Templates for analysis of individual-level prescription data

The advent of large population-based prescription databases has enabled us to study drug use with the individual user as our unit of analysis. This review presents three non-specific analytic templates that may be applied to individual-level prescription data. The ratio of prevalence odds to incidence rate can estimate the average duration for drug use. Limitations and pitfalls are discussed. Although it should be cautiously interpreted, it provides a reasonable ranking of drugs with respect to their retention in users. The Lorenz curve is an analytic tool to express skewness in drug use. It shows the proportion of drug use that is accounted for by percentiles of drug users, ranked according to their volume of drug intake. It may express the extent of heavy users as well as sporadic small-volume users and may, for example, be used to screen for an unsuspected abuse potential of a drug. The waiting-time distribution is a frequency distribution of first occurrences of drug use within a time-window. It forms the basis for a theoretical model for robust estimates of prevalence and incidence rate. On an intuitive level, it displays visual correlates of epidemiological prescribing parameters such as period prevalence, point prevalence, incidence rate, duration, prescription renewal rate, relapse of treatments and seasonality. Each measure may be incorporated into an integral matrix that reflects various traits in utilization of every drug or drug class, thereby possibly finding abnormalities that suggest sub-optimal prescribing.