2008 HSR&D National Meeting Abstract
1062 — Using the Electronic Medical Record to Reduce Both the Delay and the Work-Load Required to Detect an Influenza Epidemic
DeLisle S (VA Maryland Health Care System), Ma Z
(Johns Hopkins Hospital), South B
(VA Salt Lake City Health Care System), Smith G
(Univ. of Pennsylvania), Loftus S
(VA Maryland HCS), Samore M
(VA Salt Lake City HCS), Perl TM
(Johns Hopkins Hospital)
Measures aimed at controlling epidemics of infectious diseases critically benefit from early outbreak recognition. This work uses a mathematical model of a plausible influenza epidemic to test how the statistical characteristics of single-case detection algorithms (CDAs) impact the real-world performance of a VA-based disease surveillance system.
A manual electronic medical record (EMR) review of 15,377 outpatient encounters at the Veterans Administration health system (VA) uncovered individuals with influenza-like illness (ILI). These cases served to develop CDAs that utilized various logical arrangements of structured EMR data and automated analyses of the free-text computerized clinical notes. We applied 18 of the most successful CDAs to EMR data extracts and generated 18 distinct real-world substrate datasets of daily ILI encounters at the Baltimore VA. We then injected an age-structured, spatiotemporal modeled epidemic into each of these background datasets, and simulated prospective outbreak detection with either scanning (SatScan) or CUSUM statistical approaches.
The most desirable CDAs were those that coupled single-case detection sensitivity above 75% with high specificity or positive predictive value. Compared to ICD-9-only CDAs in use by current national surveillance systems, the best EMR-based CDAs shortened outbreak detection delay by 3 days and decreased the false-alert rate ~15-fold.
Case detection methods that take advantage of information from the full EMR can lower both the delay and the workload required to find an influenza epidemic in the community.
Automated, high-performance EMR-based recognition of diseases or events requires methodical development, but offers novel and powerful avenues toward system-wide improvements in health care quality and safety.