Objective: To evaluate the potential ability of an algorithm based on the clinical pulmonary infection score (CPIS) to identify and treat patients with bacterial ventilator-associated pneumonia (VAP) compared to a strategy based on quantitative cultures of bronchoscopic specimens.
Design: Retrospective cohort study.
Setting: Thirty-one critical care units across France.
Patients: Two hundred and one patients clinically suspected of having VAP who had been included in the "invasive strategy" group of the French multicenter randomized trial and for whose quantitative cultures bronchoscopic specimens were obtained. CPIS was determined retrospectively, based on data that had been collected for the initial study.
Measurements and results: The clinical pulmonary infection score was determined on days 1 and 3, and compared in patients identified as having developed VAP or not, as defined by bronchoscopic specimen culture results. On day 3 138 of the 201 patients (69%) had a CPIS of more than 6 that would have required prolonged antimicrobial therapy based on the algorithm. In contrast, based on bronchoscopy, only 88 (44%) patients were considered to have VAP (kappa coefficient for concordance between the two strategies, 0.33). While the sensitivity of CPIS more than 6 on day 3 for identifying VAP was 89%, its specificity was only 47%, leading to potentially unnecessary treatment of 60 (53%) of the 113 patients without VAP as diagnosed by bronchoscopy.
Conclusion: A strategy based on the CPIS to decide which patients with suspected VAP should receive prolonged administration of antibiotics would appear to over-prescribe these agents, as compared to a strategy based on bronchoscopy.