The Impact of Crosswalk Algorithm Choice on COVID-19 Utility Values Derived From a Vignette Study

OBJECTIVES: A vignette study was used to derive utility values for COVID-19 disease states, using the EQ-5D-5L instrument. Due to the lack of a validated UK value set and the preference for EQ-5D-3L utility values, the National Institute of Health and Care Excellence (NICE) recommends the use of a crosswalk algorithm to the EQ-5D-3L value set. NICE originally referenced the Van Hout algorithm, but more recently stated preference for the Hernandez-Alava algorithm, on the basis of increased functionality and reliability. This study aimed to compare the use of the two algorithms, to understand their impact on utility scores.

METHODS: A representative sample of UK adults (N=500) acting as patient proxies completed the EQ-5D-5L questionnaire for eight vignettes which qualitatively described disease states for COVID-19. EQ-5D utilities were derived using the Van Hout and Hernandez-Alava crosswalk algorithms and the EQ-5D-3L UK value set.

RESULTS: The crosswalk algorithm choice had the least impact upon the health state representing baseline HRQoL, without COVID-19 infection (mean utility score difference of 0.002). The largest impact was seen upon the state representing long COVID, where the Hernandez-Alava algorithm generated a mean utility 0.056 lower than Van Hout. In most (6/8) health states, the Van Hout algorithm generated a higher mean utility than the Hernandez-Alava algorithm.

CONCLUSIONS: The choice of EQ-5D-5L to EQ-5D-3L crosswalk algorithm had a limited impact upon the resulting EQ-5D utility values, except for the long COVID disease state where the difference exceeded the minimally clinically important difference as validated in respiratory disease. The true impact of algorithm choice in prospectively collected data should be explored. There is a clear need for a validated UK value set for EQ-5D-5L, to resolve the reliance on crosswalk algorithms and create a standard approach in the future.