A Process Mining Approach in Identifying Patient Disease Trajectory using Electronic Health Records

An electronic health record contains valuable evidence of digital footprint that could help clinicians and medical researchers understand how diseases progress over time. Process mining provides various tools to identify and model the disease trajectory, including implementing PM2, a process mining framework to conduct process mining projects. This thesis’s aims to: (1) develop methods based on process mining to identify disease trajectory models, (2) study the applicability of the method using different EHR, and 3) mine disease trajectory models from actual EHRs. Three studies were conducted to achieve these aims. The first study was a feasibility study of process mining for identifying disease trajectories using a synthetic data set. The second study identified disease trajectory using an actual EHR from a private teaching hospital in Boston, USA. The third study identified disease trajectories from England’s population-wide EHR – the NHS England’s Hospital Episode Statistic.   This study demonstrates that process mining is useful for mining disease trajectories from electronic health records using standard tools. The method is feasible for producing relatively clinically relevant disease trajectory models from actual electronic health records, which are readily available in any formal health care information systems. This study intends to expand the rich set of techniques and areas of implementation.