Agreement between self-reported data on medicine use and prescription records vary according to method of analysis and therapeutic group

OBJECTIVE: This study compared national self-reported data on medicine use and national prescription records at the individual level. STUDY DESIGN AND SETTING: Data from the nationally representative Danish health survey conducted in 2000 (n=16,688) were linked at the individual level to national prescription records covering 1999-2000. Kappa statistics and 95% confidence intervals were calculated. RESULTS: Applying the legend time method to medicine groups used mainly on a chronic basis revealed good to very good agreement between the two data sources, whereas medicines used as needed showed fair to moderate agreement. When a fixed-time window was applied for analysis, agreement was unchanged for medicines used mainly on a chronic basis, whereas agreement increased somewhat compared to the legend time method when analyzing medicines used as needed. CONCLUSION: Agreement between national self-reported data and national prescription records differed according to method of analysis and therapeutic group. A fixed-time window is an appropriate method of analysis for most therapeutic groups.