Context: Public health surveillance systems for acute hepatitis are limited: clinician reporting is insensitive and electronic laboratory reporting is nonspecific. Insurance claims and electronic health records are potential alternative sources.
Objective: To compare the utility of laboratory data, diagnosis codes, and electronic health record combination data (current and prior viral hepatitis studies, liver function tests, and diagnosis codes) for acute hepatitis A and B surveillance.
Design: Retrospective chart review.
Setting: Massachusetts ambulatory practice serving 350 000 patients per year.
Participants: All patients seen between 1990 and 2008.
Main outcome measures: Sensitivity and positive predictive value of immunoglobulin M (IgM), International Classification of Disease-Ninth Revision (ICD-9) diagnosis codes, and combination electronic health record data for acute hepatitis A and B.
Results: During the study period, there were 111 patients with positive hepatitis A IgMs, 154 with acute hepatitis A ICD-9 codes, and 77 with positive IgM and elevated liver function tests. On review, 79 cases were confirmed. Sensitivity and positive predictive value were 100% and 71% (95% confidence interval, 62%-79%) for IgM, 94% (92%-100%) and 48% (40%-56%) for ICD-9 codes and 97% (92%-100%) and 100% (96%-100%) for combination electronic health record data. There were 14 patients with positive hepatitis B core IgMs, 2564 with acute hepatitis B ICD-9 codes, and 125 with suggestive combinations of electronic health record data. Acute hepatitis B was confirmed in 122 patients. Sensitivity and positive predictive value were 9.4% (5.2%-16%) and 86% (60%-98%) for hepatitis B core IgM, 73% (65%-80%) and 3.6% (2.9%-4.4%) for ICD-9 codes, and 96% (91%-99%) and 98% (94%-99%) for electronic health record data.
Conclusions: Laboratory surveillance using IgM tests overestimates the burden of acute hepatitis A and underestimates the burden of acute hepatitis B. Claims data are subject to many false positives. Electronic health record data are both sensitive and predictive. Electronic health record-based surveillance systems merit development.