Hierarchy Pruning for Unseen Domain Discovery in Predictive Healthcare

Pengfei Hu, Xiaoxue Han, Fei Wang, Yue Ning

Keywords: EHR Prediction, Domain Generalization, Unseen Domain, Hierarchy Pruning, Domain Discovery

Abstract: Healthcare providers often divide patient populations into cohorts based on shared clinical factors, such as medical history, to deliver personalized healthcare services. This idea has also been adopted in clinical prediction models, where it presents a vital challenge: capturing both global and cohort-specific patterns while enabling model generalization to unseen domains. Since cohort boundaries naturally correspond to domain boundaries, addressing this challenge falls under the scope of domain generalization (DG), especially when domain labels are not explicitly available in EHR data. However, regular DG approaches often struggle in clinical settings due to the absence of domain labels and the inherent gap in medical knowledge. Moreover, the rich hierarchical structures embedded in medical ontologies have rarely been explored as a basis for deriving clinically meaningful domain partitions. Hence, we propose UdonCare, a hierarchy-guided method that iteratively divides patients into latent domains and decomposes domain-invariant (label) information from patient data. On two public datasets, MIMIC-III and MIMIC-IV, UdonCare shows superiority over eight baselines across four clinical prediction tasks with substantial domain gaps, highlighting the potential of medical knowledge in guiding clinical DG problems.

Primary Area: applications to physical sciences (physics, chemistry, biology, etc.)

Submission Number: 20546