|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||
Background Since 1997, all managed-care plans administered by Medicare have reported on quality-of-care measures from the Health Plan Employer Data and Information Set (HEDIS). Studies of early data found that blacks received care that was of lower quality than that received by whites. In this study, we assessed changes over time in the overall quality of care and in the magnitude of racial disparities in nine measures of clinical performance.
Methods In order to compare the quality of care for elderly white and black beneficiaries enrolled in Medicare managed-care plans who were eligible for at least one of nine HEDIS measures, we analyzed 1.8 million individual-level observations from 183 health plans from 1997 to 2003. For each measure, we assessed whether the magnitude of the racial disparity had changed over time with the use of multivariable models that adjusted for the age, sex, health plan, Medicaid eligibility, and socioeconomic position of beneficiaries on the basis of their area of residence.
Results During the seven-year study period, clinical performance improved on all measures for both white enrollees and black enrollees (P<0.001). The gap between white beneficiaries and black beneficiaries narrowed for seven HEDIS measures (P<0.01). However, racial disparities did not decrease for glucose control among patients with diabetes (increasing from 4 percent to 7 percent, P<0.001) or for cholesterol control among patients with cardiovascular disorders (increasing from 14 percent to 17 percent; change not significant, P=0.72).
Conclusions The measured quality of care for elderly Medicare beneficiaries in managed-care plans improved substantially from 1997 to 2003. Racial disparities declined for most, but not all, HEDIS measures we studied. Future research should examine factors that contributed to the narrowing of racial disparities on some measures and focus on interventions to eliminate persistent disparities in the quality of care.
Recently, signs of an improved quality of care have been evident both within and outside managed-care settings.10,11,12 These improvements may be related to efforts to measure and report clinical performance. However, little is known about whether general improvements in the quality of care are also accompanied by reductions in racial and ethnic disparities. Quality-improvement efforts may reduce such disparities, but they also may have no effect or even increase disparities.
A recent study of patients with end-stage renal disease suggested that broadly targeted interventions to improve the quality of care were associated with reduced racial disparities in hemodialysis dosing.13 However, an analysis of the administration of influenza vaccine to Medicare beneficiaries showed that higher rates of vaccination in managed-care plans, as compared with the rates in fee-for-service plans, were not associated with reduced disparities on the basis of race.7
Within the Medicare program, managed-care plans provide an opportune setting to examine the relation between improvements in the quality of health care and changes in racial disparities. Health plans are particularly well positioned to improve care, because they finance and monitor the provision of health services to enrollees.12,14 Since 1997, all health plans participating in Medicare have been required to submit publicly reported performance measures from the Health Plan Employer Data and Information Set (HEDIS) developed by the National Committee for Quality Assurance (NCQA). In previous studies of HEDIS measures, black enrollees were less likely to receive beta-blockers after myocardial infarction, eye examinations after receiving a diagnosis of diabetes, follow-up care after hospitalization for mental illness, and influenza vaccinations.6,8
In this study, we report on trends in the quality of care provided to enrollees in Medicare managed-care plans from 1997 to 2003 and assess whether racial disparities in quality changed during this period. Evaluation of HEDIS performance trends by race can provide important information about whether broad improvements are associated with narrowed or widened racial disparities in the quality of medical care.15
Methods
Study Population
We obtained HEDIS data for Medicare managed-care plans from the Centers for Medicare and Medicaid Services (CMS) covering seven reporting years (1998 to 2004) with information regarding clinical care that was delivered from 1997 to 2003. These data contained 2,691,482 observations for enrollees who were eligible for at least one of the nine HEDIS indicators described in Table 1. Each observation included the patient's health identification code and health plan, as well as variables indicating eligibility for and adherence to each HEDIS measure.
