Inadequacy of Algorithmic Ventilator-Associated Pneumonia Diagnosis in Acute Care Surgery

Clinical utility of algorithms to diagnose ventilator-associated pneumonia (VAP) in surgical patients has not been established. We aimed to test the diagnostic accuracy of two established methods to reliably diagnose VAP in acutely ill and injured surgical patients. After institutional review board approval, we prospectively collected data on 508 mechanically ventilated acute care surgery patients. Microbiologic samples were taken daily from all patients. Demographics, clinical, laboratory, and radiographic data were collected. The Johanson Criteria (JC) and Clinical Pulmonary Infection Score (CPIS) were calculated and analyzed. Sensitivity, specificity, and positive predictive values (PPV) and negative predictive value (NPV) were calculated in comparison to positive respiratory cultures. Of the 508 patients, 312 (61.4%) were acutely injured; emergent general surgery was performed in 141 (27.8%) patients, and 54 (10.6%) underwent elective operation. Positive respiratory cultures were identified in 198 (39%) of the 508 patients. JC diagnosed VAP in 291 (57.3%) patients (sensitivity 82.8%, specificity 59%, PPV 56.4%, NPV 84.3%, accuracy 68.3%). The CPIS resulted in 189 (37.2%) VAP diagnoses (sensitivity 61.1%, specificity 78.1%, PPV 64%, NPV 75.9%, and accuracy 71.5%). To address the inaccuracy of the algorithms, concordance testing was performed on the data to evaluate correlation between the algorithmic VAP diagnosis criteria and respiratory culture data. Nonconcordance with culture data diagnosis was identified with both JC (rho 0.41) and CPIS (rho 0.41). Sensitivity, specificity, PPV and NPV, and accuracy of both established clinical formulas was unacceptably low in acute care surgery patients.