Stochastic-Aware Comparative Process Mining in Healthcare.

Evidence-based innovations are critical in optimising the delivery of healthcare services. Process mining aims to provide healthcare stakeholders with insights, derived from historical data recorded in hospital information systems, to optimise healthcare processes. Healthcare processes are well-known for their complexity and control-flow variations are inherent in patient pathways undertaken by different patient cohorts. Comparative process mining can reveal insights from studying the differences between healthcare processes to better understand best-practice patient pathways. In this paper, we take a design science approach to redefine an existing method for process comparison (PCM). Where PCM considers predominantly the control-flow perspective, we extend this method with the stochastic perspective, that is, how likely a particular pathway is for certain patient cohorts, to obtain the Probabilistic Process Comparison Method (P 2 CM). Furthermore, we further automate the method. Concretely, we introduce new, stochastic-aware, methods for sub-dividing process behaviour into cohorts based on trace attributes or other trace features, methods for focusing the comparative analysis on specific pairs of interesting cohorts, and provide a new method for in-depth comparison of process differences. The approach is evaluated using three real-life healthcare datasets, of which one case study is conducted with a domain expert from an Australian hospital.