Sample size requirements to evaluate policies in addiction research using interrupted time series analysis (ITS): Tools and guidance.

Abstract

Formal power calculations are rarely presented in interrupted time-series (ITS) studies due to their technical complexity, creating a significant gap in methodological rigor. This paper aimed to make power and sample size determination more accessible for researchers, particularly in the field of addiction, by providing a suite of practical and user-friendly tools. A set of resources was developed using Monte Carlo simulation to allow researchers to estimate statistical power under a wide range of ITS design parameters. The approach allows for the explicit definition of the data-generating process, including specific autocorrelation error structures (ARMA), the presence of covariates and trends and different intervention effect types (step, pulse and trend change). The study produced three key resources: (1) a flexible R code base for conducting custom power simulations, (2) an intuitive, interactive R Shiny App that enables code-free power analysis through a web interface and (3) a series of pre-calculated look-up tables for quick sample size estimation during the initial stages of study design. Illustrative examples from addiction research demonstrate the tools' application. The provided tools bridge a critical gap by simplifying the process of conducting rigorous power calculations for ITS designs. Their adoption can enhance the planning, execution and interpretation of quasi-experimental studies, helping to ensure that research is adequately powered to detect meaningful policy and intervention effects.