The waiting time distribution as a graphical approach to epidemiologic measures of drug utilization

The emergence of large, computerized pharmacoepidemiologic databases has enabled us to study drug utilization with the individual user as the statistical unit. A recurrent problem in such analyses, however, is the overwhelming volume and complexity of data. We here describe a graphical approach that effectively conveys some essential utilization parameters for a drug. The waiting time distribution for a group of drug users is a charting of their first prescription presentations within a specified time window. For a drug used for chronic treatment, most current users will be captured at the beginning of the window. After a few months, the graph will be dominated by new, incident users. As examples, we present waiting time distributions for insulin, ulcer drugs, systemic corticosteroids, antidepressants, and disulfiram. Appropriately analyzed and interpreted, the waiting time distributions can provide information about the period prevalence, point prevalence, incidence, duration of use, seasonality, and rate of prescription renewal or relapse for specific drugs. Each of these parameters has a visual correlate. The waiting time distributions may be an informative supplement to conventional drug utilization statistics, and possibly also a useful screening tool for unusual prescribing patterns.