OBJECTIVES: The updated NICE methods manual (PMG36) formally recommends adjusting utility values for age in economic models using general population utility as a baseline. However, NICE Decision Support Unit (DSU) guidance indicates that utility estimates without a condition may be most appropriate to use as a baseline when adjusting utility values for age. Outside of analyses performed on Health Survey for England (HSE) data for cardiovascular disease (CVD) patients (Ara 2010), very few studies have explored the difference between general population utility and utility for people without a condition. Furthermore, it is unclear whether similarities hold between general population utility and utility among people without a condition when fitting an adjusted limited dependent variable mixture model (ALDVMM), which may be a more appropriate and flexible model for EQ-5D data. We sought to test if general population utility was comparable to alternatively utility baselines for three specific patient groups available from HSE EQ-5D-3L datasets.

METHODS: OLS and ALDVMM regression models were fitted individually for patient groups without specific conditions (diabetes [n=70,951], CVD [n=46,701], hypertension [n=58,129]) and compared against corresponding general population utility models using available EQ-5D-3L HSE datasets (2003-2014, total N=89,998).

RESULTS: OLS and ALDVMM models for patients without each condition were very similar to corresponding general population models. While some differences in EQ-5D-3L utility were observed among the older age groups (>90 years old) in the female cohort in some model comparisons, this may be due to relatively limited sample sizes being available at older age groups.

CONCLUSIONS: Among the three groups assessed, general population utility appeared to be a reasonable proxy for utility estimates without the conditions, indicating that use of general population utility as a baseline when adjusting utility values for age in economic models may be appropriate. However, further validation studies are required to test this assumption across other medical conditions.