Poor prediction of potentially drug-resistant pathogens using current criteria of health care-associated pneumonia

Background: Health care-associated pneumonia (HCAP) includes a broad range of patients having frequent or chronic contact with health care systems. However, the relationship between current defining criteria for HCAP and the risk of potentially drug-resistant (PDR) pathogens is controversial. Methods: We retrospectively evaluated patients admitted to Severance Hospital in South Korea with culture-positive pneumonia from January 2008 to December 2009. We analyzed the associations between risk factors for HCAP and infection with PDR pathogens, and developed a new scoring system to predict infection with PDR pathogens. Results: Among 339 patients, PDR pathogens were observed in 122 (36.0%). PDR pathogens were more common in HCAP than community-acquired pneumonia (CAP) (48.5% versus 23.8%, P < 0.001). In a logistic regression, prior hospitalization within 90 days of pneumonia (OR = 2.51, P = 0.003), recent treatment with antimicrobials (OR = 2.35, P = 0.039), and nasogastric tube feeding (OR = 15.28, P < 0.001) were independently associated with PDR pathogens. For the prediction of PDR pathogens, the sensitivity and specificity of current HCAP criteria were 66.4% and 60.4%, respectively, and 68.0% and 67.3%, respectively, for the new scoring system. Moreover, the new scoring system showed better diagnostic accuracy than current HCAP criteria (area under curve = 0.711 versus 0.634, P < 0.001). Conclusions: The current HCAP criteria are poor predictors of PDR pathogens and all patients with HCAP should not be empirically treated for these pathogens. To avoid excessive antibiotic use, individual risk stratification approaches should be considered.