Analysis of the short form-36 (SF-36): the beta-binomial distribution approach

Area of Science:

• Health outcomes research
• Statistical modeling
• Psychometrics

Background:

• Health-related quality of life (HRQoL) is a key health indicator.
• The Short Form-36 (SF-36) is a widely used generic HRQoL measure.
• Traditional analysis using multiple linear regression (MLR) may be inadequate due to non-normal, skewed, and bounded SF-36 data.

Purpose of the Study:

• To assess the performance of MLR under non-normality conditions for SF-36 scores.
• To propose and evaluate beta-binomial regression (BBR) as a suitable alternative for analyzing SF-36 data.
• To compare the clinical meaningfulness and performance of BBR versus MLR.

Main Methods:

• Analysis of SF-36 data from 193 patients with eating disorders.
• Testing normality assumptions for SF-36 domains, revealing frequent rejections.
• Fitting SF-36 scores using beta-binomial distribution and comparing MLR with BBR on real and simulated data.

Main Results:

• Normal distribution was rejected for most SF-36 domains.
• Beta-binomial distribution provided a good fit for several SF-36 domains.
• BBR demonstrated superior performance and more clinically meaningful interpretations than MLR, particularly for domains with limited categories.

Conclusions:

• Beta-binomial regression (BBR) is recommended for analyzing SF-36 scores, especially when data violate normality assumptions.
• BBR enhances the detection and interpretation of predictors of health-related quality of life.
• Further research is needed to validate BBR for other HRQoL measures and populations.