When should you admit? Tool gives answer

Model identifies risk of death from pneumonia

A new clinical model to assess the risk of death in pneumonia patients can help physicians make treatment decisions and ultimately could save millions of dollars by eliminating unnecessary hospital admissions.1 The model is the first to provide a systematic way to categorize patient risk and to provide guidance to physicians about need for admission, researchers say. (For samples of the algorithm and prediction models, see pp. 69-70.)

"Based on the large number of pneumonia cases that we have on a yearly basis, with over 600,000 hospitalizations, even a small decrease in the hospitalization rate could have a substantial impact on the use of resources," says Michael J. Fine, MD, MSC, associate professor of medicine at the University of Pittsburgh and one of the researchers who developed the instrument. It was developed as a part of the Pneumonia Patient Outcomes Research Team supported by the federal Agency for Health Care Policy and Research in Rockville, MD.

Inpatient care for patients with pneumonia cost more than $4 billion annually. Using the prediction model to detect low-risk patients could reduce hospital admissions by 31% and lead to brief observational hospital stays for another 19% of patients, Fine and his colleagues found.1

The pneumonia prediction model was developed and validated based on data of more than 50,000 inpatients with community-acquired pneumonia and an observational study of 2,287 patients.

Still, Fine cautions that the model needs further testing in clinical settings and that it should be used in conjunction with physician judgment about individual cases. "We don’t view this as a cookbook," says Fine. "There are some factors that aren’t included in our model that also go into the physician’s decision making."

When combined with home visits by nurses, some additional admission and discharge criteria, and an emphasis on physician judgment, the model worked well at the Massachusetts General Hospital in Boston, says Daniel Singer, MD, director of the clinical epidemiology unit in the hospital’s general internal medicine division.

"We found a feasible but very conservative way of implementing the pneumonia severity index," says Singer. "The most important issue is that it should be safe."

Physicians need risk guidance

The prediction model fills a gap in a critical area of physician decision making, says Singer. "The criteria for admission and discharge of pneumonia patients has not been entirely clear," he says.

The prediction model relies heavily on gender and age, giving points for each year — then subtracting 10 points for women and adding 10 for nursing home residents. A man who is 60 or older is likely to score in a higher-risk category even without many complicating factors.

Mass General developed a protocol for patients with pneumonia who came to the emergency department that allowed them to be considered for outpatient treatment if they had a risk score of 90 points or less — which placed them in risk categories I, II, and III.

"The vast majority in I, II, or III could be considered for home care, but there are extenuating circumstances that need to be considered," notes Fine. "Is the patient frail and elderly and living alone? Is the patient an alcohol or drug abuser who may have a problem consistently taking the medicine? Is it a pregnant woman?"

At Mass General, physicians considered these questions and others. "We wanted to err on the side of conservatism," says Singer.

For example, patients who were hypoxic, with a oxygen saturation of less than 90% as measured by pulse oximetry, were not considered for the protocol. Patients taking significant doses of prednisone were considered to be immunosuppressed and not candidates for outpatient treatment.

Physicians reviewed the patients’ social situations, such as inadequate support at home or homelessness. And forms providing information about the model stressed that physician judgment overrides any score, Singer says.

At Mass General, the outpatient treatment included home support. A nurse was available to visit patients one and two days after discharge from the emergency department. The home health nurses faxed reports to the physicians that included the patients’ oxygen saturation as a measure of lung function.

After using the protocol for 11 months, Singer’s group is analyzing the percentage of patients admitted, the percentage with delayed admissions, the resolution of symptoms, return to work, and both physician and patient satisfaction with care.

"There were, in fact, no deaths in the 30-day follow-up period [after treatment], and there was general satisfaction on the part of the physicians," Singer says of his initial findings.

The model is likely most useful in clinics or emergency departments with a large population of patients with pneumonia, says Singer. Mass General admits about 600 patients with community-acquired pneumonia each year, he says.

Hospitals and managed care organizations are likely to adopt the model as a part of efforts to reduce unnecessary hospital admissions, says Fine. But he stresses that it is preliminary and requires further clinical testing.

It should always be used as a guide and not a prescription, he says. "This should be used in conjunction with rather than in place of good physician judgment," he says.

[Editor’s note: For a free copy of the publication Pneumonia: New Prediction Model Proves Promising (AHCPR Pub. No. 97-R031) or the consumer fact sheet, Pneumonia: More Patients May Be Treated at Home (AHCPR Pub. No. 97-R030), contact the Agency for Health Care Policy and Research, (800) 358-9295.

For InstantFAX, call (301) 594-2800 (clinical version-977031, consumer version 977030). On the Internet: www.ahcpr.gov.

Copies of Assessment of the Variation and Outcomes of Pneumonia: Pneumonia Patient Outcomes Research Team Final Report (PB97-117808) are available from the National Technical Information Service, Springfield, VA 22161. Telephone: (703) 487-4650.]


1. Fine MJ, et al. A prediction rule to identify low-risk patients with community-acquired pneumonia. NEJM 1997; 336:243-250.