|
Using the health identification code, we matched each enrollee with HEDIS data on at least one measure with the file of Medicare enrollees for the corresponding year to obtain demographic information on the race, age, sex, and ZIP Code of residence of beneficiaries and to ascertain whether they also had Medicaid coverage. We achieved a match rate of 96 percent, or 2,573,166 observations. We excluded 229,938 observations for enrollees who were under the age of 65 years, 201,323 observations for enrollees who were of a race or ethnic background other than black or white, and 44,150 observations for enrollees who died during the year of measurement (with some overlap of enrollees in these three exclusion categories). This process yielded a total study sample of 2,122,809 observations, of which 9.2 percent were for black enrollees. To reduce the likelihood that trends in performance and disparities might be a result of the entrance and exit of health plans from Medicare, our primary analysis excluded 319,358 observations from health plans with less than five years of continuous participation in Medicare managed care. (In a secondary analysis, we did not exclude health plans with less than five years of participation and instead included observations for eligible patients from all health plans that participated in Medicare managed care for at least one year during the study period. Results were similar and are not shown.)
The primary study sample included 1,803,451 observations (9.4 percent for black enrollees) from 183 health plans for the nine HEDIS indicators. The sample size of observations over the entire study period ranged from 79,133 for the beta-blocker measure to 1,035,946 for the breast-cancer-screening measure.
Study Variables
Our dependent variables were the receipt of each HEDIS indicator (Table 1) by eligible enrollees. Our chief independent variable was black or white race, and these designations are highly accurate in Medicare enrollment data.17 Covariates included age, sex, enrollment in Medicaid, the percentage of persons 65 years of age or older within the enrollee's ZIP Code with an income of less than the federal poverty level, the percentage of persons 65 years of age or older in the enrollee's ZIP Code who had attended college, and urban residence. Data on poverty, educational level, and urban residence within a particular ZIP Code were obtained from the 2000 U.S. Census.
Statistical Analysis
We assessed demographic and socioeconomic characteristics of the population that was eligible for each HEDIS measure. For white enrollees and black enrollees in each year, we calculated the performance for each HEDIS measure as the percentage of eligible enrollees who were reported to have achieved the performance measure.
To determine adjusted rate differences between white enrollees and black enrollees, we fitted separate linear models predicting receipt of each HEDIS measure to each year's data. To assess trends, we fitted models to combined data from the first and last usable year for each measure. We assessed the overall trend in the quality of care by testing the significance of the year effect in the model without a race-by-year interaction. To assess changes in racial disparity on the risk-difference scale, we tested the significance of a race-by-year interaction term.
In order to determine the adjusted effect of variables regarding demographic characteristics, health plan, and socioeconomic factors on racial disparities in the quality of care, we fitted three versions of each of these linear regression models. Adjusted rates and differences then corresponded to those predicted at the mean values of the adjuster variables. The first model adjusted for age and sex. The second model added to the first model variables for rural residence and health plan (defined as a Medicare managed-care contract). Because each contract is typically limited to a specific state (except for a few health plans that serve contiguous areas in adjacent states) and the addition of variables by state did not significantly alter regression results after controlling for health plan, we did not include further geographic controls. The third model added to the second model variables for Medicaid eligibility and ZIP-Code-level variables for income and education. All analyses were performed using SAS statistical software (version 9.1) and are reported with two-tailed P values. Our study protocol was approved by the Human Studies Committee of Harvard Medical School and the CMS Privacy Board.
Results
The demographic and socioeconomic characteristics of the enrollees who were eligible for each of the four categories of HEDIS measures during the initial and final years of measurement are shown in Table 2. As compared with the proportion of white enrollees, a higher proportion of black enrollees was female, eligible for Medicaid, and living in urban areas that had higher rates of poverty and lower rates of educational attainment. Black enrollees made up a higher proportion of the sample for diabetes-related measures — a finding that probably reflects the higher prevalence of this disease among blacks. The demographic characteristics of white enrollees and black enrollees were very stable during the seven-year study period, except for an increase in the percentage of Medicaid recipients and a decrease in the percentage of white women in urban areas who were eligible for the breast-cancer-screening measure.
