Abstract
Process Mining (PM) effectively diagnoses inefficiencies in complex healthcare workflows, such as chemotherapy protocols. However, current methodologies often remain retrospective or rely on loosely coupled simulations, leaving a critical methodological void: the inability to quantify the aggregate, system-wide operational impact of eliminating specific, diagnosed workflow deviations. This gap prevents decision-makers from forming evidence-based strategies for resource allocation. We address this by introducing the PM(2)-Predictive Impact Model (PIM) framework, a novel, fully embedded process-native methodology that unifies conformance checking, predictive monitoring, and quantitative scenario analysis within a singular, closed-loop structure. Using event logs from an Iranian Radiotherapy and Oncology Center, we modeled a normative seven-step pathway (Fitness = 0.97, Precision = 1.00) and identified high-impact deviations, including skipped approvals and resequencing, enabling a direct causal linkage between deviation categories and system performance. PIM simulation demonstrated that removing these deviations yields statistically significant reductions in managerially relevant KPIs: Cycle Time (8.00%) and Workload (6.00%), which were robust to parameter uncertainty (p < 0.001). The PM(2)-PIM framework thus transforms retrospective diagnosis into proactive, quantitatively justified strategic planning, providing oncology services with a reproducible, low-cost, and evidence-rich basis for prioritizing interventions and achieving sustained performance gains.