Predictive model for bacteremia in adult patients with blood cultures performed at the emergency department: A preliminary report

Abstract

Useful predictive models for identifying patients at high risk of bacteremia at the emergency department (ED) are lacking. This study attempted to provide useful predictive models for identifying patients at high risk of bacteremia at the ED. A prospective cohort study was conducted at the ED of a tertiary care hospital from October 1 to November 30, 2004. Patients aged 15 years or older, who had at least two sets of blood culture, were recruited. Data were analyzed on selected covariates, including demographic characteristics, predisposing conditions, clinical presentations, laboratory tests, and presumptive diagnosis, at the ED. An iterative procedure was used to build up a logistic model, which was then simplified into a coefficient-based scoring system. A total of 558 patients with 84 episodes of true bacteremia were enrolled. Predictors of bacteremia and their assigned scores were as follows: fever greater than or equal to 38.3°C [odds ratio (OR), 2.64], 1 point; tachycardia greater than or equal to 120/min (OR, 2.521), 1 point; lymphopenia less than 0.5×10(3)/μL (OR, 3.356), 2 points; aspartate transaminase greater than 40IU/L (OR, 2.355), 1 point; C-reactive protein greater than 10mg/dL (OR, 2.226), 1 point; procalcitonin greater than 0.5 ng/mL (OR, 3.147), 2 points; and presumptive diagnosis of respiratory tract infection (OR, 0.236), -2 points. The area under the receiver operating characteristic curves of the original logistic model and the simplified scoring model using the aforementioned seven predictors and their assigned scores were 0.854 (95% confidence interval, 0.806-0.902) and 0.845 (95% confidence interval, 0.798-0.894), respectively. This simplified scoring system could rapidly identify high-risk patients of bacteremia at the ED.