Design and sample size considerations for valuation studies of multi-attribute utility instruments.

Area of Science:

• Health Economics
• Psychometrics
• Biostatistics

Background:

• The EQ-5D is crucial for health economic evaluations and reimbursement decisions.
• Current EQ-5D value sets rely on valuation studies of subsets of health states, leading to substantial prediction errors.
• Regression models are used to predict values for unvalued health states, contributing to inaccuracies.

Purpose of the Study:

• To derive a formula for the mean squared error (MSE) of EQ-5D value sets.
• To investigate the impact of valued health states, participant numbers, and correlation structures on MSE.
• To provide recommendations for optimizing EQ-5D valuation study designs.

Main Methods:

• Developed a formula for MSE in value set estimation using linear mixed models.
• Incorporated independent and Gaussian spatial correlation structures for model misspecification error.
• Analyzed the effect of varying the number of directly valued health states and participants.

Main Results:

• Valuing all 242 EQ-5D-3L health states substantially reduced MSE compared to valuing only 42 states.
• An independent correlation structure with 3773 participants yielded similar MSE to 600 participants with Gaussian spatial correlation.
• Fewer participants are needed to achieve lower MSE when all health states are valued.

Conclusions:

• Directly valuing a larger number of health states is more effective in reducing MSE than increasing participant numbers.
• Employing models with spatially correlated misspecification errors is recommended.
• Optimizing valuation study design by valuing more health states can improve the precision of EQ-5D value sets.