Improving Measures of Medicaid Participation among Older Adults Using Administrative Data and Self Reports

Recent years have witnessed a large number of program innovations affecting older adults dually enrolled in Medicare and Medicaid, such as the increased use of managed care and numerous state demonstration waivers. To assess the impact of these and other changes on older adults’ Medicaid participation and to identify the effects of Medicaid enrollment on this population, researchers need accurate data on Medicaid coverage for older adults. Many studies in this area measure Medicaid participation using self-reported program participation data from large household surveys such as the American Community Survey (ACS), the National Health Interview Survey (NHIS), or the Current Population Survey (CPS). However, there is a real concern about potential misreporting of Medicaid coverage on surveys due to program complexities, differences in program names across states, and general confusion about the difference between Medicare and Medicaid.

In this paper, we examine whether relying on self-reports of Medicaid participation contributes to over- or under-estimates of Medicaid participation among older Medicare beneficiaries. We use several years of data from the Medicare Current Beneficiary Survey (MCBS), a nationally representative survey of roughly 11,000 Medicare beneficiaries per year containing both self-reports of Medicaid participation and administrative records of monthly Medicaid program enrollment.

To investigate the potential bias caused by self-reported data, we compare Medicaid participation measures derived from self-reports with those derived from administrative records. We then investigate how differences in program participation based on self-reported versus administrative data vary by respondent traits. We examine differences in the extent of bias between community-dwelling and facility-dwelling respondents, by respondent demographic characteristics, by the presence of limitations in activities of daily living (ADLs) and instrumental activities of daily living (IADLs), and by type of Medicaid coverage (i.e., full Medicaid or Medicare Savings Program only).

We then build on this approach by developing a procedure to adjust self-reported enrollment data for reporting error similar to that used for adjusting other types of self-reported data in other large surveys. We first use the MCBS data to regress actual Medicaid participation on self-reported participation separately by race, ethnicity, sex, and other factors shown to be associated with the extent of bias. We then use the resulting parameter estimates to construct adjusted Medicaid participation rates for use in other surveys, such as the ACS, CPS, or NHIS.

Our estimates of differences in Medicaid participation based on self-reported versus administrative data are relevant for a range of research studies including studies that model program participation, studies that seek to estimate the effects of Medicaid enrollment, and studies that wish to control for sources of insurance coverage. Our efforts may inform the development of future surveys that are able to capture improved data on Medicaid enrollment through self-reports.