Deciphering data anomalies in BioSense

Area of Science:

• Public Health Surveillance
• Epidemiology
• Health Informatics

Background:

• The Centers for Disease Control and Prevention (CDC) has utilized the BioSense surveillance application since June 2004 for daily nationwide syndromic data monitoring.
• A dedicated team of CDC analysts oversees the BioSense application, playing a crucial role in identifying and interpreting data anomalies.

Purpose of the Study:

• To examine the role of CDC analysts in identifying and deciphering data anomalies within the BioSense surveillance application.
• To discuss the limitations of the current surveillance system, lessons learned, and propose future improvements for national syndromic surveillance methodology.

Main Methods:

• Data integration from diverse sources including clinical diagnoses (ICD-9-CM), medical procedures (CPT codes), over-the-counter product sales, and laboratory tests.
• Filtering of all data to exclude irrelevant information for syndromic surveillance purposes.
• Utilized statistical algorithms and analytical visualization features for anomaly detection.

Main Results:

• During June-November 2004, approximately 160 data anomalies were examined, with no disease outbreaks or deliberate pathogen exposures detected.
• Anomalies primarily involved unusual changes in daily data volume or types of clinical diagnoses and procedures.
• Routine monitoring steps include anomaly detection, scope estimation, supplemental fact-gathering, multi-source data comparison, hypothesis development, and validation, requiring rapid completion for early detection.

Conclusions:

• The BioSense system necessitates an empirical learning process to optimize the use of public health data.
• Enhancing the user interface and incorporating input from local public health partners can improve the effectiveness of syndromic surveillance.