Outcome centred process mapping in healthcare using random forest and process mining

System process mapping is essential for understanding complex systems and implementing effective management practices. In healthcare, mapping patient flow aims to reduce costs, improve the quality of care, and enhance efficiency. Process mining (PM) in healthcare is challenging due to the need for specialised knowledge and the inherent variability and complexity of healthcare processes. Comparing observed differences in patient flow pathways provides only a partial view; they must be combined with process outcomes and attributes for a comprehensive understanding. This paper proposes a combined stepwise approach using random forest (RF) and PM discovery to achieve outcome-centred process mapping from process-unaware systems. To this aim, we analysed the MIMIC-IV v2.2 dataset, containing healthcare data from patients at the Beth Israel Deaconess Medical Center (BIDMC) between the years 2008-2019. The MIMIC-IV dataset includes different types of sources within the hospital such as the emergency department (ED), patient measurements, procedures, transfer between departments, and intensive care units (ICUs). The results indicate that older patients with high ED priority and multimorbidity are particularly complex and challenging to treat, necessitating the implementation of tailored management strategies, in agreement with common clinical practice. A potential application of this approach is the real-time prediction of patients' length of stay, which could optimise clinicians' work, save healthcare resources, and improve the quality of patient care. Our approach can be applied to improve system process management by combining control-flow and data perspectives.