Keywords: Nonresponse bias, administrative data, CAHPS, Medicare
The MCAHPS survey is an essential tool for assessing quality in health care, but response rates to the surveys have been in decline, raising concerns about nonresponse bias. Using Traditional Medicare beneficiaries from 2010 to 2012 surveys linked to CMS administrative records, we compared response rates for beneficiaries across diagnoses and health care service types, and estimated ordinary least squares and logistic regressions to assess multivariate predictors of response. We then calculated the potential effect of nonresponse weighting on survey estimates. Beneficiaries with cognitive conditions like Alzheimer’s disease were less likely to respond while those with cataracts or glaucoma were more likely to respond. After controlling for demographics and other beneficiary enrollment characteristics like dual status and disability status, utilization of health care services and certain chronic conditions remain predictive. While MCAHPS respondents represent a population of beneficiaries that is more likely to have chronic conditions and use health care services, nonresponse weighting had relatively small effects on survey estimates.