Abstract O-50: PREDICTION MODEL FOR MORTALITY IN PATIENTS WITH DISSEMINATED INTRAVASCULAR COAGULOPATHY BASED ON PEDIATRIC ICU ADMISSION: THREE FACTORS IDENTIFIED

Abstract

Aims & Objectives: Predictive scoring systems measure severity of disease and prognosis of patients in the ICU. Such measurements are helpful for clinical decision-making, standardizing research, comparing the quality of patient care across different units, quantifying severity of illness, and likelihood of mortality during hospitalization. There is no severity score based on the presence of disseminated intravascular coagulation (DIC) in the pediatric population. This study Identifies risk factors and variables that predict mortality and survival rates in pediatric patients presenting to the ICU with DIC. Methods The Virtual PICU database was used to extract 11,569 charts of PICU patients with a diagnosis of DIC on admission. 6,184 patients remained after excluding charts with >40% missing data. Other missing data were addressed using imputation with the k-nearest neighbor method and calculations for certain variables (i.e., bicarbonate, base excess). This data was separated into training and testing sets and boosted tree analysis was performed. Results The three variables associated with highest mortality included elevated PTT (87,95% CI), low bicarbonate (9.8,95% CI), and negative base excess (-18,95% CI). Training and testing datasets were similar - 74.94% versus 73.09% for those who survived and 71.99% versus 70.41% for those who died. A combination of these variables had a prediction accuracy of ~73% for predicting survival and a prediction accuracy of ~70% in predicting mortality. Conclusions Elevated PTT, low bicarbonate, and base deficit predict mortality in patients with DIC on admission. Prospective studies can assess the prediction model and mortality rates in this population.