Creation of New Race-Ethnicity Codes and Socioeconomic Status (SES) Indicators for Medicare Beneficiaries
There has been considerable interest during the past decade or so in reducing racial and ethnic disparities in the use of health services. Understanding whether disparities result from sub-cultural differences in the practices of minorities or reflect the impact of different socioeconomic circumstances is key to designing interventions to reduce disparities. Studying racial/ethnic disparities in health care among Medicare beneficiaries should be close to ideal because it largely eliminates insurance- and cost-related access barriers as explanations of health care disparities. To a large extent, differences in access to health care are expected to be minimal for Medicare beneficiaries because there is a standard set of benefits for which all beneficiaries are eligible. In addition, there are assistance programs available to reduce the cost burdens associated with the monthly premiums for Part B (Supplemental Medical Insurance) as well as any coinsurance and deductibles for low income persons.
To rigorously investigate whether and where there are racial/ethnic health care disparities present in Medicare, it is critical to be able to assess the extent to which any disparities are associated with race or ethnicity rather than socioeconomic status (SES), because the impacts of these variables are often confounded, while other important factors such as age and gender are controlled as well. It has not been possible in the past to do this kind of analysis using Medicare administrative data alone because the enrollment database which contains the person-level characteristics of beneficiaries does not include an appropriate variable or surrogate to measure SES.
In addition, analyses of racial/ethnic health care disparities in the Medicare program have been limited to comparisons between White and Black/African American beneficiaries because of problems with the quality and completeness of coding on the EDB for Hispanic and Latino, Asian and Pacific Island, and American Indian and Alaska Native beneficiaries. It has been demonstrated that large numbers of beneficiaries of Hispanic and Asian decent have been erroneously identified on the Medicare enrollment database (EDB) (Arday et al., 2000; Eggers and Greenberg, 2000 Waldo, 2005). In a previous task order project (Contract Number 500-00-0024, Task 8, Health Disparities: Measuring Health Care Use and Access for Racial/Ethnic Populations, 2005) to identify racial and ethnic disparities in health care utilization and access among Medicare beneficiaries that RTI conducted for the Centers for Medicare & Medicaid Services (CMS), our first objective was to assess the accuracy of the racial/ethnic coding of beneficiaries listed in the EDB. We confirmed the same incomplete and incorrect coding of race/ethnicity. The second objective of that project was to develop an algorithm making use of surnames and other available information to upgrade the coding of the EDB race/ethnicity variable for Hispanic and Asian/Pacific Islander beneficiaries and to validate the results of the algorithm. We completed that work and a description of it is included in this report because we relied upon it. Some time prior to initiation of our work, CMS had negotiated an interagency agreement with the Indian Health Service to identify American Indian and Alaska Native beneficiaries to Medicare. We used the version of the EDB that included the upgraded American Indian/Alaska Native coding.
In addition to the incomplete and incorrect coding of race/ethnicity on the EDB that would make it difficult to conduct a rigorous assessment of the separate impact of race/ethnicity and SES on the use of health services, as we indicated above, the EDB does not contain any measures or indicators of SES. To remedy this situation, in the previous task order project (Contract Number 500-00-0024, Task 8, Health Disparities: Measuring Health Care Use and Access for Racial/Ethnic Populations, 2005), RTI successfully geocoded the addresses of 36.2 million Medicare beneficiaries to obtain a Federal Information Processing Standard (FIPS) code for the US Census Block Group (the smallest US Census area for which there are economic data reported from the Census) in which the beneficiary's address was located. We were then able to associate socioeconomic characteristics of a beneficiary's neighborhood with the beneficiary, although no analysis was done using this census data. The former task order contract under which the original surname algorithm and geocoding work was performed is the foundation on which the current task order project has been conducted.
The work conducted in task two of the current task order contract (contract 500-00-0024, task 21) with the Centers for Medicare & Medicaid Services (CMS) was undertaken with the purpose of further developing the work completed by RTI under the previous task order contract with CMS (contract 500-00-0024, task 8) in the preparation of the 2005 report titled Health Disparities: Measuring Health Care Use and Access for Racial/Ethnic Populations. The goal of this work is to empirically specify and isolate the effect of differences in SES from what are presently described as disparities in health care utilization associated with race and ethnicity among Medicare beneficiaries. The race and ethnicity variable developed to improve upon the coding of race/ethnicity available at the time on the Medicare enrollment database (EDB) used in the previous report was again used in this report. A summary of the naming algorithm work performed in the earlier study is presented in the Methods and Data section of this report.
The task activities associated with this work were made possible through an interagency agreement and transfer of funds from the Agency for Healthcare Research and Quality (AHRQ) to CMS. These task activities included the following:
- Hold a kickoff meeting, prepare monthly written progress reports, and participate in telephone conference calls as requested.
- Develop and validate a measure of SES for Medicare beneficiaries.
- Update the existing analytic data file with the new SES variable.
- Develop tabular presentations of descriptive statistics on use of selected preventive health services (cancer screening and secondary diabetes prevention), ambulatory sensitive conditions, and average length of stay and expenditures for hospitalizations common to Medicare beneficiaries based on the claims based on a previously selected weighted sample of 1.96 million. Do this for the nation as a whole as well as for designated metropolitan statistical areas (MSAs) – the top 10 for number of elderly Asians and Hispanics separately). This activity will also include limited multivariate logistic regression analyses of selected measures of utilization to summarize the impact of race/ethnicity, SES, age group, and gender.
- Prepare a final report describing the work and provide the analytic data file, a codebook for the analytic data file, and copies of the programs used to create the variables in the tabulations.
The objectives of the work performed under the second task in this task order were twofold. First, we wanted to create and validate a composite measure of SES based not on individual-level measures but on variables extracted from the 2000 U.S. Census that have been used to define the SES of beneficiaries' residential neighborhoods (block group). Because we were interested in income-related variables, the smallest aggregate unit for which this kind of data is made available by the Census is the block group, thus our choice of it as the area constituting the beneficiaries' residential neighborhood.
Our second objective in task two was to build upon the set of tabulations created for the previously completed task order discussed above. We were interested in building upon those tabulations in two ways. The first was by including SES in the tabulations along with the improved measure of race/ethnicity (that we created in the previous task order), and age group and gender. The second was to replicate the tabulations done at the national level for the 10 MSAs with the greatest concentration of Asian and Hispanics Medicare beneficiaries. Since the Census does not identify Medicare beneficiaries, we used the number of Asian, Native Hawaiian, and Pacific Islanders age 65 and over, and a similar number for Hispanics and Latinos, as our proxy for identifying the top 10 MSAs for both groups of Medicare beneficiaries.
1.4 Organization of this report
This first section of the report has described the background and purpose of this task order project. The remaining four sections of this report are devoted to descriptions of the work performed and the deliverables prepared. The next section discusses the general methods and data used to accomplish the objectives of the task. Section 3 describes how we developed and validated the SES index that we used in the tabulations. The fourth section describes the tabulations prepared in this task order and presented in separate Appendices. Section 5 discusses the limited multivariate logistic regression analyses performed to begin the onerous task of understanding the association of race/ethnicity, SES, age, and gender on disparities in health care use among Medicare beneficiaries. There are also six appendices to this report. The first and last ones (Appendix A and Appendix F) are included as part of the final report. The other appendices (B through E) are bound separately but are meant to be used in tandem with this report. They have been bound separately for the convenience of persons who may need to work with them in preparing other reports or presentations.