The parametric waiting time distribution model and its application in pharmacoepidemiologic studies

Henrik Støvring, Section for Biostatistics, Aarhus University

Information on prescription redemptions are routinely registered in populationwide electronic databases. In traditional pharmacoepidemiologic research the duration of individual prescriptions is defined based on decision rules with clinical input. We have recently proposed a parametric method to estimate prescription durations via the waiting time distribution as an alternative. I will show how covariates can be introduced in the model to improve precision. Finally, the model allows estimation of the probability of a patient being treated on a given day and I will show how this probability can be used as exposure measure to estimate the being treated.