Creation of New Race-Ethnicity Codes and Socioeconomic Status (SES) Indicators for MEdicare Beneficiaries - Chapter 5a
5. Conduct of Limited Multivariate Logistic Regression Analyses (continued)
5.6 Impact of Adding an SES-Race/Ethnicity Interaction Term to the Model
To fully investigate how race/ethnicity and SES work in explaining the variance of these health care utilization measures, we ran a third logistic model that included the interaction of race/ethnicity and SES. This model enabled us to investigate whether changes in the utilization of services differed by race/ethnicity depending on the level of the beneficiary's SES. In general, as SES level increased so did the rate of utilization, regardless of race/ethnicity. The amount of improvement in utilization, nonetheless, did to some extent depend on the beneficiary's race/ethnicity. In general, White beneficiaries seem to be deriving most of the overall improvement in amount of service use as SES status increased, while Hispanics benefited the least amount. The estimated percentages of service use by SES level and race/ethnicity with age and gender controlled are presented in Table 5.4.
Table 5.4 Percentage of Beneficiaries Receiving Selected Services by Race/Ethnicity and SES Level
|Race/Ethnicity||SES Level||Cancer Screening Services||Diabetes Secondary Prevention Services||ACSC|
and Pap Smear
Impact of SES-Race/Ethnicity Interaction on Cancer Screening Use Differences. For all three cancer screening measures, as SES status increased from the lowest level to the highest, so did the utilization for White beneficiaries; nine percentage points for males receiving a PSA test, seven percentage points for the percent of females receiving both a mammogram and a Pap smear, and five percentage points for beneficiaries receiving a colorectal exam. Although, improvements in the rate of service use for the minority groups existed, the improvement was, with only a few exceptions, much less than for White beneficiaries. This was especially true for the Hispanics, whose increase in service use from low SES to high SES did not exceed one percentage points for any of the three cancer screening measures.
For colorectal cancer screening, the amount of improvement as SES status increased did not exceed two percentage points for Blacks or Asians/Pacific Islanders and it was only three percentage points for American Indian/Alaska Natives, but for Whites there was a five percentage point increase from the lowest to the highest SES level. Likewise, the percent of male Hispanic and male Asian/Pacific Islander beneficiaries receiving a PSA test changed only one percent from lowest to highest SES level. However, for Black and American Indian/Alaska Native males, the percentage receiving a PSA test increased three and six percentage points, respectively, as SES level increased. Yet, for White males there was a nine percentage point increase from lowest to highest SES level.
The combined results for mammogram and Pap smear were slightly different than the other two cancer screening measures. White female beneficiaries had the highest utilization and Asians/Pacific Islanders had the lowest utilization, across all levels of SES, yet, the rate of increase was similar for the two groups (seven and six percentage point increases for Whites and Asian/Pacific Islanders, respectively). American Indians/Alaska Natives had a six percentage point change also. Utilization went relatively unchanged for Blacks, and Hispanics as SES level increased from lowest to highest.
Impact of SES-Race/Ethnicity Interaction on Secondary Diabetes Preventive Service Use Differences. Among the three diabetes health care utilization measures, eye exam demonstrated the most interesting interaction. At the lowest SES level, receipt of an eye exam differed across race/ethnicity groups by five percentage points, with White and Hispanic beneficiaries having the highest (57 percent) and American Indian/Alaska Native beneficiaries, the lowest (52 percent). As SES level increased, the percentage of White beneficiaries receiving an eye exam increased at a much greater rate than for the minority groups; nine percentage points for Whites compared to five percentage points for Asian/Pacific Islanders, four percentage points for Blacks, two percentage points for American Indians/Alaskan Natives, and only one percentage point for Hispanics.
The percentage of minority beneficiaries receiving the physiologic measure services changed moderately going from low SES to high SES. Asians/Pacific Islanders and Whites experienced a three percentage point increase, while Blacks had a four percentage point increase, from low to high SES. Hispanics and American Indians/Alaska Natives were the exceptions. Hispanics only saw a one percentage point increase while American Indians/Alaska Natives changed eight percentage points from low to high SES.
The one unexpected result we obtained among the diabetes use measures was associated with the receipt of instruction on self care. We found that as the level of SES increased, the percentage of beneficiaries receiving instruction in self-care decreased. This decrease occurred across all race/ethnicity groups, but most notably for White, Black, and Asian/Pacific Islander beneficiaries. At the lowest SES level more White beneficiaries received instructions in self-care (57 percent) than any other group (34 percent for American Indians/Alaskan Natives, 47 percent for Asian/Pacific Islanders, 49 percent for Hispanics, and 55 percent for Blacks). Nonetheless, the disparity between Whites and the minority groups narrowed as SES status increased. The percent of White and Black beneficiaries receiving instructions in self-care declined to 49 and 50 percent, respectively, at the highest SES level, Asians/Pacific Islanders declined to 43 percent, while for Hispanic only declined by one percent. Utilization of this service did increase slightly (one percent) among American Indians/Alaska Natives as SES level increased.
Impact of SES-Race/Ethnicity Interaction on Differences in the Use of Any ACSC. Regardless of race/ethnicity, as SES level increased, the percentage of ACSC hospitalizations for ambulatory care sensitive conditions decreased. This decreasing trend was most pronounced for Hispanics and American Indians/Alaska Natives, both of whom experienced a change of three percentage points in ACSC hospitalization from the lowest to the highest SES level. Whites and Blacks had a slightly lower decrease in the percentage of ACSC hospitalizations going from low to high SES level, two percentage points. The rate of decline for the percentage of Asian/Pacific Islander beneficiaries with ACSC hospitalizations was much smaller, declining by only one percentage point as SES level increased from the lowest to the highest level.
5.7 Conclusions from the Modeling
The conclusions we can draw about the impact of socioeconomic status (SES) on racial/ethnic health care disparities among Medicare beneficiaries from these selected multivariate analyses are, of course, limited. We have produced tabular analyses investigating the impact of SES and race/ethnicity on 45 outcomes of interest to this task order sub-task, but we have only performed multivariate analysis on a selected seven of those. Thus, our conclusions can be little more than suggestions about what may be found upon close inspection of the tables or with multivariate analyses of more numerous utilization measures.
Across these seven analyses, it is clear that SES does impact the level of service use and therefore the disparity between the service use of White Medicare beneficiaries and those who are members of racial/ethnic minority groups. Taking into account SES does seem to reduce health care disparities, probably because the minority groups are more highly represented in the lower SES levels. However, we found that disparities between Whites and some race/ethnicity groups are more affected than others when SES is controlled. It also seems to make a difference what health service is being analyzed.
Looking at the interaction of SES and race/ethnicity is enlightening in other ways, because it suggests that for some racial/ethnic groups, the magnitude of the disparity between White and minority beneficiaries differs according to the SES level of the beneficiaries. It indicates that among some minorities, even being in the highest SES level does not make their utilization more like Whites in that same SES level. This seems to be most applicable to Hispanics and Blacks.