|
|
Table 4 summarizes the results of the multivariable models. Adjustment of the HEDIS performance rates for age and sex (model 1) had little effect on estimated disparities. Additional adjustment for the enrollee's health plan and rural residence (model 2) reduced the disparities between blacks and whites in the initial and final year for six of the nine HEDIS measures and rendered the race-by-year interaction for control of glycosylated hemoglobin levels no longer statistically significant. Additional adjustment for the socioeconomic indicators of Medicaid coverage and residence in high-poverty and low-education areas (model 3) further reduced the magnitude of disparities between blacks and whites in both the initial and the final year. In all models, however, the decrease between the initial and the final year in the magnitude of disparities remained statistically significant for seven of the nine HEDIS measures we studied (P<0.01 for all race-by-year interaction terms).
|
|
In this time-trend analysis of nine clinical performance measures for enrollees in Medicare managed-care plans from 1997 to 2003, quality of care improved on all nine measures and was accompanied by a significant reduction in the disparities between blacks and whites on seven of the measures. Both trends were substantial and were not explained by changes in the sociodemographic characteristics of enrollees or in the health plans that participated in the Medicare managed-care program during the study years. In contrast, racial disparities did not decrease over time for two HEDIS measures assessing clinical outcomes for diabetes and heart disease.
An adjustment for rural residence and health plan narrowed the observed magnitude of racial disparities for most HEDIS measures — a finding suggesting that part of the racial disparity was related to the disproportionate enrollment of black beneficiaries in health plans or in regions with lower performance on these measures. Even the adjusted models, however, showed decreased racial disparities over time.
In spite of observed improvements, performance as measured by HEDIS indicators approached or exceeded 90 percent on only three measures (testing of glycosylated hemoglobin and LDL cholesterol for patients with diabetes and the frequency of prescribing beta-blockers for patients with cardiovascular disorders). On the other six measures, performance was less than 82 percent for both white enrollees and black enrollees. For these important clinical services, gaps between actual and optimal care remained substantial.2
Although racial disparities decreased to 2 percent or less for five of the six process measures, disparities remained at 7 percent or greater for the three measures assessing clinical outcomes (control of LDL cholesterol for enrollees with either diabetes or heart disease and control of glycosylated hemoglobin) in the most recent study years. Although we controlled for socioeconomic variables, the financial burden of the use of lipid-lowering and glucose-lowering medications may have contributed to the greater disparity we observed on these outcome measures, which often require sustained therapy in addition to intermittent testing.18,19
Our findings are consistent with the proposition that improvements in the quality of care are associated with reductions in racial disparities.9 By increasing the consistency of the delivery of care, interventions such as the use of reminder systems, disease management programs, and feedback to health care providers may decrease variation on the basis of nonclinical factors such as race.20,21,22,23,24 A greater awareness among beneficiaries or their health care providers about appropriate services for breast-cancer screening, diabetes control, and cardiovascular care could also explain our results.
We believe that the observed declines in racial disparities were unlikely to have resulted from specific health plan programs tailored to improve care for black enrollees. The representatives of nearly half of the health plans who responded to a recent survey did not collect data regarding race and ethnic background of enrollees.25 In addition, efforts by health plans to develop programs to eliminate disparities in the quality of care on the basis of racial and ethnic factors are relatively recent.26 Since late 2003, the CMS has provided data regarding enrollees' race and ethnic background to participating health plans and has required that they conduct at least one project to reduce disparities.27 However, the reduction in disparities we observed largely preceded these efforts.
The strengths of this study were the inclusion of a large, nationally representative sample of enrollees and the use of quality measures that have been audited and publicly reported by health plans for several years. Since all health plans participating in Medicare were required to report data regarding the quality of care, we avoided the selection bias associated with voluntary reporting programs.28 We were able to adjust for health plan effects and several measures of socioeconomic position that may have confounded or mediated the relationship between race and the quality of care. By limiting our primary analysis to plans with five or more consecutive years of participation in Medicare, we addressed the possibility that changes in the quality of care or reductions in disparities might be an artifact of health plans' selectively entering or exiting the Medicare program.
