Background: Interrupted Time Series (ITS) studies are commonly used to evaluate public health and policy interventions when randomisation is impractical or infeasible; for example, examining the effects of mass media campaigns on the use of methamphetamine among young adults. In an ITS study, measurements on a group of individuals (e.g., community) are taken repeatedly both before and after the intervention. The key benefit of the ITS design is that any secular trend in the period before the intervention can be accounted for when estimating the impact of the intervention. Several effect measures can be used to characterise both short- and long-term effects of the intervention (e.g., immediate level-change and long-term level-change). Meta-analysis of these effect estimates can usefully inform decision-making.
Objectives: In this workshop, we aim to equip review authors with the knowledge and tools to incorporate ITS in their reviews by: i) demonstrating how to digitally extract data from ITS graphs; ii) how to analyse ITS studies and meta-analyse their results; and iii) highlight design features to consider when assessing the risk of bias. This workshop will require access and basic competency in Stata or R and will assume knowledge of meta-analysis.
Description: We will use a combination of presentations and computer practicals. For the analysis of the ITS studies, we will focus on fitting segmented linear regression models and demonstrate how to set up the data for analysis and undertake the analysis. We will discuss the complexities that arise when analysing time series data (e.g., autocorrelation). We will then demonstrate how to meta-analyse the resulting effect estimates. Finally, via example, we will use the ROBINS-I framework to discuss features of ITS designs that may bias effect estimates.