Application of negative binomial modeling for discrete outcomes: a case study in aging research

Area of Science:

• Gerontology
• Biostatistics
• Clinical Trials

Background:

• Prehabilitation programs aim to improve health outcomes for frail older adults.
• Functional decline and disability are significant concerns in community-living elderly populations.
• Analyzing discrete outcome data with overdispersion presents statistical challenges.

Purpose of the Study:

• To evaluate the effectiveness of a prehabilitation program using a negative binomial regression model.
• To address overdispersion in discrete disability outcome data from a clinical trial.
• To compare the negative binomial model with the Poisson regression model for this data.

Main Methods:

• Case study utilizing negative binomial regression for discrete outcome data.
• Analysis of disability measures at 7 months from a prehabilitation clinical trial.
• Comparison of model fit and parameter efficiency between negative binomial and Poisson models.

Main Results:

• The disability data exhibited right skewness and clumping at zero (40% had no disability).
• Variance (16.4) was nearly six times greater than the mean (2.8), indicating significant overdispersion.
• The negative binomial model provided a superior fit to the data compared to the Poisson model.

Conclusions:

• The negative binomial regression model is a suitable alternative for analyzing discrete data with overdispersion.
• This model offers improved efficiency for regression parameter estimates in such cases.
• The findings support the use of negative binomial regression for clinical trial data with excess variance.