Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
, 115 (6), 428-36

Predicting Death in Patients Hospitalized for Community-Acquired Pneumonia

Affiliations

Predicting Death in Patients Hospitalized for Community-Acquired Pneumonia

B M Farr et al. Ann Intern Med.

Abstract

Objective: To validate a previously reported discriminant rule for predicting mortality in adult patients with primary community-acquired pneumonia and to determine which factors available at hospital admission predict a fatal outcome among such patients.

Design: Historical cohort study.

Setting: University hospital.

Patients: Adults admitted to the hospital for community-acquired pneumonia.

Measurements: Using stepwise logistic regression, we analyzed prognostic factors (data available at admission and recorded in the medical record) that showed a univariate association with mortality. The predictive values of three discriminant rules were measured to validate the results of a previous study.

Main results: Of 245 patients, 20 (8.2%) died. Of 42 prognostic factors identified in previous studies, 8 were associated with mortality, but only a respiratory rate of 30/min or more, a diastolic blood pressure of 60 mm Hg or less, and a blood urea nitrogen of more than 7 mmol/L remained predictive in the multivariate analysis. A discriminant rule composed of these three variables was 70% sensitive and 84% specific in predicting mortality, yielding an overall accuracy of 82%.

Conclusion: Tachypnea, diastolic hypotension, and an elevated blood urea nitrogen were independently associated with death from pneumonia in our study, confirming the value of a previously reported discriminant rule from the British Thoracic Society. This rule may be useful in triage decisions because it identifies high-risk patients who may benefit from special medical attention.

Comment in

Similar articles

See all similar articles

Cited by 38 PubMed Central articles

See all "Cited by" articles

MeSH terms

Substances

LinkOut - more resources

Feedback