Sample size determination for health psychology interventions with binomially distributed outcomes

Area of Science:

• Biostatistics
• Health Intervention Design
• Behavioral Science

Background:

• Health intervention outcomes are frequently evaluated using binomial data.
• Accurate sample size calculations necessitate modeling the pre-intervention behavior rate.
• Most sample size software neglects pre-intervention rates, leading to inflated estimates.

Purpose of the Study:

• To introduce and advocate for the conditional binomial method for sample size determination in health interventions.
• To highlight the advantages of the conditional binomial method over conventional approaches.
• To present user-friendly software implementing this method.

Main Methods:

• The study focuses on the conditional binomial method, which adjusts post-intervention outcomes based on pre-intervention behavior rates (baserates).
• This method contrasts with conventional approaches that only consider post-intervention outcomes.
• User-friendly software for implementing the conditional binomial method is presented.

Main Results:

• The conditional binomial method explicitly models pre-intervention behavior.
• This approach consistently yields smaller sample size estimates compared to traditional methods.
• The presented software facilitates the application of this more accurate method.

Conclusions:

• The conditional binomial method offers a more accurate and efficient approach to sample size calculation for health interventions with binomial outcomes.
• Utilizing baserate-adjusted outcomes leads to more precise and often smaller sample size requirements.
• The availability of user-friendly software promotes wider adoption of this improved methodology.