A Comprehensive Simulation Study to Evaluate the Effect Size and Study Length Relationship in Single-Group...

Area of Science:

• Biostatistics
• Health Services Research
• Epidemiology

Background:

• Single-group interrupted time-series analysis (ITSA) is widely used in healthcare research.
• Limited guidance exists for determining adequate sample size (number of time periods) for ITSA studies.
• Power requirements are critical for detecting statistically significant intervention effects.

Purpose of the Study:

• To estimate the number of time periods required for ITSA studies to achieve >80% and >90% statistical power.
• To assess the impact of autocorrelation, intervention timing, and effect size on power requirements.
• To provide practical tools for researchers to determine sample size for ITSA.

Main Methods:

• Simulations were conducted using varying numbers of time periods (10-100).
• Autocorrelation, intervention timing (33%, 50%, 67%), and effect sizes (level and trend changes) were manipulated.
• Statistical significance (p < 0.05, p < 0.01) and power levels were evaluated.

Main Results:

• Shorter studies and higher autocorrelation necessitate larger effect sizes for significance.
• Earlier or later intervention introduction (than 50%) requires larger effect sizes.
• Estimating level changes generally requires smaller effect sizes than trend changes, except with many time periods.
• Studies with only 10 time periods yielded unreliable estimates.

Conclusions:

• Healthcare researchers can use simulation-derived tables and the POWER_ITSA Stata package to plan ITSA studies.
• Adequate power depends on study length, autocorrelation, intervention timing, and effect size.
• Avoid studies with fewer than 10 time periods due to unreliable results.