Background: Pharmacoepidemiological studies assessing the associations between psychotropic drug use and adverse events in the elderly frequently employ automated pharmacy databases as the source of exposure data. However, information on the validity of these databases for estimating psychotropic drug exposures in elderly people is scarce.
Objective: This study evaluated the validity of the Finnish Prescription Register for estimating current exposures to psychotropic drugs in elderly people. Furthermore, the potential change in the validity over time was determined.
Methods: This was a population-based intervention study (GeMS; Geriatric Multidisciplinary Strategy for the Good Care of the Elderly) conducted between 2004 and 2007. Initially, 1000 randomly selected persons aged >or=75 years living in the City of Kuopio, Finland, in November 2003 were invited to participate in the study. Of these, 716 agreed to participate at baseline (2004) and 570 were still available for 3-year follow-up (2007). The validity of the Prescription Register was assessed by comparing it with the self-reported information collected by interviews in 2004 and in 2007 in the GeMS study. Using the self-reported data as a reference standard, sensitivity, specificity and Cohen's kappa statistic (measure of inter-rater agreement for qualitative [categorical] items) with 95% confidence intervals were computed for different categories and subcategories of psychotropic drugs, applying fixed-time windows of 4, 6 and 12 months.
Results: In 2007, the sensitivity varied between psychotropic categories and subcategories, being generally highest with the 12-month time window (0.57-0.96). The specificity was highest with the 4-month time window (0.94-0.99), showing a slight tendency to decrease with an extended time window. The sensitivity and specificity were highest for antidepressants and antipsychotics, followed by benzodiazepines. The agreement was almost perfect (kappa = 0.81-1.00) or substantial (kappa = 0.61-0.80) for all categories and subcategories of psychotropic drugs. Few differences in validity were observed between the two years.
Conclusion: Using self-reported data as a reference standard, the Prescription Register provides valid information on current exposures to antidepressants and antipsychotics in elderly people if the time window is selected with adequate consideration. However, the validity is lower for benzodiazepines, suggesting that other sources of information should be considered when performing pharmacoepidemiological studies.