2006 HSR&D National Meeting Abstract
3096 — Use Spanish Surname Match to Reduce Bias in Interpreting Medicare Data for Hispanic/White Veterans
Wei II (Michael E. DeBakey Veterans Affairs Medical Center)
Virnig BA (University of Minnesota)
Morgan RO (Michael E. DeBakey Veterans Affairs Medical Center/Baylor College of Medicine)
Medicare administrative and claims files maintained by the Centers for Medicare and Medicaid Services provide a rich source of information for examining racial and ethnic disparities in health care use. However, there have been increased concerns over the accuracy of the race code in the Medicare data. Here, we demonstrate a method of using Spanish surname matching to augment the Medicare race code and evaluate its effectiveness in reducing bias in interpreting the Medicare data.
Using three types of race code (self-report, Medicare, and surname-match), we compared the racial/ethnic representation in terms of socio-demographic characteristics, health status, health conditions, and health care utilizations among 2,086 Medicare-enrolled male veterans. Data was from six large U.S. metropolitan areas by a mail survey in 2002.
We found that the Spanish surname matching substantially increased the accuracy for the identification of Hispanics and whites compared to the Medicare race code alone--using self-reported race/ethnicity identification as the gold standard. Because the Medicare race code has higher accuracy for blacks, the surname matching did not show as much effect. The comparison of the Medicare and the self-reported race codes revealed discrepancies in the veterans’ representation at various levels of individual characteristics (e.g., geographic distribution, health conditions)--mostly in Hispanics. In general, the surname-matched race code revealed close representation of veterans similar to the self-reported race codes.
Our findings demonstrate that augmenting the Medicare race code with a match to Spanish surnames yields substantial improvement in the identification of elderly Hispanic and white male Medicare beneficiaries. Consequently, one can generate more unbiased study results and interpretations. The use of a surname match to supplement the Medicare race code may greatly enhance researchers’ and policy makers’ ability to examine health care equity.
Inaccuracy in the identification of race/ethnicity in the Medicare data can potentially misguide the policy makers who may subsequently make a law that will miss the sub-group most in need of healthcare assistance.