Institutional members access full text with Ovid®

Share this article on:


Mann-Salinas Elizabeth; Chang, Steven; Bartels, John; Hurst, Kelly; Batchinsky, Andriy
Critical Care Medicine: December 2013
doi: 10.1097/
Poster Session: Burns/Trauma 5: PDF Only

Introduction: The purpose of this project was to conduct a secondary validation of 6 novel predictors of sepsis in the burn patient (heart rate (HR) >130 bpm, mean arterial pressure (MAP) <60 mmHg, base deficit (BD) <-6 mEq/L, temp <36°C, use of vasoactive medications, and glucose >150 mg/dl). Three identified time periods were compared: sepsis (“sick”) with positive blood culture, sepsis (“sick”) with negative blood culture, and control (“non-sick”). This model was significant in predicting positive sick and all sick with area under the receiver operating curve (AUROC) of 0.775 (p < 0.001) and 0.714 (p < .001), respectively, up to 48 hours prior. Methods: Analysis was performed using 11 clinical variables from an existing within-patient comparison database of 59 burn patients from a single center. Secondary analysis compared 4 groups: non-sick vs positive sick; non-sick vs negative sick; negative sick vs positive sick; and non-sick versus all sick. Modified-random forest and CART (Classification and Regression Tree) decision tree approaches were applied to the dataset. Limitations of data excluded 2 “novel predictors” from this analysis: glucose and BD. Results: Positive sick vs not sick was the only paring with AUROC > 0.7, significant only 24 hours prior. Classifiers that achieved target AUROC include: max temp > 39.4 OR max blood transfusion (BT) ≥2 OR MAP < 58 = AUROC 0.737; MAP < 55 OR min BT ≥2 = AUROC 0.728; and any 2 of max 24 hour urine < 299, MAP < 57, Temp < 36.2°C = AUROC 0.720. High HR and low MAP were important to discriminate any sick vs non-sick patient. Conclusions: Positive sick versus not sick is the dichotomy best predicted accurately overall. Elevated HR and low MAP have been further validated as strong predictors of burn sepsis. Ongoing prospective validation of clinical predictors continues at our burn center as decision support technology solutions are in development to incorporate best predictors into useful tools for bedside providers.

© 2013 by the Society of Critical Care Medicine and Lippincott Williams & Wilkins