|
|
| |||||||||||||||||||||||||||||||||||||||||
Before generalizing from the study by Asch et al., it is therefore important to examine the study's limitations. There was potential selection bias in the sample, which included only 37 percent of those initially eligible for participation in the study. It is difficult to generalize the findings of this study, which included disproportionately few uninsured persons and patients enrolled in Medicaid, to the nation as a whole. Finally, despite their conclusion about race or ethnic background and the quality of care, the authors state that disparities were evident when their analysis was limited to categories in which racial or ethnic disparities had been found in other studies.
Given these limitations and the magnitude of other evidence, the study provides inadequate support for the conclusion that interventions focused on disparities in health care warrant less attention than general quality improvement. Both types of efforts should continue.
H. Jack Geiger, M.D.
City University of New York Medical School
New York, NY 11201
Joseph R. Betancourt, M.D., M.P.H.
Harvard Medical School
Boston, MA 02115
David R. Williams, Ph.D.
University of Michigan
Ann Arbor, MI 48106
References
It is not clear that the RAND criteria of 439 specific recommendations, generated by panels of experts, apply seamlessly to medical care as it is delivered on a daily basis. Are all of the 439 recommendations of equal importance? Is there a general consensus in the medical community about the importance of each measure? If these measures do not vary among groups with markedly differing health outcomes, just how relevant are the criteria?
In the Discussion section, virtually no mention is made of the validity of the medical criteria. The results of this study cast doubt on the relevance of the criteria to medical outcomes.
Robert Sherrick, M.D.
Kalispell Diagnostic Service
Kalispell, MT 59901
References
Quality indicators provide one measure of complexity and clinical work. For example, the care of a patient with diabetes, hypertension, depression, and new headache might be compared with the care of a patient who requires cataract extraction; the former entails up to 56 quality indicators and 2.18 relative-value units (RVUs)1,2; the latter entails 5 quality indicators and 18 RVUs.2 In this example, the incentive index (RVU÷indicator) is nearly 100 times as great for procedural care (3.60) as for cognitive care (0.04).
With reimbursement disparities of this magnitude, should we be surprised at the diminishing workforce in generalism? The number of trainees choosing generalist careers is plummeting.3 Yet a larger supply of generalists is associated with increased quality.4 Analysis of current incentives can shed light on the manner by which the existing U.S. economic model puts all Americans at risk for poor quality of care.
Christine A. Sinsky, M.D.
Medical Associates Clinic and Health Plans
Dubuque, IA 52001
csinsky1{at}mahealthcare.com
References
Geiger et al. and Sherrick bring up important methodologic concerns, and we have addressed them through extensive examination and modeling. We adjusted for nonresponse and tested the sensitivity of our findings to nonresponse bias. Although nonresponse bias explains some of the small differences we found among racial and ethnic groups, nonrespondent blacks would have to have had implausibly low (near zero) overall quality scores to produce differences of the magnitude (approximately 20 percent) often found in the literature on disparities. The criteria we used to measure the quality of care went through a scientifically established process. The evidence linking each measured process to a relevant health outcome was assembled and presented to a group of nationally known experts nominated by their specialty societies. The modified Delphi method is established and validated and has been shown to predict future trial results.1 Limited empirical evidence also supports the relationship of similar measurement sets to observed outcomes.2,3
As Sherrick points out, other studies have documented very different outcomes among groups that had only limited overall differences in process quality in our study. Many factors other than the quality of medical care processes affect health outcomes. These factors include access, environment, disease severity, health habits, and adherence. Undoubtedly, we have a long way to go to reduce disparities in outcomes for all patients. Nonetheless, it is reasonable to focus on validated process measures that are linked to outcomes because they are under the control of providers and managers and therefore directly amenable to quality-improvement efforts.
Sinsky's point that more highly remunerated care may have been delivered at higher rates of indicator compliance is interesting. Our study was not designed to address the relationship between reimbursement and quality, and future research should investigate it.
Steven M. Asch, M.D., M.P.H.
Veterans Affairs Greater Los Angeles Healthcare System
Los Angeles, CA 90073
Eve A. Kerr, M.D., M.P.H.
Veterans Affairs Ann Arbor Healthcare System
Ann Arbor, MI 48105
Elizabeth A. McGlynn, Ph.D.
RAND Health
Santa Monica, CA 90407
References
| |||||||||||||||||||||||||||||||||||||||||
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | TERMS OF USE | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2009 Massachusetts Medical Society. All rights reserved. |
Previous
