A K-Reversible Approach to Model Clinical Trajectories

A clinical trajectory can be defined as the path followed by patients between an initial heath state S_i such as being healthy to another state S_j such as being diagnosed with a specific clinical condition. Being able to identify the common trajectories that a group of patients take can benefit clinicians at identifying the current state of patient and potentially provide early treatment to avoid going towards specific paths. In this paper we present our approach that enables a clinical dataset of patient encounters to be clustered into groups of similarity and run through our algorithm which produces an automaton displaying the most common trajectories taken by patients. Furthermore, we explore a dataset of patients that have experienced mild traumatic brain injuries (mTBI) to show that our approach is effective at clustering and identifying common trajectories for patients that develop headaches, sleep, and post traumatic stress disorder (PTSD) post concussion.