Standard variables fail to identify patients who will not respond to fluid resuscitation following thermal injury: brief report

Approximately 13% of thermally injured patients fail resuscitation, in that they die during the first 48 h postburn despite full resuscitative efforts. The purpose of this study was to characterize these patients, and to develop a predictor of resuscitation failure. Records of 3807 thermally injured patients admitted to this burn centre during 1980–1997 were reviewed. Patients were classified as surviving to hospital discharge (“NONFAIL/LIVE”), as surviving resuscitation but dying later (“NONFAIL/DIE”), or as failing resuscitation (“FAIL”). Ordinal logistic regression was used to develop a predictor of membership in each of these three groups. With respect to total burn size, full-thickness burn size, and inhalation injury, the three groups represented a gradation in injury severity from least severe (NONFAIL/LIVE) to most severe (FAIL). The predictive model had an overall accuracy of 91.6%; however, it correctly classified NONFAIL/LIVE patients more often (97.7% accuracy) than it did NONFAIL/DIE patients (57.5%) or FAIL patients (16.1%). Patients who failed resuscitation were more severely injured than those who survived resuscitation, but was not possible accurately to predict who will fail resuscitation using data available on admission.