Comparing care pathways between COVID-19 pandemic waves using electronic health records: a process mining case study

Abstract Purpose: The COVID-19 pandemic caused rapid shifts in the workflow of many health services, but evidence of how this affected multidisciplinary care settings is limited. We propose a Process Mining approach that utilises timestamped data from Electronic Health Records to compare care provider patterns across pandemic waves. Methods: We collected routine events from Scottish hospital episodes in adults with COVID-19 status and linked health provider inputs to generate standardised treatment logs. Conformance checking metrics were used to select the optimal model (Inductive Miner infrequent [IMi]) for downstream analysis. Visual diagrams from the discovered Petri Nets indicated the interactions on pre-coded provider and activity-level subsets. We used cross-log conformance checking and graph similarity to measure distances between adverse and less adverse groups across pandemic waves. Results: We included 1,153 patients with COVID-19 (302 [26%] in Wave 1 and 851 [74%] in Wave 2) with 55,212 relevant care provider events. At the conformance checking stage, the IMi model, achieved good log fitness (LF=0.95) and generalisation (G=0.89), but limited precision (PR=0.27) across all granularity levels. More structured care procedures in Wave 1 were present, compared to mixed multidisciplinary patterns in Wave 2. Care activities differed in patients with extended stay (GED=348, PR=0.231 vs GED=197, PR=0.429 in shorter stays). Patients in out-of-hours care and intensive therapy were linked with more standardised patterns. Conclusion: Process Mining can be incorporated alongside clinical oversight to provide visual and quantitative comparisons of care interactions in COVID-19 episodes and enhance further research in complex cases.