[Prediction of bacteremia in patients with suspicion of infection in emergency room]

Background and objectives: To evaluate the relationship between some clinical and analytical data and the presence of bacteremia in order to establish a clinical decision rule.

Patients and methods: All the patients with blood cultures obtained from the emergency room in a two months period were analyzed. Patients were randomly assigned to derivation or validation sets. A logistic regression of the significant values in the univariate analysis was performed and a score obtained. The prevalence of bacteraemia for every score was calculated. The diagnostic efficacy curves and the performance of the predictive model were calculated.

Results: 412 patients were enrolled. The blood cultures were positive in 12.8% of them. The significant values in the univariate analysis were Charlson index ≥2 and PCT > 0.4ng/ml. Four groups of increasing risk of bacteraemia were designed, from 0 to 35% in the derivation set and from 2.9% to 27.2% in the validation set. In the diagnostic efficacy curve, the AUC was 0.8 in the derivation set and 0.74 in the validation set. The model presented a negative predictive value of 95.2% in the derivation set and 95.3% in the validation set.

Conclusions: A model that includes Charlson index and PCT makes possible to define a group of patients with a very low risk of bacteremia.