Clinical pulmonary infection score for ventilator-associated pneumonia: accuracy and inter-observer variability

Abstract

Objective: Although quantitative microbiological cultures of samples obtained by bronchoscopy are considered the most specific tool for diagnosing ventilator-associated pneumonia, this labor-intensive invasive technique is not widely used. The Clinical Pulmonary Infection Score (CPIS), a diagnostic algorithm that relies on easily available clinical, radiographic, and microbiological criteria, could be an attractive alternative for diagnosing ventilator-associated pneumonia. Initially, the CPIS scoring system was validated upon 40 quantitative cultures of bronchoalveolar lavage fluid from 28 patients, and only few other studies have evaluated this scoring system since then. Therefore, little is known about the accuracy of this score.

Design: We compared the scores of a slightly adjusted CPIS with results from quantitative cultures of bronchoalveolar lavage fluid in 99 consecutive patients with suspicion of ventilator-associated pneumonia, using growth of > or =10(4) cfu/ml in bronchoalveolar lavage fluid as a cut-off for diagnosing ventilator-associated pneumonia. In addition, the CPIS were calculated for 52 patients by two different intensivists to determine the inter-observer variability.

Results: Ventilator-associated pneumonia was diagnosed in 69 (69.6%) patients. When using a CPIS >5 as diagnostic cutoff, the sensitivity of the score was 83% and its specificity was 17%. The area under the Receiver Operating Characteristic curve was 0.55. The level of agreement for prospectively measured Clinical Pulmonary Infection Score (< or =6 and >6) was poor (kappa =0.16).

Conclusions: When compared to quantitative cultures of bronchoalveolar lavage fluid, the CPIS has a low sensitivity and specificity for diagnosing ventilator-associated pneumonia with considerable inter-observer variability.