Disparities in Health Measures By Race and Ethnicity Among Beneficiaries in Medicare Advantage: A Review of the Literature
Data and Methods for Comparing Racial/Ethnic Disparities in Medicare Advantage
Study inclusion and exclusion criteria
This literature review summarizes findings from 20 studies published between January 1, 2018 and April 1, 2023. These 20 studies include 18 studies that compare measures of quality of care and beneficiary experience between people of color in Medicare Advantage plans and White enrollees, and two studies that report findings for people of color relative to all Medicare Advantage enrollees rather than just White enrollees. Two of the 20 studies presented estimates for specific subgroups (e.g., Black enrollees in rural areas in Medicare Advantage plans) without presenting overall estimates for a particular racial or ethnic group (e.g., Black enrollees in Medicare Advantage plans overall).
Studies were selected for the review if they included data for at least one year from 2013 or later. Twelve studies used data from a year between 2018 and 2023, while the remaining (8 studies) used data from a year between 2013 and 2017 (Appendix Table 2). The data period is relevant because the Affordable Care Act made substantial changes to how Medicare Advantage plans are paid, which were not fully phased in for several years after that legislation was enacted in 2010, and so their effects may not be fully captured by studies that use older data.
To meet the inclusion criteria, studies also had to include a transparent discussion of methods and data sources, including discussion of limitations. Most studies included in this literature review are articles from peer-reviewed journals, but this review also includes studies published by independent policy and research groups as well as government reports. The brief excludes studies that were fully funded by advocacy or industry groups.
To collect relevant studies, keyword searches were conducted of PubMed, Google Scholar, and other academic search engines, as well as the websites of governmental, research, and policy organizations that publish work related to health care. Additional studies were found using a snowball technique based on bibliographies of previously pulled studies. While the approach was designed to be as comprehensive as possible in including studies that meet the criteria, it is possible that some relevant studies were overlooked.
All differences reported in the text are statistically significant (with p-value less than or equal to 0.05) unless noted otherwise (e.g., for results that are reported as similar). In a few studies, researchers distinguished differences that were statistically significant, but not practically significant due to very small differences in estimates, although the threshold for practical significance varied across studies (Appendix Table 1).
Methods and data used in studies to compare Medicare Advantage enrollees by race and ethnicity
Studies in this review presented stratified estimates for some but not all of the racial and ethnic groups listed in current federal minimum standards. Current federal minimum standards for collecting and presenting data on race and ethnicity, as specified by the Office of Management and Budget (OMB), include the following ethnic groups: 1) not Hispanic or Latino and 2) Hispanic or Latino; and the following racial groups: 1) American Indian and Alaska Native, 2) Asian, 3) Black or African American, 4) Native Hawaiian or Other Pacific Islander, and 5) White. These standards were last updated in 1997, with OMB proposing to update them again in 2023 to reflect the increasing diversity of the U.S population and evolved immigration patterns. A report by the Office of Inspector General (OIG) found that Medicare’s enrollment data are inconsistent with federal data collection standards, limiting researchers’ ability to ensure that their racial/ethnic stratifications are in alignment with federal data collection standards. This report describes results for each of the racial or ethnic groups included in the 20 studies. If a particular racial or ethnic group is not mentioned, it means that estimates for that particular racial or ethnic group for that specific measure were not presented.
Data sources used in these studies varied in how they identified race and ethnicity of enrollees. For example, a study that used the Health Outcomes Survey used self-reported data to identify enrollees’ race and ethnicity. The Healthcare Effectiveness Data and Information Set (HEDIS), which was used as a source of data in eight studies, uses an imputed methodology that combines CMS’ administrative data, surname, and residential information to identify enrollees by race or ethnicity. This method is recommended for providing estimates for White, Black, Hispanic, and Asian and Pacific Islander enrollees, but not for American Indian and Alaska Native enrollees. Therefore, some studies were able to present estimates for American Indian and Alaska Native enrollees on CAHPS measures, but not HEDIS measures. Two studies used claims data from a single insurer (UnitedHealth), which uses a proprietary imputation method that cross-references enrollees’ names and zip code to a nationally recognized supplier of consumer marketing data to generate a weighted prediction of race/ethnicity from over 180 ethnicities.
Analyses varied in data sources used to compare measures of quality of care and beneficiary experience by race and ethnicity. Most studies reviewed here (17 out of 20 studies) used nationally representative data sources (Appendix Table 2), such as the Medicare Current Beneficiary Survey and the HEDIS. Two studies used claims data collected from a single health plan (UnitedHealth) that covered enrollees within the plan nationally, and one study used electronic health records data from primary care facilities across 10 states. No studies were identified that used Medicare Advantage encounter data.
Most studies used multivariate regression models to account for differences in beneficiary characteristics, including differences in health status. Of the 20 studies in this review, most (17 studies) used multivariate models to account for differences in the characteristics of enrollees in Medicare Advantage including demographic, socioeconomic, and health risks, though they varied in methodology and transparency (Appendix Table 2). A few studies further controlled for plan and contract-level characteristics, such as county-level market penetration rates, plan type (e.g., HMO versus PPO), integrated health system status, and whether the contract had one or more Special Needs Plans. In the case of studies that presented both bivariate and multivariate estimates, the adjusted estimates are reported in this review. The remaining three studies that did not present findings from multivariate models examined measures of beneficiary experience, rather than quality of care.
The use of more advanced statistical methods varied. Two studies created matched samples (e.g., using propensity score matching to balance samples of enrollees in vertically-integrated plans versus other Medicare Advantage plans) or included inverse probability of treatment weights in the regression model as a further attempt to adjust for differences in the likelihood of certain groups to enroll in certain Medicare Advantage plans over others (i.e., integrated plans versus non-integrated plans) (Appendix Table 2). No studies were identified that used a quasi-experimental design, such as difference-in-differences or an instrumental variable approach, to isolate the effect of race/ethnicity among Medicare Advantage enrollees on outcomes of interest.