Validation of a diagnostic score model for the prediction of Legionella pneumophila pneumonia

Background: Community-acquired pneumonia (CAP) due to Legionella has a high mortality rate in patients who do not receive adequate antibiotic therapy. In a previous study, we developed a simple Legionella Score to distinguish patients with Legionella and non-Legionella pneumonia based on clinical information at diagnosis. In the present study, we validated this Legionella Score for the presumptive diagnosis of Legionella CAP. Methods: This validation cohort included 109 patients with Legionella CAP and 683 patients with non-Legionella CAP. The Legionella Score includes six parameters by assigning one point for each of the following items: being male, absence of cough, dyspnea, C-reactive protein (CRP) >= 18 mg/dL, lactate dehydrogenase (LDH) >= 260 U/L, and sodium < 134 mmol/L. Results: When the Legionella CAP and non-Legionella CAP were compared by univariate analysis, most of the evaluated symptoms and laboratory test results differed substantially. The six parameters that were used for the Legionella Score also indicated clear differences between the Legionella and non-Legionella CAP. All Legionella patients had a score of 2 points or higher. The median Legionella Scores were 4 in the Legionella CAP cases and 2 in the non-Legionella CAP cases. A receiver operating characteristics curve showed that the area under the curve was 0.93. The proposed best cutoff, total score >= 3, had sensitivity of 93% and specificity of 75%. Conclusion: Our Legionella Score was shown to have good diagnostic ability with a positive likelihood of 3.7 and a negative likelihood of 0.10. (c) 2019 Japanese Society of Chemotherapy and The Japanese Association for Infectious Diseases. Published by Elsevier Ltd. All rights reserved.