MODELLING HEALTH SYSTEM CAPACITY AND MEASUREMENT DISPERSION TO EXPLAIN CROSS-NATIONAL VARIATION IN EQVT UTILITY VALUES
Author(s)
Annushiah Vasan Thakumar, BSc, PhD1, Xun Li, BSc, MSc2, Ling Jie Cheng, PhD, MPH, BSN (Hons), RN3.
1School of Pharmacy, Faculty of Health & Medical Sciences, Taylor's University, Subang Jaya, Malaysia, 2School of Engineering, Computing, and Mathematics, Oxford Brookes University, Oxford, United Kingdom, 3National Perinatal Epidemiology Unit, Nuffield Department of Women's & Reproductive Health, University of Oxford, Oxford, United Kingdom.
1School of Pharmacy, Faculty of Health & Medical Sciences, Taylor's University, Subang Jaya, Malaysia, 2School of Engineering, Computing, and Mathematics, Oxford Brookes University, Oxford, United Kingdom, 3National Perinatal Epidemiology Unit, Nuffield Department of Women's & Reproductive Health, University of Oxford, Oxford, United Kingdom.
OBJECTIVES: Utility values vary widely across countries, yet the role of measurement dispersion remains unclear. This study examined predictors of mean EQ-VT utility scores and whether modelling dispersion improves prediction and inference.
METHODS: We conducted a secondary analysis of 40 EQ-5D-5L valuation studies implemented with the EQ-VT protocol between 2012 and 2024. Seventeen indicators spanning study-level factors, demography, health outcomes, economy, healthcare financing, and healthcare resources were linked at the country level to study-specific mean EQ-VT utility values. Univariable ordinary least squares (OLS) models and penalised regression using least absolute shrinkage and selection operator (LASSO) were estimated, alongside forward selection models; sensitivity analyses incorporated the standard deviation of the mean EQ-VT values to account for variance heterogeneity. Model performance was compared using adjusted R², root mean square error and Akaike information criterion, and Bayesian information criterion.
RESULTS: Across 40 valuation studies, the mean sample age was 44.4 years, 47.7% of participants were male, and 78.6% used EQ-VT version 2. In univariable models, studies from countries with more hospital beds and more physicians per 1,000 population had higher mean EQ-VT utilities; hospital bed density showed the strongest association (β = 0.020 per additional bed per 1,000 population; 95%CI: 0.005 to 0.036). LASSO selected EQ-VT version, hospital beds, and physician density, but only EQ-VT version remained statistically significant in post‑LASSO OLS. Forward selection retained hospital beds per 1,000 population (β = 0.016; 95%CI: 0.001 to 0.032), although overall model fit was modest. When the standard deviation of mean EQ-VT values was included to account for dispersion, none of the predictors remained statistically significant.
CONCLUSIONS: Country-level capacity indicators explained variation in mean EQ-VT utilities in conventional models but lost significance once dispersion was modelled. Ignoring dispersion may overstate the role of health system characteristics. Thus, cross-population borrowing should consider both mean utilities and their variability.
METHODS: We conducted a secondary analysis of 40 EQ-5D-5L valuation studies implemented with the EQ-VT protocol between 2012 and 2024. Seventeen indicators spanning study-level factors, demography, health outcomes, economy, healthcare financing, and healthcare resources were linked at the country level to study-specific mean EQ-VT utility values. Univariable ordinary least squares (OLS) models and penalised regression using least absolute shrinkage and selection operator (LASSO) were estimated, alongside forward selection models; sensitivity analyses incorporated the standard deviation of the mean EQ-VT values to account for variance heterogeneity. Model performance was compared using adjusted R², root mean square error and Akaike information criterion, and Bayesian information criterion.
RESULTS: Across 40 valuation studies, the mean sample age was 44.4 years, 47.7% of participants were male, and 78.6% used EQ-VT version 2. In univariable models, studies from countries with more hospital beds and more physicians per 1,000 population had higher mean EQ-VT utilities; hospital bed density showed the strongest association (β = 0.020 per additional bed per 1,000 population; 95%CI: 0.005 to 0.036). LASSO selected EQ-VT version, hospital beds, and physician density, but only EQ-VT version remained statistically significant in post‑LASSO OLS. Forward selection retained hospital beds per 1,000 population (β = 0.016; 95%CI: 0.001 to 0.032), although overall model fit was modest. When the standard deviation of mean EQ-VT values was included to account for dispersion, none of the predictors remained statistically significant.
CONCLUSIONS: Country-level capacity indicators explained variation in mean EQ-VT utilities in conventional models but lost significance once dispersion was modelled. Ignoring dispersion may overstate the role of health system characteristics. Thus, cross-population borrowing should consider both mean utilities and their variability.
Conference/Value in Health Info
2026-05, ISPOR 2026, Philadelphia, PA, USA
Value in Health, Volume 29, Issue S6
Code
MSR178
Topic
Methodological & Statistical Research
Topic Subcategory
PRO & Related Methods
Disease
No Additional Disease & Conditions/Specialized Treatment Areas