National nosocomial infection surveillance system: from benchmark to bedside in trauma patients.

Introduction: Ventilator-associated pneumonia (VAP) is an important cause of morbidity and mortality in the injured patient. Identification of those with VAP is important both in immediate clinical decision making as well as for the epidemiologic evaluation of the disease and benchmarking of rates across institutions with variable practice patterns. Despite this, controversy exists over the optimal method of VAP diagnosis. Many centers currently use invasive culture methods such as bronchoalveolar lavage (BAL) for diagnosis. Another diagnostic method, and the most common epidemiologic tool used to track VAP, is the definition employed by the National Nosocomial Infections Surveillance (NNIS) system. This relies on a combination of clinical and culture data. Our goal was to evaluate the accuracy of the NNIS definition as compared with BAL diagnosis in trauma patients.

Methods: Records of all ventilated patients admitted to the trauma intensive care unit at a Level I center who were evaluated for the presence of pneumonia over a 2.5-year period were reviewed. VAP diagnosis was established if > or =10 cfu/mL were cultured on BAL. VAP rates and time of onset were compared with the hospital infection control database, which defines VAP by NNIS criteria. Assuming BAL to be correct, sensitivity, specificity, and positive and negative predictive values were calculated for NNIS VAP.

Results: From September 1, 2001, through December 31, 2003, 292 patients underwent BAL for suspected pneumonia. The pneumonia rate in this group was 34 per 1,000 ventilator days. The NNIS definition showed excellent overall agreement, with a rate of 36 per 1,000 ventilator days. The use of the NNIS definition for bedside decision making, however, is less accurate. Sensitivity and positive predictive value were reasonably good (84% and 83%, respectively), whereas specificity and negative predictive value suffer (69% and 69%, respectively). Most importantly, the use of NNIS would have led to no treatment in 16% of patients diagnosed with VAP by BAL.

Conclusions: Compared with strict bacteriologic criteria for VAP, the NNIS definition has good overall agreement and seems to have utility as an epidemiologic benchmarking tool in trauma patients. However, the NNIS definition has less utility as a bedside decision-making tool in this population, leading to under-treatment in a significant number of patients.