Comparison of different methods to estimate prevalence of drug use by using pharmacy records

Abstract

Several methods to estimate prevalence of drug use are available, which may complicate a valid comparison of these estimates. Standardization may contribute to more valid comparisons. We compared different methods to estimate prevalence of drug use by using pharmacy records. Data were obtained from the Dutch population-based PHARMO-database comprising medication histories of 300,000 subjects. Five point prevalences and a 1-year prevalence of cholesterol-lowering drug use were estimated in 1995. Four point prevalences differed in data handling before estimating prevalence (e.g., correction for irregular drug use or construction of episodes of drug use). The numerator of the fifth point prevalence estimate represented the number of defined daily doses (DDDs) instead of the number of patients filling a prescription. The first four point prevalences ranged from 11.0-12.1 per thousand. Prevalence ratio (male:female) was 1.2 for these methods. The fifth method resulted in an estimate similar to the other point prevalences (11.9 DDDs/1000 inhabitants). However, the prevalence ratio was 1.4 due to larger average number of DDDs prescribed to men. One year-prevalence was 4-5 per thousand higher than point prevalences. The comparison of these methods indicated that the choice of prevalence measure (point versus period prevalence) substantially influenced the prevalence estimate, whereas the influence of data handling was negligible. For standardization purposes in drug utilization research, we recommend estimating point prevalence instead of period prevalence. The various methods of data handling before estimating point prevalence yielded similar results and therefore we cannot recommend one specific method. However, defined daily doses should not be used to estimate (point) prevalences of drug use because this measure is significantly influenced by prescribed dosage regimens.