Correspondence
N Engl J Med 1997; 337:1851-1852December 18, 1997DOI: 10.1056/NEJM199712183372517
We have major questions about the methods and conclusions of Morgan et al. (July 17 issue) 1 in their study of the Medicare–health maintenance organization (HMO) revolving door. The authors admit that the study area, southern Florida, is unique, having “more Medicare beneficiaries than each of 35 states, one of the nation's highest per capita health care costs, and the nation's second highest Medicare capitated rate.” Nevertheless, the authors generalize their results to the whole nation without reservation or qualification — a serious leap of faith and lapse of judgment.
In addition, the authors excluded Medicare recipients who were continuously enrolled in HMOs during the study period. Without taking this population into account, no accurate conclusions can be made about whether HMOs save Medicare money.
A vitally important point is that there is no evidence that hospitalization rates were calculated with the use of equivalent time periods for the different groups or that the rates were annualized. The results are without value if different time periods were not appropriately handled. Also, the focus on inpatient admissions as an overall utilization measure is weak. Outpatient care represents almost half of Medicare's fee-for-service costs.2
Important changes have occurred in the population enrolled in Medicare HMOs since the study's 1993 cutoff date. At Aetna U.S. Healthcare, the addition and expansion of desirable coverage, such as coverage for prescription drugs, have resulted in substantial growth in the Medicare membership. These changes have attracted a population that is very different from the population only three years ago. For example, in 1996 the prevalence of chronic illnesses among the members of Aetna U.S. Healthcare's Medicare HMO was higher than the prevalence two years earlier and higher than that among Medicare beneficiaries overall. This latter difference suggests that HMOs are experiencing adverse selection.3,4
David G. Josephson
James R. Grana, Ph.D.
Nicholas A. Hanchak, M.D.
U.S. Quality Algorithms, Blue Bell, PA 19422
1
Free Full Text | Web of Science | Medline
2
1997 Annual report to Congress. Washington, D.C.: Physician Payment Review Commission, 1997:Table 1-1.
3
National Center for Health Statistics, Adams PF, Marano MA. Current estimates from the National Health Interview Survey, 1994. Vital and health statistics. Series 10. No. 193. Washington, D.C.: Government Printing Office, 1995:83-4.
4
Medline
The authors reply:
To the Editor: Josephson et al. express concern that southern Florida's uniqueness limits the usefulness of our conclusions about Medicare-HMO enrollment and disenrollment. Their position is puzzling, since our results are strikingly consistent with those reported nationally.1 In addition, southern Florida's high rate of enrollment in Medicare HMOs, its mature HMO market, and the dominance of that market by the independent-practice-association model suggest that southern Florida is leading national trends. It does not take a leap of faith to see the relevance of our results to national Medicare policy.
We regret any confusion that Josephson et al. had in interpreting our methods of calculating rates. For the HMO-enrollment and HMO-disenrollment groups, inpatient-admission rates were calculated for the year before enrollment and for the three months after disenrollment, respectively. For comparison, all rates for our fee-for-service beneficiaries were standardized according to time span (i.e., the length of the period) and temporal occurrence (i.e., when the preenrollment or postdisenrollment period occurred during the time frame of the study).
We focused on inpatient admissions because part of the justification for establishing the Medicare-HMO program was its presumed ability to reduce total Medicare inpatient costs. It seems self-evident that an examination of the use of inpatient care with respect to HMO enrollment and disenrollment is entirely relevant to an understanding of whether Medicare is meeting its goal. Our findings strongly suggest that the fee-for-service system is absorbing high-use, and high-cost, beneficiaries, while the HMOs are picking up the relatively low-use, and low-cost, beneficiaries. Furthermore, our data suggest that some beneficiaries may perceive substantial barriers to access within HMOs, a point that seems to be frequently lost in the general discussion of Medicare HMOs. Our exclusion of the continuously enrolled HMO beneficiaries does nothing to negate either point.
Josephson et al. cite changes in the enrollment population at Aetna U.S. Healthcare as evidence that Medicare HMOs are attracting patients with chronic illnesses. If, as they suggest, HMOs are experiencing adverse selection because of enrollment trends, these trends should be examined on a population basis. As we discussed in our article, encounter-level billing data for HMO enrollees are not routinely provided by Medicare HMOs to the Health Care Financing Administration and are not available for analysis. We would welcome the opportunity to merge industry data with data from the Health Care Financing Administration, so that we could track diagnoses, utilization, and content of care as people move in and out of the HMO program.
Robert O. Morgan, Ph.D.
Miami Veterans Affairs Medical Center, Miami, FL 33125
Beth A. Virnig, Ph.D., M.P.H.
Carolee A. DeVito, Ph.D., M.P.H.
University of Miami, Miami Beach, FL 33139
Nancy A. Persily, M.P.H.
George Washington University, Washington, DC 20037