Dr. Cohen has a small but productive basic science lab using a mouse model of trauma and hemorrhagic shock, as well as an active clinical research group collecting clinical blood samples to study coagulopathy in trauma patients. The ability to work translationally was very appealing, as well as his support in taking several very useful statistical classes.

1. Clinical acute traumatic coagulopathy
   Despite advances in resuscitation practices, almost 1/3 of critically injured patients have abnormal blood clotting on arrival to the emergency department. Our group has an ongoing prospective, longitudinal study of coagulation abnormalities in more than 500 trauma patients admitted to SFGH, using thromboelastography, platelet aggregometry, real-time thrombin generation assays, and clotting factor analysis to explore coagulopathy after trauma.
2. Bioinformatics in critical care
   Our usual mode of study to identify the relative influence of a novel factor or patient characteristic to outcomes uses univariate statistics and regression modeling to control for potential confounders; however, the sheer volume and complexity of clinical information available makes these approaches subject to model over-fitting and poor generalizability. In collaboration with other investigators at SFGH and at the Berkeley School of Public Health, we are exploring the use of advanced statistical techniques such as principal components analysis, hierarchical clustering, and superlearner-based regression analysis to better separate critical information from statistical noise in a busy critical care setting.