Predicting positive blood cultures in patients presenting with pneumonia at an Emergency Department in Singapore

ABSTRACT

<p><b>INTRODUCTION</b>Routine blood cultures have been recommended for all patients in treatment guidelines for community-acquired pneumonia (CAP). This practice has become a major area of resource utilisation, despite the lack of evidence in its clinical utility. Calls for abandoning the practice is balanced by the occasions of uncovering an unexpected pathogen or an unusual antimicrobial resistance pattern. The aim of this study is to identify factors that predict positive blood cultures among patients hospitalised for pneumonia upon presentation at the Emergency Department (ED).</p><p><b>MATERIALS AND METHODS</b>A case control study was carried out on patients treated for pneumonia in the ED who had routine blood cultures performed as part of their management. The pneumonia severity index (PSI) was used to categorize patients into low- and high-risk for 30-day mortality. Logistic regression was carried out to determine factors significantly associated with positive blood cultures, from which a predictive probability equation was used to identify patients whose blood cultures were negative at a pre-determined cut-off, with minimum number of culture positive misclassification. A scoring system was devised, with scores predicting which patients would be likely to have a positive or negative blood culture.</p><p><b>RESULTS</b>A total of 1407 patients with pneumonia were treated at ED from May to December 2006, from whom 1800 blood cultures were performed. Of these, 140 cultures (7.8%) grew organisms, comprising 96 (5.3%) true positive cultures and 44 (2.4%) contaminated cultures. Logistic regression analysis identified ill patients with higher PSI classes, smokers and Malay patients to be more likely to have positive blood cultures. Patients who had prior treatment with antibiotics, chronic obstructive pulmonary disease and cough were less likely to have positive blood cultures. An index to predict a negative blood culture resulted in the accurate classification of all but 4 positive patients while still correctly classifying 27.8% of blood culture negative patients. The area under the ROC curve was 0.71 (95% CI, 0.65-0.76). A simplified scoring system was devised based on the predictive model had a sensitivity of 82% and specificity of 38.2% for a positive blood culture.</p><p><b>CONCLUSION</b>Routine blood cultures yielded negative results in 94% of patients presenting with pneumonia. The development of the clinical scoring system is a first step towards selecting patients for whom blood cultures is performed and improve cost-effectiveness.</p>