A simulation-based sample size estimation for interrupted time-series studies with count outcome

Interrupted time series (ITS) analyses have been widely used to evaluate public health interventions. However, there is a lack of appropriate methods for sample size estimation in ITS studies with count outcomes, whether or not autocorrelation is present. We conducted a simulation-based study to comprehensively explore the impacts of intervention patterns, pre-intervention levels and trends, and intervention effect sizes on sample size determination. Moreover, we considered both observation-driven and parameter-driven autocorrelated cases, which were modelled using the quasi-maximum likelihood estimation and the maximum significant ρ correction methods, respectively. The simulation results demonstrated that all these parameters significantly influence sample size requirements, underscoring the necessity of their full consideration during the design stage. These simulation-based methods and the relevant R codes facilitate practical and flexible implementation for sample size estimation of ITS studies with count outcomes under varying autocorrelation structures.