Strategies for prediction of drug-resistant pathogens and empiric antibiotic selection in community-acquired pneumonia

Abstract

Although most patients with community-acquired pneumonia (CAP) are appropriately treated with narrow-spectrum antibiotics, predicting which patients require coverage of drug-resistant pathogens (DRP) remains a challenge. The 2019 American Thoracic Society/Infectious Diseases Society of America CAP guidelines endorse using locally validated prediction models for DRP. Here we review risk factors for DRP and provide a summary of available risk prediction models. Both inadequate initial empiric spectrum as well as unnecessary broad-spectrum antibiotic use are associated with poor outcomes in CAP. Multiple prediction models for DRP-based patient-level risk factors have been published, with some variation in included predictor variables and test performance characteristics. Seven models have been robustly externally validated, and implementation data have been published for two of these models. All models demonstrated better performance than the healthcare-associated pneumonia criteria, with most favoring sensitivity over specificity. We also report validation of the novel, risk factor-based treatment algorithm proposed in the American Thoracic Society/Infectious Diseases Society of America guidelines which strongly favors specificity over sensitivity, especially in nonsevere pneumonia. Using patient-level risk factors to guide the decision whether to prescribe broad-spectrum antibiotics is a rational approach to treatment. Several viable candidate prediction models are available. Hospitals should evaluate the local performance of existing scores before implementing in routine clinical practice.