OBJECTIVES: Proxy ratings primarily provided by informal caregivers are usually administered if people living with dementia (PlwD) are cognitively unable to rate health independently. Typically it is reported, that proxies underestimate PlwD health. However, analyses of self- and proxy-rated discrepancies in individual responses that focus on HRQoL dimensions are currently lacking. Therefore, we examined discrepancies between self- and proxy-rated health on an individual response level.

METHODS: We analyzed EQ-5D-5L baseline data of a cluster-randomized, controlled intervention trial, administered as a self- and proxy-proxy-assessment of n=174 dyads (PlwD and their caregivers). Initially, we calculated agreement statistics (weighted kappa) for each dimension of the EQ-5D. For individual response levels of each dimension, we analyzed the degree of inconsistencies between self- and proxy ratings ranging from 0 (identical responses in all dimensions) to 5 (different responses in all five dimensions). Further, we grouped the PlwD into “no problems” and “problems” depending on the EQ-5D-5L dimensions to compare the response distances between PlwD and caregivers.

RESULTS: PlwD had a mean age of 80.1 years (49.1% female) and caregivers a mean age of 67.9 years (67.8% female). Every third PlwD lived alone (62%), and 90% of PlwD were mild to moderate cognitively impaired (MMSE mean score 19.2). In general, PlwD rated their health better than proxies, with lowest kappa coefficients for usual activity (0.23) and highest for pain/discomfort (0.40). Nine dyads (5%) showed identical ratings. Among the grouped PlwD without and with problems, proxies estimated the PlwD as either worse (underestimation) or better (overestimation), coupled with more frequent mid-response options by proxies. Over- and underestimation was more present in mobility, usual activities, and self-care.

CONCLUSIONS: The two trends of contrary response behavior (over- and underestimation) of both rater types underline the need for further research when using HRQoL measures as proxy instruments and examining influencing factors in that context.