Multicentre validation and update of a Legionella prediction score to guide microbiological testing in community-acquired pneumonia.

Abstract

With notification rates for Legionnaires' disease (LD) rising in Europe and the United States, accurate case detection becomes increasingly important for individual patient management and public health surveillance. A six-predictor score has been proposed to identify community-acquired pneumonia (CAP) patients at increased risk of LD. We externally validated and updated this score to improve its practical usability. We analysed data from 196 patients with LD across 20 Swiss hospitals and 196 matched Legionella test-negative CAP controls (August 2022-March 2024). We evaluated the availability of the six original score predictors (fever, no/dry cough, hyponatremia, elevated CRP, elevated LDH, low platelet count) in routine care and assessed the score's discriminative performance. The dataset was split into development and validation cohorts to assess if model simplicity and predictive performance could be improved. The original score yielded 91% sensitivity (95% CI: 86-96%) and 35% (95% CI: 28-42%) specificity at a cut-off ≥2. LDH was infrequently measured, and platelet count was a poor predictor. The simplified SwissLEGIO score (fever >38°C, sodium <133 mmol/L, CRP >180 mg/L, no/dry cough, prior β-lactam therapy) showed high sensitivity (88-92%) and improved specificity (46-58%) at a cut-off ≥2. The SwissLEGIO score is an easy-to-apply screening tool to rule out LD in hospitalised CAP patients. A scores <2 could reduce Legionella-specific testing by 36-52% (at a disease prevalence of 4%) while maintaining high sensitivity for case detection.