Bayesian statistical models to estimate EQ-5D utility scores from EORTC QLQ data in myeloma

Area of Science:

• Health Economics
• Biostatistics
• Clinical Trials

Background:

• Health-related quality of life (HRQoL) data modeling is challenging due to data skewness and zero-inflation.
• Accurate estimation of utility scores is crucial for economic evaluations in healthcare.

Purpose of the Study:

• To develop and compare various two-part models (TPMs) for modeling health-related quality of life data.
• To assess the performance of different TPMs in estimating European Quality of Life-5 Dimensions (EQ-5D) utility scores.

Main Methods:

• Utilized data from the UK Medical Research Council Myeloma IX trial.
• Developed four TPMs: normal regression, variance-function normal regression, log-transformed data, and gamma regression with log link.
• Validated models using a hold-out dataset, comparing predicted vs. observed EQ-5D scores, R², and root mean square error.

Main Results:

• Model performance varied in the derivation set based on evaluation criteria.
• TPM with normal regression favored R²/adjusted-R².
• TPM with gamma regression showed superior predictive performance across all criteria in the validation dataset.

Conclusions:

• TPM regression models offer flexible approaches for estimating mean EQ-5D utility weights.
• The TPM with gamma regression is recommended for its robust predictive performance in this context.
• These models facilitate accurate economic evaluations using patient-reported outcome data.