Introduction
This study focuses on healthcare process automation by leveraging process mining and simulation to model clinical pathways for Chronic Obstructive Pulmonary Disease (COPD). Traditional methods of analyzing patient journeys require extensive manual effort, making it difficult to generalize findings across populations.
By utilizing AnyLogic healthcare simulation software, the research automates the validation of process-mined clinical models, ensuring they accurately reflect real-world patient flows. This approach supports data-driven decision-making in healthcare, helping to optimize patient care, resource allocation, and disease management strategies.
Simulation model
To enable automated clinical pathway modeling, AnyLogic is used to transform process-mined models into agent-based simulations. Each patient is represented as an autonomous agent navigating through different stages of COPD—risk, diagnosis, hospitalization, and oxygen therapy.
These stages are structured within a State Chart (CPSC - Clinical Pathway State Chart), where transitions occur based on probabilities extracted from real-world hospitalization data.
The project framework consists of:
- Process mining using Disco – extracting patient pathways from medical records.
- Simulation in AnyLogic – modeling thousands of patient trajectories to analyze variations in disease progression.
- Validation via PM4PY – comparing synthetic patient logs from simulation with actual hospital data to assess fitness, precision, and generalization.
By automating clinical pathway analysis, the simulation virtualizes patient flow patterns and supports healthcare decision-makers in optimizing care delivery.
Results
The AnyLogic-based simulation successfully replicates COPD patient journeys, with process-mined models achieving fitness above 65% and precision exceeding 83%. The study demonstrates that simulation-based validation improves process mining by quantifying model accuracy, identifying deviations in clinical pathways, and predicting patient care outcomes.
Future work will focus on enhancing simulation accuracy and incorporating "what-if" scenario analysis to improve predictive modeling and support data-driven decision-making in healthcare process automation.