Predictors of bacteremia in emergency department patients with suspected infection

Abstract Objectives The goal of this study is to identify clinical variables associated with bacteremia. Such data could provide a rational basis for blood culture testing in emergency department (ED) patients with suspected infection. Methods This is a secondary analysis of a prospective cohort of ED patients with suspected infection. Data collected included demographics, vital signs, medical history, suspected source of infection, laboratory and blood culture results and outcomes. Bacteremia was defined as a positive blood culture by Centers for Disease Control criteria. Clinical variables associated with bacteremia on univariate logistic regression were entered into a multivariable model. Results There were 5630 patients enrolled with an average age of 59.9 ± 19.9 years, and 54% were female. Blood cultures were obtained on 3310 (58.8%). There were 409 (12.4%) positive blood cultures, of which 68 (16.6%) were methicillin-resistant Staphylococcus aureus (MRSA) and 161 (39.4%) were Gram negatives. Ten covariates (respiratory failure, vasopressor use, neutrophilia, bandemia, thrombocytopenia, indwelling venous catheter, abnormal temperature, suspected line or urinary infection, or endocarditis) were associated with all-cause bacteremia in the final model (c-statistic area under the curve [AUC], 0.71). Additional factors associated with MRSA bacteremia included end-stage renal disease (odds ratio [OR], 3.9; 95% confidence interval [CI], 1.9-7.8) and diabetes (OR, 2.0; 95% CI, 1.1-3.6) (AUC, 0.73). Factors strongly associated with Gram-negative bacteremia included vasopressor use in the ED (OR, 2.8; 95% CI, 1.7-4.6), bandemia (OR, 3.5; 95% CI, 2.3-5.3), and suspected urinary infection (OR, 4.0; 95% CI, 2.8-5.8) (AUC, 0.75). Conclusions This study identified several clinical factors associated with bacteremia as well as MRSA and Gram-negative subtypes, but the magnitude of their associations is limited. Combining these covariates into a multivariable model moderately increases their predictive value.