Validation of a Prediction Score for Drug-Resistant Microorganisms in Community-acquired Pneumonia

Rationale: Recommended initial empiric antimicrobial treatment covers the most common bacterial pathogens; however, community-acquired pneumonia (CAP) may be caused by microorganisms not targeted by this treatment. Developed in 2015, the PES (Pseudonzonas aeruginosa, extended-spectrum beta-lactamase-producing Enterobacteriaceae, and methicillin-resistant Staphylococcus aureus) score was developed in 2015 to predict the microbiological etiology of CAP caused by PES microorganisms. Objective: To validate the usefulness of the PES score for predicting PES microorganisms in two cohorts of patients with CAP from Valencia and Mataro. Methods: We analyzed two prospective observational cohorts of patients with CAP from Valencia and Mataro. Patients in the Mataro cohort were all admitted to an intensive care unit (ICU). Results: Of the 1,024 patients in the Valencia cohort, 505 (51%) had a microbiological etiology and 31 (6%) had a PES microorganism isolated. The area under the receiver operating characteristic curve was 0.81 (95% confidence interval [95% CI], 0.74-0.88). For a PES score >= 5, sensitivity, specificity, the negative and positive predictive values as well as the negative and positive likelihood ratios were 72%, 74%, 98%, 14%, 0.38, and 2.75, respectively. Of the 299 patients in the Mataro cohort, 213 (71%) had a microbiological etiology and 11 (5%) had a PES microorganism isolated. The area under the receiver operating characteristic curve was 0.73 (95% CI 0.61-0.86). For a PES score >= 5, sensitivity, specificity, the negative and positive predictive values, and the negative and positive likelihood ratios were 36%, 83%, 96%, 11%, 0.77, and 2.09, respectively. The best cutoff for patients admitted to the ICU was 4 points, which improved sensitivity to 86%. The hypothetical application of the PPS score showed high rates of overtreatment in both cohorts (26% and 35%, respectively) and similar rates of undertreatment. Conclusions: The PES score showed good accuracy in predicting the risk for microorganisms that required different empirical therapy; however, its use as a single strategy for detecting noncore pathogens could lead to high rates of overtreatment. Given its high negative predictive value, the PES score may be used as a first step of a wider strategy that includes subsequent advanced diagnostic tests.