Can the clinical pulmonary infection score impact ICU antibiotic days?

BACKGROUND: The Clinical Pulmonary Infection Score (CPIS) has been used in the intensive care unit (ICU) as a decision tool for initiation of antibiotics in suspected pneumonia and also for discontinuing antibiotics if the CPIS score is <or=6 on day three of therapy, but it is not in common clinical use. We sought to determine if application of a CPIS score<or=6 at three days could reduce antibiotic use and if a blinded committee would have a greater percentage of patients with CPIS>6 on day one receiving antibiotics empirically for pneumonia.
METHODS: Over 11 months, we evaluated empiric antibiotics prospectively in two ICUs of a large tertiary university teaching hospital. A pneumonia committee (PC) reviewed all patients and defined pneumonia according to the guidelines of the U.S. Centers for Disease Control and Prevention (CDC). The CPIS was calculated for all patients at day one and day three of antibiotic therapy. The percentage of patients with a CPIS<or=6 was compared for the ICU and PC, and the total antibiotic days potentially saved by using CPIS<or=6 as the criterion for treatment were determined. Receiver operating characteristic (ROC) curves and inter-observer reliability were determined.
RESULTS: Three hundred twelve patients received empiric antibiotics, 83 of whom were believed to have pneumonia by the ICU staff (2,283 antibiotic days). On day one, the 55 patients started on antibiotics had a CPIS<or=6, with 1,460 antibiotic-days, and only 28 patients had a CPIS>6 (823 antibiotic-days). In contrast, the PC determined 19 patients (23%) to have pneumonia by the CDC definition (731 antibiotic-days), with eight of these patients having a CPIS<or=6 and 11 a CPIS>6. Pneumonia committee review resulted in fewer patients believed to have pneumonia and a greater percentage with a CPIS>6 (odds ratio [OR] 2.7; 95% confidence interval [CI] 0.86, 8.6; p=0.05). Restriction of antibiotics to patients with a CPIS>6 would have saved 1,460 antibiotic-days at day one and 1,053 days if treatment was delayed until day three. Clinical Pulmonary Infection Score ROC curves for the PC showed an area under the curve (AUC) of 0.82 (95% CI 0.72, 0.91), whereas the AUC for the ICU group was 0.85 (95% CI 0.79, 0.92). The sensitivity and specificity of a CPIS>6 for the PC were 79% and 75%, respectively, with correct prediction 76% of the time. The inter-observer reliability of the CPIS had a kappa value of 0.88.
CONCLUSIONS: This prospective evaluation confirms that 50% of antibiotic-days in our ICU are used empirically for pneumonia when that infection is not likely to be present by either CDC or CPIS criteria. Although the CPIS has good reliability and acceptable sensitivity and specificity, PC review and CPIS<or=6 were commonly divergent (42-47%). Thus, better strategies should be developed for identification of pneumonia and empiric antibiotic administration in the ICU.