Pharmacy data in epidemiological studies: an easy to obtain and reliable tool

Purpose The agreement between drug use measured in computerized pharmacy records and patient interviews or questionnaires is generally good. However, most investigations on this subject studied selected populations or subsets. We studied the coverage of Dutch pharmacy data for our study cohort, and the agreement between the different sources. Methods We used the data from 8592 subjects of an on-going population-based study, focused on the impact of microalbuminuria (PREVEND). Data on drug use was collected in a questionnaire and at community pharmacies. Drug use was measured in the year preceding the questionnaire. Agreement between the sources was measured using kappa-values, sensitivity and positive predictive value. Results Pharmacy data could be collected for 7568 (88%) of the study cohort. Pharmacy data and questionnaires showed good agreement for antihypertensives, lipid lowering drugs, oral antidiabetics and oral contraceptives, but poor agreement for nitrates, hormone replacement therapy and painkillers. Conclusions Pharmacy data could be collected for a large proportion of our cohort. For chronically used drugs pharmacy data generally agrees well with questionnaires. However, for drugs used for shorter periods, as needed, or also available over-the-counter, the agreement is not so good. Pharmacy data can be a valuable source of drug information in epidemiological studies. Copyright (C) 2002 John Wiley Sons, Ltd.