Including Organ Dysfunctions in a Predictive Score for Nosocomial Pneumonia After Cardiothoracic Surgery

Abstract

Background: Clinical diagnosis of ICU-acquired pneumonia after cardiothoracic surgery is challenging. Johanson criteria (chest radiograph infiltrate, purulent tracheal secretions, fever, and leukocytosis) fail in half the cases. A high Clinical Pulmonary Infection Score (CPIS) and ≥ 2-point increase in Sequential Organ Failure Assessment (SOFA) score (SOFA↑ ≥ 2) may improve diagnosis. The aim of the study was to evaluate whether CPIS or SOFA↑ ≥ 2 contributes to predict ICU-acquired pneumonia in subjects after cardiothoracic surgery.

Methods: We used a prospective observational design. Spiegelhalter-Knill-Jones scoring systems including CPIS or SOFA↑ ≥ 2, together with other clinical and laboratory variables, were developed in a derivation cohort. A positive quantitative pulmonary sample culture was required to confirm ICU-acquired pneumonia. Area under the receiver operating characteristic curve (AUROC) was computed for each of the 2 scoring systems. The best system was evaluated in a validation cohort.

Results: Derivation and validation cohorts included 172 and 108 subjects, with 410 and 216 suspected ICU-acquired pneumonia episodes, respectively. AUROC was 0.53 ± 0.03 for CPIS (P = .29) and 0.54 ± 0.03 for SOFA↑ ≥ 2 (P = .29). Adding purulent tracheal secretions and leukocytosis to SOFA↑ ≥ 2 (SOFA model) increased AUROC to 0.65 ± 0.03 (P < .001). Adding catecholamine use to CPIS (CPIS model) increased AUROC only slightly, to 0.57 ± 0.03. The probabilities predicted by the SOFA model were reliable, especially when high or low.

Conclusions: A clinical scoring system including at least SOFA↑ ≥ 2 increase barely improved ICU-acquired pneumonia prediction in subjects after cardiothoracic surgery.

Keywords: CPIS; SOFA score; hospital-acquired pneumonia; nosocomial pneumonia; prediction model development; sepsis.