Comparison of the Accuracy of Local and International Prediction Models for Massive Transfusion in Major Trauma Patients

Abstract

Objective To validate a local prediction model (Rainer's score) for massive transfusion in trauma patients and compare its accuracy with international prediction models (TASH score and ABC score) using local trauma data. Methods Patients were recruited retrospectively from the trauma registry of a regional trauma centre during the period from January 2005 to December 2010. Patients with Injury Severity Score (ISS) ≥9 and aged ≥12 years were included. Burn and drowning patients, patients with known severe anemia (haemoglobin <7 g/dL) and renal failure were excluded. Massive transfusion was defined as delivery of ≥10 units of packed red blood cells within 24 hours. The sensitivity specificity positive and negative predictive values, positive and negative likelihood ratio and accuracy were calculated for different prediction models. The overall discriminatory capacities of models were compared using the area under the receiver operating characteristic (ROC) curve. Results A total of 1030 patents met the inclusion criteria and 27 patients required ≥10 units packed cells within 24 hours. The accuracy was the best in the TASH score (97.3%) when compared to the Rainer's score (96.5%) and ABC score (95.1%). Sensitivity was better in Rainer's score (33.3%) and ABC score (33.3%) than in the TASH score (25.9%). The area under ROC curve for TASH score, Rainer's score and ABC score were 0.911, 0.886 and 0.809 respectively. Conclusions We validate the prediction model (Rainer's score) by a set of local, non-university institution data. Rainer's score has high accuracy (97%) in predicting the need for massive transfusion and is comparable to TASH score and ABC score.