Ruling out Legionella in community-acquired pneumonia

Abstract

Background: Assessing the likelihood for Legionella sp. in community-acquired pneumonia is important because of differences in treatment regimens. Currently used antigen tests and culture have limited sensitivity with important time delays, making empirical broad-spectrum coverage necessary. Therefore, a score with 6 variables recently has been proposed. We sought to validate these parameters in an independent cohort.

Methods: We analyzed adult patients with community-acquired pneumonia from a large multinational database (Community Acquired Pneumonia Organization) who were treated between 2001 and 2012 with more than 4 of the 6 prespecified clinical variables available. Association and discrimination were assessed using logistic regression analysis and area under the curve (AUC).

Results: Of 1939 included patients, the infectious cause was known in 594 (28.9%), including Streptococcus pneumoniae in 264 (13.6%) and Legionella sp. in 37 (1.9%). The proposed clinical predictors fever, cough, hyponatremia, lactate dehydrogenase, C-reactive protein, and platelet count were all associated or tended to be associated with Legionella cause. A logistic regression analysis including all these predictors showed excellent discrimination with an AUC of 0.91 (95% confidence interval, 0.87-0.94). The original dichotomized score showed good discrimination (AUC, 0.73; 95% confidence interval, 0.65-0.81) and a high negative predictive value of 99% for patients with less than 2 parameters present.

Conclusions: With the use of a large independent patient sample from an international database, this analysis validates previously proposed clinical variables to accurately rule out Legionella sp., which may help to optimize initial empiric therapy.

Keywords: Legionella; Pneumonia; Prediction algorithm.