Evaluating Conformance and Performance in Oncology Pathways via Telemedicine Using Process Mining: A Study in a Chilean Hospital Network

Tele-oncology expanded rapidly during the COVID-19 pandemic and remains central to cancer care, yet evidence on guideline adherence in tele-oncology is limited, particularly in Latin America. In Chile, the Explicit Health Guarantees (GES) program mandates oncology care standards but provides no process model for monitoring adherence. We evaluated the conformance and temporal performance of telemedicine-based oncology pathways against a guideline-derived reference model using process mining, based on event logs from 182 patients treated in 2020–2023 in a Chilean hospital network. Conformance was computed using PM4Py with token-based replay and alignment-based fitness, complemented by cohort, abstraction-level and robustness analyses with bootstrap confidence intervals and multiple-comparison-adjusted testing. Overall conformance was low (4.95% of cases fully conforming; mean alignment fitness 0.19, 95% CI 0.16–0.22) against a highly precise reference model (precision 0.90). Censoring-aware analyses showed that low conformance was driven predominantly by genuine structural deviations rather than incomplete observation alone. Event-level cohort signatures were observed, but patient-level analyses suggested that these patterns should be interpreted as exploratory. Crude differences in care duration by out-of-model activity status were largely explained by differential observation time after exposure adjustment. Process-mining conformance checking provides reproducible indicators for governing tele-oncology pathways when metrics are interpreted alongside methodological caveats.