Presentation (CTI)

OBJECTIVES: Process mining enables data-driven reconstruction of real-world care pathways but is still rarely applied in oncology. This study reports the first implementation of process mining at IRST “Dino Amadori,” focusing on early-stage breast cancer. The objective was to map actual patient flows, assess adherence to clinical guidelines (PDTA), and identify opportunities for process optimization.
METHODS: We analyzed the oncological pathways of 2,785 patients who underwent their first oncology consultation between 2015 and 2023. Administrative and clinical data were extracted and preprocessed to construct event logs, enabling process discovery with the PM4Py library. Stratification by cancer stage improved model clarity. Data-driven pathways were compared with the regional clinical guidelines (PDTA) to assess adherence and identify process deviations.
RESULTS: The resulting process maps provided a clear visualization of the main care pathways, particularly after refining the event logs through stratification and filtering techniques. The analysis confirmed alignment with core clinical milestones while also highlighting opportunities to optimize pathway adherence. A focused evaluation of time to chemotherapy (TTC) showed that approximately 84% of patients initiated treatment within the clinically recommended 60-day window. Throughout the project, iterative feedback from a multidisciplinary group of clinicians and hospital administrators played a pivotal role in interpreting results and validating model relevance. This collaboration enabled a more nuanced understanding of deviations from expected care pathways.
CONCLUSIONS: Our study demonstrates the feasibility of process-oriented data science in oncology. The approach offers a reproducible methodology for deriving actionable insights and supporting evidence-based improvements in complex healthcare settings.