Process Mining for Continuous Improvement in Healthcare: A Systematic Literature Review

Hospitals are a mix of good intentions and messy realities: many clinical processes run non-linear timescales and involve multiple systems and people. This thesis seeks to answer: (a) what is process mining (PM); (b) how teams are actually applying it; and, (c) What have studies told us so far about how to apply PM for continuous improvement in healthcare? The process mining approach is a data-driven technique, which processes event logs (case id, activity performed, timestamp and additional attributes) to bring back “how work actually gets done” and for generating concise process maps, performance metrics and variant reports. It lies between data science and improvement practice and aids discovery (what is happening), conformance checking (is practice following the model) and enhancement (where to change the process). The most common, practical steps that the literature and case work have to offer are simple: you select a candidate process, extract and link the event data, clean up and enrich the log in question and run discovery or conformance checking algorithms (alpha/fuzzy/heuristic/inductive types), offer simple graphs/KPIs for learning purposes (cycle time, throughput, waiting time), perform focused Kaizen/DMAIC type interventions & validate change by re-mining. Disco, ProM and PM4Py are also popular tools for these stages. What do the studies tell us? Process-mining literature and an ED case study find PM consistently uncovers bottlenecks, variant behavior and waiting-time hotspots and can support simulation but with ‘flesh on the bones’ enabling it to suggest concrete actions (e.g., from discharge-lounge or staff re-allocation) to test. The thesis closes with a minimal-viable checklist for mid-size hospitals to help teams make messy logs into repeatable, evidence-based improvement.