Abstract
AbstractInterrupted time series design is a quasi-experimental study design commonly used to evaluate the impact of a particular intervention (e.g., a health policy implementation) on a specific outcome. Two of the most often recommended analytical approaches to interrupted time series analysis are autoregressive integrated moving average (ARIMA) and Generalized Additive Models (GAM). We conducted simulation tests to determine the performance differences between ARIMA and GAM methodology across different policy effect sizes, with or without seasonality, and with or without misspecification of policy variables. We found that ARIMA exhibited more consistent results under certain conditions, such as with different policy effect sizes, with or without seasonality, while GAM were more robust when the model was misspecified. Given these findings, the variation between the models underscores the need for careful model selection and validation in health policy studies.
Publisher
Cold Spring Harbor Laboratory
Reference20 articles.
1. Understanding and using time series analyses in addiction research
2. Cook, T. D. , Campbell, D. T. , & Shadish, W. (2002). Experimental and quasi-experimental designs for generalized causal inference (Vol. 1195). Houghton Mifflin.
3. The Excess Winter Deaths Measure
4. The temporal relationship between per capita alcohol consumption and harm: A systematic review of time lag specifications in aggregate time series analyses
5. Methodology and reporting characteristics of studies using interrupted time series design in healthcare;BMC medical research methodology,2019