Response Patterns in Health State Valuation Using Endogenous Attribute Attendance and Latent Class Analysis

Area of Science:

• Health Economics
• Decision Science
• Psychometrics

Background:

• Discrete choice experiments (DCEs) are crucial for health state valuation.
• Simplifying heuristics, like attribute non-attendance, can bias DCE models.
• Attribute non-attendance means respondents focus on a subset of attributes.

Purpose of the Study:

• To explore attribute non-attendance in health state valuation DCEs.
• To compare latent class (LC) and endogenous attribute attendance (EAA) models for health DCEs.
• To assess the impact of non-attendance on quality-adjusted life year (QALY) weights.

Main Methods:

• Applied LC and EAA models to health state valuation DCE data.
• Adjusted LC and EAA models for the QALY framework.
• Compared model fit and estimated QALY weights under different non-attendance assumptions.

Main Results:

• Explicitly modeling attribute non-attendance significantly improves model fit.
• The impact on QALY weights depends on the source of non-attendance.
• LC and EAA models yield similar QALY weights if non-attendance reflects preference heterogeneity.
• QALY weights differ from standard models if non-attendance simplifies choices.

Conclusions:

• Attribute non-attendance significantly impacts health state valuation models.
• The interpretation of non-attendance (heterogeneity vs. simplification) is critical for QALY weight estimation.
• Policymakers should use a range of weights from LC and EAA models for sensitivity analysis due to unknown non-attendance causes.