OBJECTIVES: Controlling pharmaceutical expenditures is of particular interest to governments, as pharmaceuticals present one of the main components of health care expenditures. Patent expiries of drugs are important, as dispensing (cheaper) generic drugs potentially results in lower pharmaceutical expenditures. Therefore, modelling generic substitution patterns is highly relevant as this can provide useful cost-cutting decision support. The aim of this study was to model the duration until patients switch from branded to generic drugs, in relation to various influencing variables METHODS: Data were obtained from Dutch pharmacy dispensing records from IADB.nl. We focused on antidepressant prescription data. To identify a pattern on the underlying diffusion process, the analysis was applied to four antidepressants whose patent recently expired (fluoxetine, paroxetine, citalopram, sertraline). Duration analysis techniques were used to estimate the probability of patients to switch to a generic drug over time, and to estimate the effect of different covariates (e.g. patient, pharmacist and general practitioner characteristics) on this switching probability. Since interval censored data were used, discrete duration methods were applied which resulted in the estimation of a binary regression model. RESULTS: A higher probability for patients to switch from a branded to a generic drug within the first five months after patent expiry was identified. Switching probabilities were mainly affected by the general practitioners' and pharmacists' inclination to provide generic drugs, the age, the experience of the patient on the specific drug and the amount of different generic drugs introduced in the market. CONCLUSIONS: Despite differences in pharmacy dispensing patterns for different antidepressants studied, we generally found common patterns in generic substitution. Next to the inclination of health care professionals, patient characteristics, time and the amount of available generic alternatives affect generic substitution significantly. However, future research in drug classes other than antidepressants should be conducted to verify our findings.