DEVELOPMENT AND EXTERNAL VALIDATION OF A NOVEL COPULA-BASED METHOD FOR CROSS-COUNTRY EQ-5D UTILITY VALUE CONVERSION
Author(s)
Chris D. Poole, BSc, PhD1, Rashad Carlton, PharmD, MSPH2, Paul Turner, PhD3, Natalia Hakimi Hawken, MSc, PhD4;
1Cencora, Director, Economic Modelling, Uxbridge, United Kingdom, 2Cencora, Economic Modelling, Conshohocken, PA, USA, 3Cencora, Uxbridge, United Kingdom, 4Merck, Schiphol Rijk, Netherlands
1Cencora, Director, Economic Modelling, Uxbridge, United Kingdom, 2Cencora, Economic Modelling, Conshohocken, PA, USA, 3Cencora, Uxbridge, United Kingdom, 4Merck, Schiphol Rijk, Netherlands
OBJECTIVES: International health technology assessments frequently require conversion of EQ-5D utility values between country-specific tariffs, yet no standardized methodology exists. Therefore, this study aimed to develop and validate a novel method for converting UK EQ-5D-3L values to country-specific equivalents by repurposing the NICE-recommended 3L/5L mapping copula model's (Hernandez-Alava [2020]) intermediate caliper-matching algorithm.
METHODS: The conversion method utilized the Hernandez-Alava model's mechanism for identifying EQ-5D health states falling within an empirically specified caliper width (0.100) of target UK values. For each UK utility input, caliper-matched health states were identified, alternative country tariffs applied, and mean values calculated to derive country-specific equivalents. External validation employed the Van Wilder (2019) systematic review catalogue containing paired observations of UK and local EQ-5D values across diseased populations. Predicted local values were compared to observed values using Pearson and Spearman correlations, mean absolute error (MAE), root mean squared error (RMSE), and Bland-Altman analysis to assess agreement and systematic bias.
RESULTS: Validation included 389 paired observations across 13 target countries. Overall performance showed moderate correlation (Spearman ρ=0.556; Pearson r=0.391) with MAE=0.051 and RMSE=0.091. Bland-Altman analysis revealed slight systematic underprediction (mean bias=-0.034; 95% CI: -0.200 to 0.132). Performance varied substantially by country: USA demonstrated strong correlation (n=278; ρ=0.769; MAE=0.026), along with Sweden (6; 0.909; 0.035) and Brazil (6; 0.842; 0.094) while Singapore showed poorer correlation and larger errors (27; 0.266; 0.226) along with Korea (34; 0.379; 0.092) and Denmark (14; 0.356; 0.050).
CONCLUSIONS: This novel conversion method provides a theoretically grounded approach for cross-country EQ-5D utility transformation. While overall validation demonstrates acceptable accuracy, country-specific performance varies considerably, possibly due to unmeasured between-cohort confounding. The method performs well for US conversions and offers a standardized alternative to ad-hoc approaches currently employed in international health economic evaluations. Future work will optimize caliper selection and address validation cohort confounding.
METHODS: The conversion method utilized the Hernandez-Alava model's mechanism for identifying EQ-5D health states falling within an empirically specified caliper width (0.100) of target UK values. For each UK utility input, caliper-matched health states were identified, alternative country tariffs applied, and mean values calculated to derive country-specific equivalents. External validation employed the Van Wilder (2019) systematic review catalogue containing paired observations of UK and local EQ-5D values across diseased populations. Predicted local values were compared to observed values using Pearson and Spearman correlations, mean absolute error (MAE), root mean squared error (RMSE), and Bland-Altman analysis to assess agreement and systematic bias.
RESULTS: Validation included 389 paired observations across 13 target countries. Overall performance showed moderate correlation (Spearman ρ=0.556; Pearson r=0.391) with MAE=0.051 and RMSE=0.091. Bland-Altman analysis revealed slight systematic underprediction (mean bias=-0.034; 95% CI: -0.200 to 0.132). Performance varied substantially by country: USA demonstrated strong correlation (n=278; ρ=0.769; MAE=0.026), along with Sweden (6; 0.909; 0.035) and Brazil (6; 0.842; 0.094) while Singapore showed poorer correlation and larger errors (27; 0.266; 0.226) along with Korea (34; 0.379; 0.092) and Denmark (14; 0.356; 0.050).
CONCLUSIONS: This novel conversion method provides a theoretically grounded approach for cross-country EQ-5D utility transformation. While overall validation demonstrates acceptable accuracy, country-specific performance varies considerably, possibly due to unmeasured between-cohort confounding. The method performs well for US conversions and offers a standardized alternative to ad-hoc approaches currently employed in international health economic evaluations. Future work will optimize caliper selection and address validation cohort confounding.
Conference/Value in Health Info
2026-05, ISPOR 2026, Philadelphia, PA, USA
Value in Health, Volume 29, Issue S6
Code
P25
Topic
Patient-Centered Research
Topic Subcategory
Health State Utilities
Disease
No Additional Disease & Conditions/Specialized Treatment Areas