|
|
|
|
|
Marc Elliott, PhD, Health, RAND, 1776 Main Street, P.O. Box 2138, Santa Monica, CA 90407 and Kelly Chong, MHA, M Phil, West Los Angeles Veterans Administration, 11301 Wilshire Blvd., MIRECC (210A), Los Angeles, CA 90073, (310) 478-3711 x48338, kchong1@ucla.edu.
Many seniors are incapable of responding to health surveys without some assistance, which may range from help with reading or writing to a proxy, such as a family member, responding in the place of the senior. Many surveys, such as the nationally representative 2001 Medicare FFS CAHPSâ Survey of 110,215 Medicare FFS beneficiaries, allow the use of proxy respondents so as to not omit the experiences of these most vulnerable seniors. Questions arise about the extent to which proxy responses are equivalent to those which would have been obtained from a beneficiary with similar experiences who would have been capable of responding unassisted. To date, this question has been addressed through case-mix adjustment (CMA), a regression-based approach, with “assisted” and “proxy respondent” cases distinguished from “unassisted” cases, controlling for age, health status, and education.
The CMA approach suggests that assistance results in global ratings that are moderately lower and that proxy use resulted in global ratings that are substantially lower than would have been obtained without assistance (-0.21 and –0.37 standard deviations, respectively; p<0.001 in each case). By this method, objective report items are equivalent for assisted and unassisted respondents, but are somewhat lower with proxies (-0.12 standard deviations; p<0.001).
When subgroups, such as unassisted and proxy respondent cases, have very little overlap, CMA and other regression methods can be inefficient, in that they give equal weight to all cases, including those cases with almost no chance of membership in the group to which they are being compared. We therefore employed propensity score weighting to focus the comparison on the subgroup of beneficiaries who most resembled the beneficiaries who employed assistance or proxy respondents. These analyses suggest that the true effect of proxies is about half as large as originally estimated for global ratings (-0.17 standard deviations) and negligible for objective report items (-0.04 standard deviations).
These findings suggest that with propensity score weighting to adjust for what are primarily underlying health differences, proxy and assisted responses may be quite accurate representations of the health experience of seniors. These findings are encouraging for the use of proxy respondents in settings such as nursing homes where cognitive limitations are greater than in the general Medicare beneficiary population.
Learning Objectives:
Keywords: Medicare, Survey
Presenting author's disclosure statement:
I do not have any significant financial interest/arrangement or affiliation with any organization/institution whose products or services are being discussed in this session.