BackgroundStandard diagnostic criteria are not useful for identifying bloodstream infection in patients with an acute burn injury patients. The study objective was to develop and validate a tool using common laboratory, clinical, and patient parameters for early identification of bloodstream infection after acute burn injury (within 10 days after a burn). MethodsWe retrospectively and prospectively reviewed for tool development the hospital course of patients with an acute burn injury (n = 156) and validated the tool in different cohorts (retrospective [n = 26] and prospective [n = 90]). The Pearson correlation identified independent variables associated with bloodstream infection (P < .1) in the development cohort that were then analyzed using binary logistic regression to identify the simplest model (P < .05; adjusted odds ratio >1). Classification and regression tree analysis was used to identify tool parameter breakpoints. Performance metrics were completed to evaluate and validate the tool. ResultsThe best model (P < .05) was: Ln [odds of bloodstream infection] = –96.749 + 3.230 (platelet volatility) + 2.235 (max temperature [°C]) + 0.339 (% full burn) + 0.242 (% partial burn) + 0.045 (max heart rate [bpm]), with a threshold probability categorizing bloodstream infection of >48%. The sensitivity, specificity, accuracy, false positive rate, false negative rate, and positive (+) and negative (–) likelihood ratios of the tool in the developmental cohort (n = 156) were 89%, 98%, 96%, 2%, 11%, 53, and 0·11, respectively; and in the prospective validation cohort (n = 90 were 91%, 90%, 90%, 10%, 9%, 9, and 0·1, respectively (n = 90). ConclusionThe validated bloodstream infection screening tool in patients with acute burn injury has excellent predictive ability to assist in the identification of patients for whom blood cultures should be requested.
Read full abstract