Design, analysis, power, and sample size calculation for three‐phase interrupted time series analysis in evaluation of health policy interventions

Summary

This methodological paper presents a simulation-based framework and ready-to-use computer programs for determining appropriate sample sizes and statistical power in three-phase interrupted time series studies evaluating health policy and environmental interventions. The authors illustrate the design and analysis approach using a National Institutes of Health-funded exemplar, demonstrating how power varies with sample size, effect size, autocorrelation, and the type of intervention effect being tested (level change, trend change, or both). The work provides investigators with practical tools to ensure sufficient statistical power when implementing this increasingly common quasi-experimental study design.

UK applicability

The methodological framework and computational tools presented are broadly applicable to UK health policy evaluation contexts, particularly for assessing the impact of National Health Service reforms, public health interventions, or environmental health policies. However, the abstract does not address UK-specific regulatory or health system contexts, limiting direct guidance for UK implementation.

Key measures

Statistical power; sample size; segmented autoregressive error models; autocorrelation coefficients (−0.9 to 0.9); effect sizes; testing level changes and trend changes

Outcomes reported

The study developed and demonstrated simulation-based approaches for calculating statistical power and sample size in three-phase interrupted time series (ITS) studies evaluating health policy, systems, or environmental interventions. Power estimates were generated across varying levels of autocorrelation, effect sizes, and testing conditions (level change, trend change, or both).

Topic tags