3 - develops cough, Nurse’s Hotline Day 4 – Managed Care Org Absenteeism Day 5 – Worsens, Ambulance Dispatch (EMS) ED Logs • Day 6 • Day 7 • Day 8 - Traditional Surveillance *Farzad Mostashari, NYC Do. H
Messy Data • • • Noisy, periodic (weekly, seasonally) Multiple data streams Duplicate records Syndromic coding not standardized Data quality Means for evaluation not well developed
Bio-ALIRT • “Bio-Event Advanced Leading Indicator Recognition Technology” • Program to develop technology for early detection of a covert biological attack • Defense Advanced Research Projects Agency (DARPA) • Began in fy 2001
Biosurveillance Data Space LATER DETECTION EARLY DETECTION INTELLIGENCE ANIMALS Vets Zoos Agribusiness Environmental BIOSENSORS HUMAN BEHAVIORS OTC Pharm Absenteeism Utilities Coughs NON TRADITIONAL USES Tests ordered Poison Centers Influenza isolates 911 Calls Humidity Traffic EMS Runs Temperature Survey Public Transport Cafeteria Nurse Calls Pollution Video Surv Newsgroup Test Results Diagnosis Web Queries Allergy Index CLINICAL DATA Complaints Pollen counts Wind Speed/ direct. GOLD MEDICAL STANDARDS ER Visits Prescriptions Radiograph Reports Medical Examiner Test Results Sentinel MD Investigations Limited Utility Some Potential Promising
Bio. Sense (under development) • Complementary project to President’s initiatives Bio. Watch and Bio. Shield. • Focuses on disease symptoms related to syndromic categories (BT agents) • Data source examples: – Patient encounter (ICD 9, outpatient) – OTC sales of home health remedies – Lab tests ordered – Nurse call line
Common Interests/Challenges DARPA – Bio. Alirt • • • Surveillance for BT Non-traditional data Early detection Evaluation of algorithms Privacy protection CDC – Bio. Sense • • • Surveillance for BT Non-traditional data Early detection Evaluation of algorithms Privacy protection
Themes (system) • Local vs. Regional vs. National vs. Global focus • Interoperability / Transportability • Interdisciplinary science and technologies – Culturalism – Language – Social networks • Case/Adverse Event definitions • Information/knowledge management • Leadership
Themes (functionality) • • Timeliness for response potential Data quality factors System evaluation Data access Standards Signal detection thresholds Analytic methodologies
Analytic Obstacles/Opportunities • • • ‘Opportunistic’ data ‘Syndromes’ Empirical inductive inference Evaluation of utility and public health value Multiple data streams in time – – – Multivariate time series ( uncharacterized transfer functions) Time alignment Differential quality
- Slides: 22