Validation of a scoring tool to predict drug-resistant pathogens in hospitalised pneumonia patients

Abstract

Health care-associated pneumonia (HCAP) affects a heterogeneous group of patients in frequent contact with health care systems. However, HCAP criteria poorly predict infection with drug-resistant (DR) pathogens. To validate our previously reported risk-scoring model (predictive of DR pathogen infection) in patients admitted to hospital with pneumonia. We evaluated 580 patients admitted with culture-positive bacterial pneumonia. We identified risk factors, evaluated the risk-scoring model's capacity to predict infection by DR pathogens and compared the model's diagnostic accuracy with that of current HCAP criteria. DR pathogens were observed in 227/580 patients (39.1%). Of 269 HCAP patients, 153 (56.9%) were infected with DR pathogens. Overtreatment was more common in HCAP than in community-acquired pneumonia (58.7% vs. 41.2%, P < 0.001). Recent hospitalisation, admission from a long-term care facility, recent antibiotic treatment and tube feeding were independently associated with DR pathogens. For pathogen prediction, the risk-scoring model showed better diagnostic accuracy than HCAP criteria (area under receiver operating-characteristic curve = 0.723 vs. 0.673, P < 0.001). According to current HCAP criteria, half of the HCAP patients were treated unnecessarily with broad-spectrum antibiotics. Risk scoring by stratifying risk factors could improve the identification of patients likely to be infected with DR pathogens.