Background Discriminating Mycoplasma pneumoniae (MP) from Streptococcus pneumoniae (SP) and viral etiologies of community-acquired pneumonia (CAP) is challenging but has important implications regarding empiric antibiotic therapy. We investigated patient parameters upon hospital admission to predict MP infection. Methods All patients hospitalized in a tertiary care hospital between 2013 and 2017 for CAP with a confirmed etiology were analyzed using logistic regression analyses and area under the receiver operator characteristics (ROC) curves (AUC) for associations between demographic, clinical and laboratory features and the causative pathogen. Results We analyzed 568 patients with CAP, including 47 (8%) with MP; 152 (27%) with SP and 369 (65%) with influenza or other viruses. Comparing MP and SP by multivariate logistic regression analysis, younger age (odds ration [OR] 0.56 per 10 years, 95% CI 0.42-0.73), a lower neutrophil/lymphocyte ratio (OR 0.9, 0.82-0.99) and an elevated C-reactive protein/procalcitonin (CRP/PCT) ratio (OR 15.04 [5.23-43.26] for a 400 mg/μg cut-off) independently predicted MP. With a ROC curve AUC of 0.91 (0.80 for the >400 mg/μg cutoff), the CRP/PCT ratio was the strongest predictor of MP vs. SP. The discriminatory value resulted from significantly lower PCT values (p < 0.001) for MP, while CRP was high in both groups (p = 0.057). Comparing MP and viral infections showed similar results with again the CRP/PCT ratio providing the best information (AUC 0.83; OR 5.55 for the >400 mg/μg cutoff, 2.26-13.64). Conclusions In patients hospitalized with CAP, a high admission CRP/PCT ratio predicts M. pneumoniae infection and may improve empiric management.
Keywords: M. pneumoniae; biomarker; community acquired pneumonia; procalcitonin; respiratory infections.