TOPLINE

A six-predictor score for detecting Legionnaires disease (LD) was simplified to a five-predictor SwissLEGIO score that outperformed the original score in a recent Swiss study of patients with community-acquired pneumonia (CAP).

METHODOLOGY

• Researchers evaluated the performance and usability of a simplified LD prediction score using data from patients with community-acquired LD across 20 Swiss hospitals and matched control patients with CAP who tested negative for Legionella (non-LD CAP) from two Swiss hospitals.
• LD was confirmed in patients with CAP by a positive Legionella culture, a positive urinary antigen test, or polymerase chain reaction from respiratory samples (n = 196; median age, 69 years; 68.4% men); patients with non-LD CAP required at least one documented negative test for Legionella (n = 196; median age, 72 years; 64.3% men).
• The original score allocated one point to each of six predictors: high fever (> 39.4 °C), absence of sputum (no/dry cough), hyponatraemia (sodium levels < 133 mmol/L), elevated C-reactive protein (CRP) levels (> 187 mg/L), elevated lactate dehydrogenase (LDH) levels (> 225 U/L), and low platelet counts (< 171 G/L).
• The researchers split the patients into a development cohort (118 matched pairs) and an internal validation cohort (78 matched pairs) to test whether the simplified score could improve performance.
• The performance of the scores was assessed by calculating sensitivity, specificity, negative predictive value (NPV), and positive predictive value, with predictive values calculated assuming an LD prevalence of 4%.

TAKEAWAY

• In the original score, LDH was the most commonly missing predictor because of not being tested at 11 of the 20 hospitals, which was absent in 57.7% of patients with community-acquired LD and 10.7% of those with non-LD CAP.
• The original score demonstrated moderate-to-good discriminative performance (area under the curve [AUC], 0.78; 95% CI, 0.73-0.82), maintaining a sensitivity of 91.2% (95% CI, 86.2%-96.2%) but a specificity of 35.3% (95% CI, 28.4%-42.1%) at a cutoff ≥ 2.
• In the validation cohort, a modified four-predictor score (CRP levels > 180 mg/L, sodium levels < 133 mmol/L, no/dry cough, and fever > 38.0 °C) demonstrated better predictive performance (AUC, 0.78; 95% CI, 0.70-0.85; sensitivity, 85%; 95% CI, 76.9%-93.1%; specificity, 49.6%; 95% CI, 37.8%-61.3%) than the original score.
• Moreover, adding prior beta-lactam therapy as a fifth predictor created the SwissLEGIO score, with an AUC of 0.77 (95% CI, 0.69-0.85), a sensitivity of 87.9% (95% CI, 79.9%-95.9%), and a specificity of 46.0% (95% CI, 34.0%-57.9%) at a cutoff ≥ 2. A high NPV of 98.9% was reported for ruling out community-acquired LD in patients with a score below 2.

IN PRACTICE

"The SwissLEGIO score can complement microbiological testing by providing clear and pragmatic criteria to decide who should undergo diagnostic testing for Legionella. The score is particularly useful in settings where routine testing is not recommended, providing a more efficient alternative to epidemiological risk factor- or severity-guided testing," the authors wrote.

SOURCE

The study was led by Melina Bigler, Swiss Tropical and Public Health Institute, Basel, Switzerland. It was published online on July 03, 2026, in the International Journal of Infectious Diseases.

LIMITATIONS

The score required external validation beyond the Swiss cohort before guideline adoption; the case-control design (rather than the cohort design) limited direct prevalence estimation, and screening relied mainly on urinary antigen testing, most sensitive for the L pneumophila serogroup, which may have affected generalisability to other Legionella species/serogroups.

DISCLOSURES

This study was funded by the Swiss Federal Office of Public Health. One author disclosed receiving honoraria and travel support and serving on advisory boards for multiple pharmaceutical companies.

This article was created using several editorial tools, including AI, as part of the process. Human editors reviewed this content before publication.