Variation in Health State Utility Values Across Preference-Based Instruments: A Systematic Review and Meta-analysis...

Health economic evaluations require health state utility values (HSUVs) to estimate quality-adjusted life years (QALYs). However, HSUVs derived from different preference-based instruments, valuation methods and value sets may not be directly comparable, which can affect interpretation of economic evaluations. Although reference cases, such as that of the National Institute for Health and Care Excellence (NICE), specify a preferred instrument, such as the EQ-5D, alternative measures are frequently used in practice. Using behavioural risk factors as an applied case, this study examined variation in HSUVs across instruments and synthesised utility estimates for body mass index (BMI) and smoking categories.

A systematic review of population-based observational studies published between 2009 and 2024 was conducted across PubMed, Embase, Web of Science and EconLit. Random-effects meta-analyses were used to estimate pooled HSUVs by instrument type and risk factor category. Mixed-effects meta-regression models explored associations between behavioural risk factors, health-related quality of life (HRQoL) instrument type and variation in HSUVs while accounting for clustering of multiple estimates within studies. Preference-based instruments included the EQ-5D, Health Utilities Index (HUI), Short Form 6-Dimension (SF-6D) and Quality of Well-Being (QWB) scale. Given the anticipated heterogeneity across studies, pooled estimates were used to describe variation in reported HSUVs.

Forty-four articles were included, contributing 186 utility estimates across five instrument categories. Substantial variation in HSUVs was observed across instruments with consistent patterns evident across BMI and smoking categories. Relative to the EQ-5D-3L, the HUI, SF-6D and QWB yielded systematically lower utility values. The EQ-5D-5L yielded higher values in the BMI model, but the corresponding estimate in the smoking model was not statistically significant. Variation associated with instrument categories, which may also reflect related valuation and derivation differences, was comparable, and in some cases greater than, differences observed across BMI and smoking categories, although substantial between-study heterogeneity remained (I² > 90%). Lower HSUVs were observed among individuals who were underweight, overweight or obese compared with those of normal weight and among former and current smokers compared with never smokers. Former smokers reported lower values than current smokers, consistent with selection effects. Alcohol use was reported in too few studies to support reliable instrument-stratified pooling, and reported values overlapped substantially across drinking categories.

Reported HSUVs varied systematically across preference-based instruments, with differences comparable to, and in some cases exceeding, those associated with behavioural risk factors. Because instrument versions, value sets, valuation methods, mapping procedures and population characteristics could not be fully separated, these differences should be interpreted as reflecting broader utility measurement and valuation approaches rather than independent instrument effects. Utility-source selection should be transparently justified, aligned with relevant reference-case recommendations and explored through sensitivity analyses where alternative credible inputs are available.