Dynamic treatment allocation of blood products in the treatment of trauma injury

We present results from a dynamic treatment regime analysis using data from the PRospective, Observational, Multi-Center Massive Transfusion sTudy (PROMMTT). Our goal was to determine the optimal longitudinal allocation of blood products in the treatment of severe trauma in order to maximize survival. We estimated the cross-validated area under the ROC curve to assess the ability of available clinical variables to predict death. We defined our parameters of interest based on the causal inference literature, for example, the average treatment effect, defined as the change in the risk of death under interventions on the patterns of blood product allocation. The models on which the parameters were based used SuperLearner, an ensemble learner that takes a library of algorithms and determines an optimal combination by minimizing the cross-validated risk and a bias reduction step based on targeted maximum-likelihood. The analyses identified dynamic starting and stopping rules that are based on covariate measurements and treatment history, which decrease the probability of death. These methods are generalizable to other longitudinal studies in public health.