Our study had several limitations. It was not designed to address the factors that may have caused the observed results or to determine whether similar trends would have been observed for aspects of the quality of care beyond those assessed by the public reporting of HEDIS measures. Furthermore, patients in Medicare fee-for-service and non-Medicare settings were not included in the HEDIS data set. Previous studies have shown similar racial disparities in fee-for-service and managed-care settings,7,29 but whether our finding of decreasing racial disparities over time extends beyond Medicare managed care remains an open question.
Because enrollment data for Medicare did not reliably identify enrollees who were Hispanic, Asian, or Native American during our study years,17 we chose not to analyze trends among these ethnic and racial groups. Such studies are clearly needed. The data also lacked detailed clinical information to provide risk-adjusted outcome measures. Although unmeasured clinical factors might partially explain cross-sectional differences in outcome measures by race, such factors would be less likely to explain changes in racial disparities over time.
Several studies have suggested that racial differences in the quality and outcomes of care may be related to differences in the site of care between white and minority patients.30,31 Although we were able to analyze the contribution of health plans to racial variation in the quality of care, the current HEDIS reporting protocol does not collect information on providers and practices within plans. In addition, the attitudes of patients about health in general or about their ability to modify their diet or physical activity32,33 may contribute to racial disparities in clinical outcomes, but these variables were not available in our analysis.
Our findings have two important policy implications. For policymakers, health plan leaders, purchasers, and providers who are concerned with improving the quality of care and with reducing disparities, appropriate data are essential to gauge progress on each of these objectives. Measures of quality should be stratified by race, ethnic background, and socioeconomic position — an approach that is now rarely possible with publicly reported data on the quality of care.9,34,35 Second, although racial disparities decreased on some measures of quality, interventions that are focused on black enrollees or their health care providers may still be necessary to eliminate the disparities that remain.
In summary, improvements in the quality of care among enrollees in Medicare managed-care plans since 1997 have been accompanied by reduced racial disparities in most, but not all, measures of clinical performance we studied. Effective collaborative efforts by policymakers, health-plan administrators, clinicians, and patients may be needed to eliminate these disparities entirely.
Supported by an institutional National Research Service Award (5 T32 HP11001-15) from the Health Resources and Services Administration, by the Primary Care Research Fund of Brigham and Women's Hospital, and by a grant (P01-HS10803) from the Agency for Healthcare Research and Quality.
We are indebted to Robert E. Wolf and Matthew Cioffi for statistical and programming contributions; to Kimberly Elmo and Judy Giles for help with data acquisition; and to the Research Data Assistance Center at the University of Minnesota.
Source Information
From the Division of General Medicine and Primary Care, Department of Medicine, Brigham and Women's Hospital (A.N.T., E.C.S., J.Z.A.); the Department of Health Care Policy, Harvard Medical School (A.N.T., A.M.Z., J.Z.A.); and the Department of Health Policy and Management, Harvard School of Public Health (E.C.S., J.Z.A.) — all in Boston.
Address reprint requests to Dr. Ayanian at the Department of Health Care Policy, Harvard Medical School, 180 Longwood Ave., Boston, MA 02115, or at ayanian{at}hcp.med.harvard.edu.
References
1852.
| |||||||||||||||||||||||||||||||||||||||||||||||||||
Related Letters:
Trends in Racial Disparities in Care
Kuller L. H., Freedman B. I., Wagenknecht L. E., Bowden D. W., Keppel K. G., Pearcy J. N., Weissman J. S., Akpunonu B. E., Mutgi A. B., Khuder S. A., Vaccarino V., the National Registry of Myocardial Infarction Investigators , Jha A. K., Epstein A. M., Orav E. J., Trivedi A. N., Zaslavsky A. M., Ayanian J. Z.
Extract |
Full Text |
PDF
N Engl J Med 2005;
353:2081-2085, Nov 10, 2005.
Correspondence
This article has been cited by other articles:
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2008 Massachusetts Medical Society. All rights reserved. |
Previous
