Accuracy of a score predicting the presence of an atypical pathogen in hospitalized patients with moderately severe community-acquired pneumonia

Background Atypical pathogens (AP), present in some patients with community-acquired pneumonia (CAP), are intrinsically resistant to betalactam drugs, the mainstay of empirical antibiotic treatment. Adding antibiotic coverage for AP increases the risk of adverse effects and antimicrobial selection pressure, while withholding such coverage may worsen the prognosis if an AP is causative. A clinical model predicting the presence of AP would allow targeting atypical coverage for patients most likely to benefit. Methods This is a secondary analysis of a multicentric randomized controlled trial that included 580 adults patients hospitalized for CAP. A predictive score was built using independent predictive factors for AP identified through multivariate analysis. Accuracy of the score was assessed using area under the receiver operating curve (AUROC), sensitivity, and specificity. Results Prevalence of AP was 5.3%. Age < 75 years (OR 2.7, 95% CI 1.2–6.2), heart failure (OR 2.6, 95% CI 1.1–6.1), absence of chest pain (OR 3.0, 95% CI 1.1–8.2), natremia < 135 mmol/L (OR 3.0, 95% CI 1.4–6.6) and contracting the disease in autumn (OR 2.7, 95% CI 1.3–5.9) were independently associated with AP. A predictive score using these factors had an AUROC of 0.78 (95% CI 0.71–0.85). A score of 0 or 1 (present in 33% of patients) had 100% sensitivity and 35% specificity. Conclusion Use of a score built on easily obtained clinical and laboratory data would allow safe withholding of atypical antibiotic coverage in a significant number of patients, with an expected positive impact on bacterial resistance and drug adverse effects. Trial registration: NCT00818610.