Examining Race and Ethnicity Information in Medicare Administrative Data

Racial and ethnic disparities are observed in the health status and health outcomes of Medicare beneficiaries. Reducing these disparities is a national priority, and having high-quality data on individuals’ race and ethnicity is critical for researchers working to do so. However, using Medicare data to identify race and ethnicity is not straightforward. Currently, Medicare largely relies on Social Security Administration data for information about Medicare beneficiary race and ethnicity. Directly self-reported race and ethnicity information is collected for subsets of Medicare beneficiaries but is not explicitly collected for the purpose of populating race/ethnicity information in the Medicare administrative record. As a consequence of historical data collection practices, the quality of Medicare’s administrative data on race and ethnicity varies substantially by racial/ethnic group; the data are generally much more accurate for whites and blacks than for other racial/ethnic groups. Identification of Hispanic and Asian/Pacific Islander beneficiaries has improved through use of an imputation algorithm recently applied to the Medicare administrative database. To improve the accuracy of race/ethnicity data for Medicare beneficiaries, researchers have developed techniques such as geocoding and surname analysis that indirectly assign Medicare beneficiary race and ethnicity. However, these techniques are relatively new and data may not be widely available. Understanding the strengths and limitations of different approaches to identifying race and ethnicity will help researchers choose the best method for their particular purpose, and help policymakers interpret studies using these measures.