Reports of fewer activity limitations: recovery, survey fatigue, or switching respondent?

Very little is known about the validity and reliability of disability dynamics reported in household surveys, and some researchers argue that measurement error likely plays a large part in reported recovery from disability.

We assessed the reliability of reported recovery from activity limitations elicited from 2 types of questions. We assessed competing hypotheses explaining reported recoveries from disability: people are less likely to recover from more severe disabilities, switching between self- and proxy-response affects reported recovery, and survey fatigue reduces reported disability.

Using the second panel of the Medical Expenditure Panel Survey, we estimated kappas for 2 types of questions in the same interview about limitations in activities of daily living and instrumental activities of daily living. We estimated multinomial logit models of consistently reported recovery, consistently reported ongoing limitations, and inconsistent responses. Recovery is a function of severity, switching respondent, measures of survey burden (such as family size), age, and education.

Within an interview, we found substantial reliability for both instrumental activities of daily living and activities of daily living limitations (kappa = 0.62 and 0.70, respectively). Sample members with more severe disabilities are less likely to report recovery, which is consistent with accurate reporting. Controlling for severity, the type of respondent affects reported recovery. Measures of survey burden did not affect reports.

Researchers can be confident in reports of recovery in the Medical Expenditure Panel Survey, especially when disability status was self-reported in both interviews. Researchers may also want to control for proxy respondents and switching respondents in their analyses.