Development and validation of a predictive model for bacteremia in patients hospitalized by the emergency department with suspected infection

Positive blood cultures usually indicate disseminated infection that is associated with a poor prognosis and higher mortality. We seek to develop and validate a predictive model to identify factors associated with positive blood cultures in emergency patients. Secondary analysis of data from two prospective cohorts (EPISEPSIS: developing cohort, and DISEPSIS: validation cohort) of patients with suspected or confirmed infection, assembled in emergency services in 10 hospitals in four cities in Colombia between September 2007 and February 2008. A logistic multivariable model was fitted to identify clinical and laboratory variables predictive of positive blood culture. We analyzed 719 patients in developing and 467 in validation cohort with 32% and 21% positive blood cultures, respectively. The final predictive model included variables with significant coefficients for both cohorts: temperature > 38° C, Glasgow < 15 and platelet < 150.000 cells/mm³, with calibration (goodness-of-fit H-L) p = 0.0907 and p = 0.7003 and discrimination AUC = 0.68 (95% CI = 0.65-0.72) and 0.65 (95% CI = 0.61-0.70) in EPISEPSIS and DISEPSIS, respectively. Specifically, temperature > 38 °C and platelets < 150.000 cells/mm³ and normal Glasgow; or Glasgow < 15 with normal temperature and platelets exhibit a LR between 1,9 (CI 95% = 1,2-3,1) and 2,3 (CI 95% = 1,7-3,1). Glasgow < 15 with any of low platelets or high temperature shows a LR between 2,2 (CI 95% = 1,1-4,4) and 2,6 (CI 95% = 1,7-4,3). Temperature > 38° C, platelet count < 150,000 cells/mm³ and GCS < 15 are variables associated with increased likelihood of having a positive blood culture.