Predictors of Bacteraemia in Emergency Department Patients with Pneumonia

Objective To determine the proportion of patients investigated in an Australian emergency department (ED) for non-hospital-acquired pneumonia with bacteraemia and to identify risk factors for bacteraemia. Method Retrospective cohort study of patients with an ED diagnosis of non-hospital-acquired pneumonia in whom blood cultures were taken. Data collected included demographics, history of chronic obstructive pulmonary disease (COPD), immunosuppression, intravenous (IV) drug use or diabetes, prior antibiotic use, clinical features, biochemistry, haematology and blood culture results, ED disposition and pneumonia severity index (PSI) class. Outcomes of interest were the proportion of blood cultures that identified bacteraemia and identification of independent predictors of bacteraemia. Data analysis was by descriptive statistics, odds ratios (OR) and multivariate analysis. Results Two hundred patients were studied. The bacteraemia rate was 7% (95% CI=4-11%). IV drug use (OR 16.7, 95% CI=2.65-105) and pulse rate (OR 1.29, 95% CI=1.01-1.65, per 10 beat rise) were independently associated with bacteraemia. Overall, 1/199 patients had a significant broadening of therapy based on a blood culture result (0.5%, 95% CI=0.09-2.8%). On post hoc analysis, using PSI class IV/V or known IV drug use as criteria for blood culture ordering had sensitivity 92.9% (95% CI=64.1-99.6%) and negative predictive value 98.9% (95% CI=93.5-99.9%) for bacteraemia. Conclusion 7% of blood cultures from patients with non-hospital acquired pneumonia showed bacteraemia. Using the combination of PSI class IV/V or IV drug use as criteria for blood culture ordering shows promise.