Respiratory Infections in Nursing Homes
February 28, 2002 4 minutes read
Respiratory Infections in Nursing Homes
Abstract & Commentary
Synopsis: Eight variables predicted mortality from lower respiratory tract infections in nursing home residents, and these variables can be used to make treatment decisions.
Source: Mehr DR, et al. JAMA. 2001;286:2427-2436.
A community-based study using 36 Missouri nursing homes identified 4959 residents with lower respiratory tract infections (LRI) over 3 years from 1995-1998. These facilities were chosen to reflect national characteristics in ownership, size, and location. The definition of LRI was derived from those used in nursing home surveillance,1 and included not only chest x-ray-proven pneumonias, but also at least 2-3 other significant symptoms such as cough, sputum production, fever, new pulmonary physical findings, etc. Mehr and colleagues felt this would more accurately represent the conditions most physicians are faced with in this setting.
With funding from the Agency for Healthcare Research and Quality and other sources, the Center for Family Medicine Science at the University of Missouri, Columbia, coordinated this prospective study involving multiple collaborators and institutions. Project nurses visited each facility nearly daily to collect data and followed patients thorough hospitalizations. After eliminating cases that refused participation, were transferred, or did not meet the definition of LRI, 1406 cases were left for evaluation in 1044 residents, and 30-day mortality from all causes was obtained. Abstraction of clinical and radiograph information was doubly reviewed by independent clinicians. Initially, 25 categories of variables were identified, which might predict mortality and were further analyzed using a variety of statistical methods.
Although a large number of variables were associated with 30-day mortality, 8 were identified that could be used to construct a predictive model useful in the clinical setting. Logistic models were then used to develop a risk score with points for various levels of the variables. For example, 0-2 points were given for WBC count; > 24 × 103 received 2 points, while levels 14.1-24 received 1 point. Serum BUN elevations went up to 6 points, giving it the most weight, with the next most important factor (allocated 4 points) being body mass index, with less than 13 kg/m2 predicting the most risk.
Pneumonia on chest x-ray was not a significant predictor of mortality in this model. Because Mehr et al wanted to mimic the actual conditions that clinicians work with in nursing homes, data were collected from x-ray reports rather than review of radiographs, and many of them were done portably and of uncertain value. Increased respiratory rate also was not a good predictor, although it had been found to be in previous studies.
Using their risk-scoring model, they compared predicted and observed mortality from LRI in nursing home residents and found remarkable agreement in nearly all cases. Predicted mortality ranged from 2% in the lowest risk cases to more than 50% in the highest risk cases.
Comment by Mary Elina Ferris, MD
This large study gives interesting and useful results from nursing home surveillance for LRI over a 3-year period. Although previous predictive models have been published and used to predict mortality in community-acquired pneumonia,2 that model attaches increased risk with advancing age and would suggest that hospitalization is indicated for almost all nursing home pneumonias. Indeed, Mehr et al note that if the older model had been applied to their cases, fully 85% would have fallen into the highest risk categories, whereas in reality only 48% of their cases were both predicted and observed to fall into high risk. Furthermore, 30% of their own hospitalizations would have been predicted to be low risk using their new model, and potentially great savings could have been realized in this early identification.
This study demonstrates that medical assumptions about community pneumonia are not necessarily applicable to the special population residing in nursing homes. Of particular interest was the lack of usefulness of the chest x-ray in predicting mortality, but as Mehr et al note, their wider definition of LRI to include bronchitis and other significant respiratory symptoms, as well as the quality of portable x-rays, may have rendered the chest x-ray less useful.
Wide variation in medical decisions regarding hospitalization of nursing home patients has been found in the past,3 perhaps not surprising considering how little evidence there has been to guide physicians and families. Unnecessary hospitalizations can potentially be avoided if greater discrimination is used with these newly identified 8 variables, or at least better predictions made when families ask for prognosis.
Dr. Ferris, Clinical Associate Professor, University of Southern California, is Associate Editor of Internal Medicine Alert.
References
1. McGeer A, et al. Am J Infect Control. 1991;19:1-7.
2. Fine MJ, et al. N Engl J Med. 1997;336:243-250.
3. Saliba D, et al. J Am Geriatr Soc. 2000;48:154-163.
You have reached your article limit for the month. Subscribe now to access this article plus other member-only content.
- Award-winning Medical Content
- Latest Advances & Development in Medicine
- Unbiased Content