Discovering heart failure patient pathways through event enrichment and abstraction

Integration of electronic health data from patients with heart failure into process mining could provide insights to the patient journey. However, the current approaches have challenges in addressing all aspects of patient care.

A retrospective cohort study of 234 patients with heart failure who were admitted to an outpatient heart failure clinic was conducted. We developed a two-stage pipeline to integrate multidimensional data aspects, such as patient-reported outcome measures, biomarkers, medication changes and reasons for hospitalization. The pipeline consists of: (1) Data Enrichment, where we derive indicators from disparate sources like outpatient visits and self-reported data and we structure this information into an event log; and (2) Supervised Event Abstraction where low-level events are translated into clinically meaningful concepts using a knowledge graph. We demonstrate the practical integration of clinical event enrichment and knowledge-graph-based abstraction for exploratory heart failure pathway analysis.

We identified four clusters based on patients' backgrounds. Group A included older, multi-morbid patients with advanced HF and more frequent HF hospitalizations and renal-function worsening, while Groups B, C, and D represented ischemic, arrhythmia-associated, and younger lower-comorbidity profiles with simpler care pathways.

These findings demonstrate that this process mining approach offers a practical framework to better understand specific patient pathways and identify mitigation strategies for adverse disease trajectories and mortality. It allows for a more granular understanding of patient journeys and the identification of at-risk cohorts by integrating clinical and patient-reported outcomes. This is a significant step towards more data-driven, patient-centered healthcare.