Currently existing predictive models for massive blood transfusion in major trauma patients had limitations for sequential evaluation of patients and lack of dynamic parameters.

To establish a predictive model for predicting the need of massive blood transfusion major trauma patients, integrating dynamic parameters.

Multi-center retrospective cohort study.

Four designated trauma centers in Hong Kong.

Trauma patients aged >12years were recruited from the trauma registries from 2005 to 2012. MBT was defined as delivery of ≥10units of packed red cells within 24h. Split sampling method was adopted for model building and validation. Multivariate logistic regression was adopted for model building, with weight assigned based on logarithmic of adjusted odds ratios. The performance of the dynamic MBT score (DMBT) was compared with the PWH score and the Trauma Associated Severe Hemorrhage (TASH) score in the validation data set.

4991 patients were included in the study. The DMBT was established with 8 parameters: systolic blood pressure, heart rate, hemoglobin, hemoglobin drop within the first 2h, INR, base deficit, unstable pelvic fracture and hemoperitoneum in radiological imaging. At cut-off score of 6 the DMBT achieved sensitivity of 78.2% and specificity of 89.2%. In the validation set, the AUCs of the DMBT, PWH score, and TASH score were 0.907, 0.844, and 0.867 respectively.

The DMBT score allows both snapshot and sequential activation along the trauma care pathway and has better performance than the PWH score and TASH score.