346

Abstract

Introduction: The aim of this study is to construct bacteremia prediction model using commonly available clinical variables in hospitalized patients with community acquired pneumonia (CAP). Methods: A prospective database including 2422 patients who were hospitalized with the diagnosis of CAP via emergency department were analyzed retrospectively. Independent risk factors were investigated by using multivariable logistic regression in 60% of cohort. We assigned weighted point to each predictive factors and made a prediction rule by using calculated clinical score. This model was validated both internally and externally with the remaining 40% of cohort and cohort of an independent academic hospital. Low risk group of bacteremia was defined as patients who have a risk of bacteremia under 3%. Results: Total of 2422 patients were included in this study. The overall rates of bacteremia were 5.7% in the cohort. The significant factors of predicting bacteremia were following 7 variables; SBP<90mmHg, HR>125beats/min, BT<35°C or >40°C,WBC<4000 or 12000 cells/µL,Platelet<130000 cells/µL cells, and Albumin<3.3g/dL, and CRP>17mg/dl. After using our prediction rule in validation cohorts, 78.7% and 74.8% of internal and external validation cohort were classified as low risk bacteremia group. The area under receiver operating characteristics curves were 0.75, and 0.79 in internal and external validation cohorts, respectively. Conclusions: This model could provide a guideline to perform the blood cultures in hospitalized CAP patients with potential reduction of blood cultures.