Presentation (CTI)

OBJECTIVES: To compare utilities from the 2026 UK EQ-5D-5L value set with 5L-to-3L mapping and crosswalk approaches across the complete EQ-5D-5L state space and adjacent transitions.
METHODS: All 3,125 EQ-5D-5L states were valued using the Rowen et al. (2026) UK value set, Hernández Alava et al. 5L-to-3L mapping, and van Hout et al. crosswalk. For mapping, utilities were estimated for 2024 Office for National Statistics (ONS) adult age-sex strata and averaged using population weights. States were equally weighted; ONS weights standardised mapping age-sex covariates, not health-state prevalence. Severity score summed five domain levels (range: 5 [11111] to 25 [55555]). Differences were summarised overall and by severity score. 12,500 adjacent one-level worsening transitions were generated, holding other domains constant, and utility decrements and non-monotonicity were assessed (utility increase after worsening).
RESULTS: Mean utility was 0.322 using Rowen et al., 0.242 using ONS-weighted Hernández Alava et al. mapping, and 0.184 using van Hout et al. Rowen utilities were higher than mapping utilities for 79.4% of states (mean directional difference: 0.080); 70.5% had an absolute between-method utility difference >0.05 units. Mean directional difference peaked at 0.094 at severity score 15, but was −0.023 at score 22 and −0.041 for 55555. Across transitions, mean absolute decrement difference versus mapping was 0.052 utility units. 40.7% and 12.9% had absolute decrement differences >0.05 and >0.10 utility units, respectively. Hernández Alava et al. mapping generated 1,425/12,500 (11.4%) non-monotonic transitions, versus none under Rowen et al. or van Hout et al.
CONCLUSIONS: The 2026 UK EQ-5D-5L value set was not interchangeable with mapped or crosswalk-derived utilities. Differences were profile- and transition-specific, with mean differences becoming negative at selected higher severity scores. Mixing approaches may produce inconsistent utility estimates for the same EQ-5D-5L health state.