Dr. Crane
Clinical question: Can select patient demographic data and risk factors help predict Legionella infection better than clinical gestalt or current Infectious Diseases Society of America (IDSA)/American Thoracic Society (ATS) testing guidelines?
Background: Legionella occupies a small fraction of the identified causes of community-acquired pneumonia (around 2%) and is commonly diagnosed via urine antigen testing (UAT) or polymerase chain reaction (PCR) performed on respiratory tract specimens. IDSA/ATS 2019 guidelines suggest testing only patients with severe pneumonia, recent travel, or during local outbreaks.
Study design: A retrospective cohort study
Setting: 177 U.S. hospitals belonging to Premiere Healthcare Database and 12 Cleveland Clinic Health System hospitals
Synopsis: Between 2010 and 2015, 43,070 of 166,689 patients were included, with ICD-9/10 codes for community-acquired pneumonia. Inclusion required patients to have received three days of antibiotics, have an infiltrate on chest X-ray, and have been tested for Legionella. A logistics regression model identified 30-day proximity to Legionella outbreak (OR 3.4), June–October admission (OR 3.4), hyponatremia (OR 3.3), smoking (OR 2.4), and diarrhea (OR 2.0) as positive predictors, while an admission within the prior six-months (OR 0.27) and chronic pulmonary disease (OR 0.49) were negative predictors. Qualifying as severe pneumonia did not correlate with test positivity or negativity. Simulating a model of testing patients with the highest predicted probabilities yielded a 3.73% positivity rate, which was 33.5% above the guidelines-based strategy and 150% higher than actual physician testing. This study may overestimate Legionella cases by including only patients that underwent Legionella testing. Furthermore, the practicality of this model is limited by requiring clinicians to know about recent Legionella outbreaks at their institution. The author’s Legionella risk calculator can be accessed at https://riskcalc.org/Legionella.
Bottom line: A logistic regression model-based risk calculator likely outperforms clinical gestalt and IDSA/ATS guidelines in predicting Legionella infection; though an appropriate threshold for testing remains undetermined and the practicality of this model is limited by requiring clinicians to know about recent Legionella outbreaks at their institution.
Citation: Rothberg MB, et al. A risk model to identify Legionella among patients admitted with community-acquired pneumonia: A retrospective cohort study. J Hosp Med. 2022;17(8):624-32.
Dr. Crane is an assistant professor of medicine at the University of Virginia School of Medicine, Charlottesville, Va.