Measuring safety treatment effects using full Bayes non-linear safety performance intervention functions

► We propose a first-order autoregressive non-linear intervention model for safety evaluation. ► The model makes use of the Poisson-lognormal hierarchy and incorporates random parameters. ► Simple expressions are given to compute novelty, direct and indirect treatment effects. ► The non-linear model is shown to provide a wider variety of treatment profiles than the linear model. ► The non-linear dynamic model is also shown to fit the data better than the linear intervention model.