Background: Knowledge of the determinant factors responsible for the presence of antimicrobial-resistant pathogens in severe nursing home-acquired pneumonia (NHAP) is deemed essential for antibiotic selection.
Methods: Data for institutionalized patients with cases of severe pneumonia confirmed by culture of protected bronchoalveolar lavage fluid samples (> or =10(3) cfu/mL) during a 36-month period were analyzed. A classification tree with a sensitivity of 100% was developed using binary recursive partitioning to predict which patients are unlikely to have drug-resistant pathogen (DRP)-related pneumonia.
Results: Of the 88 patients who satisfied the inclusion criteria, 17 had at least 1 DRP recovered from the lower respiratory tract. The predictor variables were the Activity of Daily Living score and previous use of antibiotics. Prospective application of the model in 47 patients over a 24-month period yielded a sensitivity of 100% (95% confidence interval [CI], 71.3%-100%) and a specificity of 69.4% (95% CI, 51.9%-83.6%).
Conclusions: The use of the tree may provide a more rational basis for selecting initial therapy for severe NHAP after it is validated in a large prospective study.