Reliability of proxy response on mental health indices for aged, community-dwelling women.

The correspondence between respondent and proxy response was evaluated on 4 mental health measures (Affect Balance Scale, Center for Epidemiological Studies Depression Scale, Mental Status Questionnaire, and Mini-Mental State Examination) with a sample of 538 respondent-proxy pairs. Results indicated that respondent and proxy responses were strongly associated, particularly for the cognitive measures. This association was found even for respondents classified as depressed or cognitively impaired. Although there was evidence of proxy bias, with proxies underrating affective status and overrating cognitive status, the magnitude of the bias proved small for all scales but the Mental Status Questionnaire. Examination of response comparability by proxy characteristics showed that choice of proxy affected agreement and bias. Implications of these findings for survey research are discussed.