Bias in proxies' reports of disability: data from the National Health Interview Survey on disability.

OBJECTIVES: These studies examined whether differences between self-reports and proxy reports of disabilities reflect proxy response biases or only respondent selection factors.
METHODS: The data were from the National Health Interview Survey on Disability (1994-1995, phases 1 and 2). In study 1, reports of disabilities were regressed on respondent status, self vs proxy, and demographic factors. In study 2, the ratios of the proportions of self-reports and proxy reports of disabilities were regressed on independent lay ratings of observability of these disabilities and their "interactional" nature. In study 3, the disability reports for people who differed in respondent status in one phase but self-reported the same disability in the other phase were compared.
RESULTS: In study 1, proxies under-reported disabilities for people aged 18 to 64 years but overreported for people 65 years or older. In study 2, the observability and interactional scores accounted for more than 60% of the variance of self and proxy differences in an inverse relationship, study 3 confirmed the basic findings of study 1.
CONCLUSIONS: Use of proxies in representative surveys on disability introduces systematic biases, affecting national disability estimates.