Abstract
In this study, we present a framework for analyzing associations between patient cohorts and the trauma resuscitation procedures their patients received. Our framework works by quantifying associations between discovered patient cohorts and treatment patterns. We evaluated our framework on a trauma resuscitation dataset collected in a level 1 trauma center. Our experimental results show that using weights learned by our algorithm improves measurements of patient similarity. Four patient cohorts were then found via clustering, and statistically significant resuscitation patterns were discovered using process mining techniques. Though only tested on the trauma resuscitation process, our framework can be generalized to analyze other medical processes.