Printer Friendly

Correspondence: to the editor.

Krakauer et al.(1) provided an evaluation of HCFA's Medicare hospital mortality report that is important and timely. Krakauer et al. assessed HCFA's model by measuring it against an alternative model derived from detailed clinical data (their "gold standard"). They reported that when risk-adjusted mortality rates were computed using first the HCFA model and then the clinical model, the two sets of results were similar (rank order correlation, .88).(1) The authors interpreted this finding as evidence of the HCFA model's validity.

However, evidence presented within the article, but not discussed by the authors, raises some doubts about the validity of the particular model that was used as the gold standard. Certain risk factors included in the model (e.g., chronic pulmonary disease, malignant tumor, history of angina/myocardial infarction, history of asthma, and hypothermia), which would be assumed, on clinical grounds, to increase the risk of death, had negative coefficients. It is difficult to accept that the presence of these conditions would improve a patient's prognosis.

The negative coefficients are more likely due to estimation problems with the model. We speculate that the problem may lie in the incomplete framework used for classifying principal diagnoses. Table 1 of Krakauer et al. reveals only 46 percent of patients fell into specific disease groups with the remainder grouped together in an all-other category. With principal diagnosis underspecified, counterintuitive negative coefficients could mean that some serious health deficits are serving as proxies for chronic as opposed to acute illnesses among the unclassified patient group. HCFA avoids this problem by subdividing patients into 17 exhaustive and mutually exclusive analytic diagnostic categories and estimating a distinct model for each such diagnostic group.

Signs that the clinical model may have been misspecified raise questions about the interpretation of the strong correlation between the two sets of rankings. It is possible that neither the HCFA nor the clinical model possesses sufficient predictive power to explain much of the variation in hospital mortality. In that case, the sequence in which hospitals were ranked by either model would be similar to the order of the hospitals' observed (unadjusted) mortality rates. To find out, we used data from HCFA's mortality report to generate rank order correlations between hospital-specific mortality residuals and observed mortality rates for hospitals (N = 5,414). Like Krakauer et al., we excluded hospitals with fewer than 50 cases or more than 45 percent deaths. The resulting rank order correlation was .75, indicating considerable similarity between the HCFA model's rankings and rankings based on mortality rates that had not been adjusted at all.

Several studies have shown that the HCFA model suffers from some serious limitations.(2,3,4) The fact that a model constructed from MedisGroups clinical values produces results that differ little from a model that employs only the very limited information contained in claims data is not very reassuring about the usefulness of data derived from MedisGroups to rank hospitals in terms of quality of care.

Jesse Green, Ph.D. Neil Wintfeld, Ph.D. Health Policy Research New York University

1. Krakauer, H., R. C. Bailey, K. J. Skellan, et al. "Evaluation of the HCFA Model for the Analysis of Mortality Following Hospitalization." Health Services Research 27, no. 3 (1992): 317-35.

2. Green, J., N. Wintfeld, P. Sharkey, and L. J. Passman. "The Importance of Severity of Illness in Assessing Hospital Mortality." Journal of the American Medical Association 263 (1990): 241-46.

3. Green, J., L. J. Passman, and N. Wintfeld. "Analyzing Hospital Mortality: The Consequences of Diversity in Patient Mix." Journal of the American Medical Association 265 (1991): 1849-53.

4. Park, R. E., R. H. Brook, J. Kosecoff, et al. "Explaining Variations in Hospital Death Rates: Randomness, Severity of Illness, Quality of Care." Journal of the American Medical Association 264 (1990): 484-90.
COPYRIGHT 1993 Health Research and Educational Trust
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1993 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:response to H. Krakauer, R.C. Bailey and K.J. Skellan, Health Services Research, vol. 27, p. 317, 1992
Author:Green, Jesse; Wintfeld, Neil
Publication:Health Services Research
Date:Aug 1, 1993
Previous Article:Hospital board effectiveness: relationships between governing board composition and hospital financial viability.
Next Article:Correspondence: the authors reply.

Terms of use | Copyright © 2017 Farlex, Inc. | Feedback | For webmasters