Predictors of bacteraemia in patients with suspected community-acquired pneumonia

INTRODUCTION: The diagnostic yield of blood cultures is limited in patients with community-acquired pneumonia (CAP). Yet, positive blood culture results provide important information for antibiotic treatment and for monitoring epidemiologic trends. We investigated the potential of clinical predictors to improve the cost-benefit ratio of obtaining blood cultures.

METHODS: Data from two prospective cohort studies of adults with suspected CAP, admitted to non-ICU wards, were combined. Two models were created, one using readily available parameters and one additionally including laboratory parameters.

RESULTS: 3,786 patients were included (2,626 (69%) with X-ray confirmed CAP). Blood cultures were obtained from 2,977 (79%) patients (and from 2,107 (80%) with X-ray confirmed CAP). 266 (8.9%) of the patients with a blood culture had bacteraemia. Clinical predictors of bacteraemia were absence of pre-admission antibiotic treatment, pleuritic pain, gastro-intestinal symptoms, tachycardia, tachypnea, hypotension and absence of hypoxia. After including laboratory results in the model, younger age, C-reactive protein, leukocytosis or leukopenia, low thrombocyte count, low sodium level, elevated urea and elevated arterial pH were added, while gastro-intestinal symptoms and hypotension were no longer significant. The area under the receiver operating characteristics curve was 0.66 (95% confidence interval 0.63-0.70) for the first model and 0.76 (95% confidence interval 0.73-0.79) for the second model.

CONCLUSION: In conclusion, in patients hospitalized with CAP, bacteraemia was moderately predictable using clinical parameters only. We recommend against the use of a risk prediction model for the decision to obtain blood cultures.