Sample size and power considerations for ordinary least squares interrupted time series analysis: a simulation study

Area of Science:

• Epidemiology
• Biostatistics
• Health Services Research

Background:

• Interrupted time series (ITS) analysis is increasingly utilized in epidemiology.
• There is a lack of guidance on power and sample size for ITS studies.

Purpose of the Study:

• To assess statistical power for detecting intervention effects in various real-life ITS scenarios.
• To identify factors influencing power in ITS analysis.

Main Methods:

• Monte Carlo simulations were used to generate 1,000 ITS datasets per scenario.
• Variables included time points, sample size per time point, intervention effect size, and intervention timing.
• Simulations modeled both slope and step changes in the outcome.

Main Results:

• Sample size per time point significantly impacted statistical power.
• Even with 12 pre- and 12 post-intervention time points and moderate effect sizes, low sample size per time point led to underpowered analyses.
• Percentage bias was also evaluated.

Conclusions:

• Adequate power in ITS studies requires collective consideration of multiple factors.
• The study provides insight into sample size requirements for ordinary least squares (OLS) ITS analysis.
• Developed Stata code is available to estimate sample size needs.