Performance Indicators for Pneumonia in the Elderly
Performance Indicators for Pneumonia in the Elderly
Abstract & Commentary
Synopsis: Length of stay was a useful performance indicator with regard to the treatment of pneumonia, but mortality was not.
Source: Hand R, et al. Mortality and length-of-stay as performance indicators for pneumonia in the elderly. J Investig Med 1997;45:183-190.
The medical world is going crazy about community-acquired pneumonia. It is as though it were an emerging disease instead of one that has been with us since hominids evolved. Why the clamor? For one, new and resistant pathogens have actually emerged in the last decade including hantavirus and penicillin-resistant, pneumococci-producing clinical dilemmas afresh. New antimicrobials including fancy macrolides and second-generation fluoroquinolones have been developed, and pharmaceutical companies are eager to delineate a clinical niche for them to occupy.
Disease management enterprises charged with quality measurements and quality assurance of medical care also value pneumonia as a diagnosis ideally suited for the study of performance indicators. A group of investigators at the University of Illinois was curious if they could manipulate an HCFA database for pneumonia in the Chicago area to determine performance indicators that affect outcome for patients requiring admission for that diagnosis. They chose one pneumonia diagnosis related group (DRG) for study and used all discharges or deaths at 20 area hospitals during the period 1988-1992.
Figure
Scatter Plots of Length of Stay and Mortality
Scatter plots of the 1991 and 1992 length of stay (days) and mortality (rate) for 20 Illinois hospitals. The plots illustrate the correlation of length of stay and lack of correlation of mortality at the hospitals between the two years. This is reflected in the significant correlation coefficient of 0.766 (P < 0.00005) for length of stay and the nonsignificant correlation coefficient of 0.301 (P = 0.0986) for mortality.
Reprinted with permission from: Hand R, et al. J Investig Med 1997;45:183-190.
One hypothesis of the work was that if the study could not account for differences in mortality among study hospitals, then, they argue, that variation was indeed random. The reader may recognize the hint of an industrial approach to quality assessment that surfaced in earlier works of Shewhart and Deming that addressed the commercial marketplace.
If the Results section is a foreshadow of such studies to come, then the average reader will struggle amidst statistical machinations. Let’s consider the simple findings first.
More than 16,000 patients with DRG89 surfaced from the 20 hospitals during the four-year study at about 4000 per year. Discharges remained stable from hospital to hospital during the study period.
There were trends to shorter lengths of stay and lower mortality with coefficients for lengths-of-stay reaching statistical significance. Scatter plots allow visualization that the lengths of stay showed correlation from year to year but that the scatter was diffuse for mortality and therefore random (e.g., if a mortality rate at one hospital was 17.6% for 1992, but only 12.2% in 1991.) (See Figure.) Other hospitals showed such a variation suggesting that the higher mortality rates were a result of random variation. Because of this randomness, the differences in mortality from year to year at a given hospital are, in the authors’ words, "meaningless."
Using a method called a power calculation, the power for coefficient and the power for rank coefficient was highly significant for length of stay but not significant for mortality.
COMMENT BY JOSEPH F. JOHN, MD, FACP
The value of this study lies in the model it advances for examining those performance indicators in specific medical and surgical conditions that can be systematically reproduced. It is useful to determine those indicators whose variation is random. Such indicators, like mortality in the present study, even if weighted for severity, are unlikely to stabilize an aggregated performance indicator. On the other hand, indicators whose power for coefficient is high (30.90) can be reliably used to analyze performance. Thus, in this study, increased lengths of stay would reflect poorly on hospital care of community-acquired pneumonia.
Peer review organizations and other organizations have traditionally focused on mortality as an outcome indication. Just today in a local paper, I read that a certain health system offered lower mortality if you used their hospitals for coronary artery bypass surgery. Other performance indicators, like length of stay, were not mentioned.
A tempting practical application of this paper is to run back to your office of quality management and determine if the length of stay for community-acquired pneumonia DRG 89 at your hospital is below eight days. If so, you would compare well with the Chicago hospitals studied. It will be important to determine if conditions other than pneumonia should employ mortality as a performance indicator.
Subscribe Now for Access
You have reached your article limit for the month. We hope you found our articles both enjoyable and insightful. For information on new subscriptions, product trials, alternative billing arrangements or group and site discounts please call 800-688-2421. We look forward to having you as a long-term member of the Relias Media community.