OBJECTIVES: Comparison of annual medication costs in a population-based study using a prediction formula based on 7 day medication history to cost data provided by health insurance companies. METHODS: The KORA Asthma and Allergy study evaluated cost of illness due to asthma and allergies in a population-based case-control design. Medication costs originated from a 7 day medication history (interview) and from health insurance data. Drugs documented per interview were assigned an average price per defined daily dose (DDD) for each standard package size group. Weekly medication costs were extrapolated by multiplying price per DDD (medium package size) by predicted length of intake according to general ATC group. For consenting subjects, all medications reimbursed by the health insurance companies for 1998 were obtained. The annual total costs as well as cost differences between disease groups were compared between both approaches. RESULTS: Of 1534 subjects participating in the KORA study, 1249 were insured publicly and 63.8% of those consented to release their health insurance data. Of 614 persons with prescribed medications according to insurance data, 233 (38%) reported no prescribed medications during the interview. Median (inter-quartile range) annual costs for this group were 37€ (16-103€). For the other 381 subjects (62%), annual insurance costs were 260€ (116-638€) whereas predicted costs were higher (364€; 104-863€). For subjects with asthma or allergy, predicted costs agreed better with annual costs. Costs extrapolated from interview data correlated significantly (r=0.63) with the annual costs. CONCLUSION: Estimation of annual medication costs for chronic disease patients from seven day medication history data is feasible and estimates of group averages are similar to full annual data from health insurance companies. For population-based samples the latter approach is logistically more difficult, is less accepted by subjects, and does not encompass costs of most of OTC drugs.