Mapping EQ-5D-5L Utilities in Dementia: Integrating Self and Proxy Reports

EQ-5D is widely used in dementia research, but cognitive impairments often necessitate proxy assessments, resulting in differences between self- and proxy-reported data. Some EQ-5D dimensions are better reported by people with dementia (PwD), particularly in the mild to moderate stages, whereas others are more accurately captured by proxies. This study evaluates whether a combined utility value, integrating both PwD and proxy reports, can be predicted when data from only 1 respondent type are available.

Data from 2 dementia studies, ACTIFCARE and EPIC, were used to develop mapping models aimed at predicting combined utility values. These models integrate dimension-specific EQ-5D-5L responses from both respondent types to enhance health-related quality-of-life (HRQoL) assessments. Response mapping with ordered probit models was used to predict EQ-5D-5L responses when only 1 respondent type's data were available. Model performance was evaluated by comparing observed and predicted data across dementia severity stages.

Combined utility values provided a more accurate reflection of HRQoL, showing greater sensitivity to health status changes. Proxy data proved to be more effective in predicting PwD responses for certain EQ-5D-5L dimensions, suggesting that proxy data collection may be particularly useful in specific situations. The mapping models performed well for mild-to-moderate dementia but were less accurate in severe dementia because of limited data.

Combined utility values improve HRQoL assessments, particularly in detecting subtle health changes in mild-to-moderate dementia. These models support their use in economic evaluations of dementia interventions, although challenges remain in severe dementia because of data limitations.