Objective: The objective of this study was to compare the observed rates of ventilator-associated pneumonia when using the National Healthcare Safety Network vs. the American College of Chest Physicians criteria.
Design: Prospective, observational cohort study.
Setting: A 1250-bed academic tertiary care medical center.
Patients: Adult medical and surgical intensive care unit patients requiring mechanical ventilation for >48 hrs.
Measurements and main results: Patients were prospectively and independently screened for ventilator-associated pneumonia from January 2009 to January 2010 using the National Healthcare Safety Network and American College of Chest Physicians criteria. All American College of Chest Physicians classifications, including the corresponding radiographs and laboratory data, were prospectively reviewed by one of the investigators (JD) and confirmed by a second investigator (MHK). All National Healthcare Safety Network classifications were administratively determined using the hospital's infection prevention surveillance system. Over 1 yr, 2060 patients met the inclusion criteria. Of these, 83 patients (4%) had ventilator-associated pneumonia according to the American College of Chest Physicians criteria as compared with 12 patients (0.6%) using the National Healthcare Safety Network criteria. The corresponding rates of ventilator-associated pneumonia were 8.5 vs. 1.2 cases per 1,000 ventilator days, respectively. Agreement of the two sets of criteria was marginal (κ statistic, 0.26). Cultures were positive in 88% of ventilator-associated pneumonias in the American College of Chest Physicians group as compared to 92% in the National Healthcare Safety Network group.
Conclusions: There is poor agreement between clinical and administrative surveillance methods for the diagnosis of ventilator-associated pneumonia. Although there may be some benefit to using more stringent criteria for surveillance of ventilator-associated pneumonia, use of the administratively applied National Healthcare Safety Network criteria may significantly underestimate the scope of the clinical problem.