How should we capture health state utility in dementia? Comparisons of DEMQOL-Proxy-U and of self- and proxy-completed EQ-5D-5L.

Objective Investigate the construct validity, criterion validity and responsiveness of DEMQOL-Proxy-U and of self- and proxy-completed EQ-5D-5L. Methods The analysis used a three-wave, individual-level dataset of 1,004 people living with dementia in residential care that included self-completed EQ-5D-5L and formal- and informal-carer proxy-completed EQ-5D-5L and DEMQOL-Proxy-U utility values, as well as other non-utility cognitive (FAST, CDR, CMAI) and HRQOL measures (QOL-AD-NH, QUALID). Construct validity, criterion validity and responsiveness was assessed using correlation, Bland-Altman plots and panel data regression models. Results Self-completed EQ-5D-5L failed to reflect clinically important differences and changes in FAST, CDR and CMAI, but did capture the resident’s own view of HRQOL (QOL-AD-NH). As dementia severity increased, collection of EQ-5D-5L-proxy and DEMQOL-Proxy-U data was more feasible than self-completed EQ-5D-5L. These formal- and informal-carer proxy-measures also better reflected changes in FAST, CDR and CMAI but didn’t capture the resident’s own view of HRQOL (QOL-AD-NH), despite adequately capturing the proxies own view of the resident’s HRQOL (QUALID). This indicates discrepancies between a proxies-view and residents-view of the impact tangible declines in health, cognition or functional abilities have on HRQOL. The EQ-5D-5L-proxy and DEMQOL-Proxy-U were generally poor substitutes. Regardless of which proxy completed it, EQ5D-5L-proxy was typically more responsive than DEMQOL-Proxy-U to changes in CDR, FAST and CMAI, indicating that use of DEMQOL-Proxy-U is not always justified. Conclusion Disparities in the measurement properties of different utility measures mean that choices about how to measure utility in trials could affect economic evaluation outcomes and hence how resources are allocated for dementia care.