Use of insurance claims in epidemiologic research: Identification of peptic ulcers, gi bleeding, pancreatitis, hepatitis and renal disease

Abstract

Data bases that are created to administer health care benefits and to record financial aspects of medical care in the United States are useful for epidemiologic research in certain circumstances. Case finding and verification in such records is complex and rarely described in detail in research reports because of space limitations. This report describes portmarketing surveillance for peptic ulcer, upper gastrointestinal hemorrhage and hepatic dysfunction in 68, 180 people who took non‐steroidal anti‐inflammatory drugs (NSAIDs), by using an automated data base together with review of selected medical records. Cases included both mild diseases treated on an outpatient basis and more severe illnesses that required hospital admission. Using computer programs we identified 9642 subjects who had at least one insurance claim suggesting a possible case event. We manually reviewed chronological profiles of their insurance claims, to select 833 subjects for medical record reviews, of whom 152 were verified as cases. Our algorithms for case identification were sensitive, but not very specific. If we had restricted our manual review to people who had at least five relevant claims during the study period, we would have avoided 42% of the reviewing workload and lost only 5% of the cases.