Print E-mail
Volume 47, Number 6, December 2015

Derivation of a clinical prediction rule for bloodstream infection mortality of patients visiting the emergency department based on predisposition, infection, response, and organ dysfunction concept 

Chun-Fu Yeh, Kuan-Fu Chen, Jung-Jr Ye, Ching-Tai Huang

Received: February 25, 2013    Revised: June 1, 2013    Accepted: June 28, 2013   


Corresponding author:

*Corresponding author. Department of Emergency Medicine, Chang Gung Memorial Hospital, Maijin Road, Keelung 222, Taiwan.
** Corresponding author. Division of Infectious Diseases, Department of Internal Medicine, Chang Gung Memorial Hospital, 5 Fu-Shin Street,
Gueishan 333, Taoyuan, Taiwan.
E-mail addresses:, (K.-F. Chen), (C.-T. Huang). 


Background and purpose: 

Bloodstream infection (BSI) is a serious infection with a high mortality. We aimed to construct a predictive scoring system to stratify the severity of patients with BSI visiting the emergency department (ED). 



We conducted a retrospective cohort study consisting of patients who visited the ED of a tertiary hospital with documented BSI in 2010. The potential predictors of mortality were obtained via chart review. Multivariate logistic regression was utilized to identify predictors of mortality. Penalized maximum likelihood estimation (PMLE) was applied for score development. 



There were 1063 patients with bacteremia included, with an overall 28-day mortality rate of 13.2% (n = 140). In multiple logistic regression with penalization, the independent predictors of death were “predisposition”: malignancy (β-coefficient, 0.65; +2 points); “infection”: Staphylococcus aureus (S. aureus) bacteremia (0.69; +2 points), pneumonia (1.32; +4 points), and bacteremia with an unknown focus (0.70; +2 points); “response”: body temperature <36°C (1.17; +3 points), band form >5% (1.00; +3 points), and red blood cell distribution width (RDW) >15% (0.63; +2 points); and “organ dysfunction”: pulse oximeter oxygen saturation <90% (0.72; +2 points) and creatinine >2 mg/dL (0.69; +2 points). The area under receiver operating characteristic curve (AUROC) for the model was 0.881 [95% confidence interval (CI), 0.848–0.913], with a better performance than the Pitt bacteremia score (AUROC: 0.750; 95% CI 0.699–0.800, p < 0.001). The patients were stratified into four risk groups: (1) low, 0–3 points, mortality rate: 1.5%; (2) moderate, 4–6 points, mortality rate: 10.5%; (3) high, 7–8 points, mortality rate: 28.6%; and (4) very high, ≥9 points, mortality rate: 65.5%. 



The new scoring system for bacteremia could facilitate the prediction of the risk of 28-day mortality for patients visiting the ED with BSI. 


Key words:

Bacteremia, Bloodstream infection, Mortality, Pitt bacteremia score predictive model