Adjusting a prognostic score for burned children with logistic regression

Abstract

This study was aimed at the adjustment and cross-validation of a prognostic score for burned children on hospital admission. DEMI score assesses seven risk factors related to four main topics: depth (percent of partial-thickness and full-thickness burn), extension (percent of total body surface area burned and injury to dorso-gluteal area), morbidity on admission (preexisting illness and complications caused by inadequate transfer), and presence of Inhalation syndrome. These items were selected in a previous case control study with bivariate analysis on an estimation sample of 168 children with acute burns admitted to the burn unit of the Hospital de Pediatria Juan P. Garrahan between July 1991 and October 1993. This study uses the same retrospective sample to adjust this score by means of multivariate analysis. Only four items were included in the logistic regression formula: percentage total body surface area burned, inadequate transfer, inhalation syndrome, and dorso-gluteal affection. Resulting abbreviation of DEMI score was then cross-validated on a prospective validation sample of 137 patients, yielding high correlation with the outcome (R = 0.78) and high sensitivity (80.95%), specificity (96.55%), and positive (80.95%) and negative (96.55%) predictive values. We conclude that this new DEMI score is a simple and accurate tool to predict mortality risk in burned children.