MIPO: Mutual Integration of Patient Journey and Medical Ontology for Healthcare Representation Learning

TL;DR

MIPO addresses the challenge of learning robust, interpretable representations from EHRs under data scarcity by mutually integrating patient journeys with medical ontologies. It introduces a hierarchical Transformer-based framework with a graph-embedding module, comprising a V-Encoder for visit-level fusion and a P-Encoder for sequence-level modeling, guided by two predictive tasks: sequential diagnosis prediction and ontology-based disease typing, combined into a joint objective $\ L = L_P + L_V$. The approach yields superior predictive performance on MIMIC-III and eICU compared with baselines such as GRAM, KAME, and attention-based RNNs, and demonstrates robustness under limited data conditions. Interpretability is supported by qualitative analyses showing disease-wise clustering of learned embeddings that align with CCS ontology, underscoring MIPO’s potential for real-world healthcare deployment.

Abstract

Representation learning on electronic health records (EHRs) plays a vital role in downstream medical prediction tasks. Although natural language processing techniques, such as recurrent neural networks, and self-attention, have been adapted for learning medical representations from hierarchical, time-stamped EHR data, they often struggle when either general or task-specific data are limited. Recent efforts have attempted to mitigate this challenge by incorporating medical ontologies (i.e., knowledge graphs) into self-supervised tasks like diagnosis prediction. However, two main issues remain: (1) small and uniform ontologies that lack diversity for robust learning, and (2) insufficient attention to the critical contexts or dependencies underlying patient journeys, which could further enhance ontology-based learning. To address these gaps, we propose MIPO (Mutual Integration of Patient Journey and Medical Ontology), a robust end-to-end framework that employs a Transformer-based architecture for representation learning. MIPO emphasizes task-specific representation learning through a sequential diagnosis prediction task, while also incorporating an ontology-based disease-typing task. A graph-embedding module is introduced to integrate information from patient visit records, thus alleviating data insufficiency. This setup creates a mutually reinforcing loop, where both patient-journey embedding and ontology embedding benefit from each other. We validate MIPO on two real-world benchmark datasets, showing that it consistently outperforms baseline methods under both sufficient and limited data conditions. Furthermore, the resulting diagnosis embeddings offer improved interpretability, underscoring the promise of MIPO for real-world healthcare applications.

Figures (4)

• Figure 1: The model architecture of MIPO. The graph is formatted as a hierarchical tree, in which, the root node is virtual. To construct the tree, the leaf nodes (solid circles) denote fine-grained diagnoses, and the non-leaf nodes (dotted circles) denote coarse-grained disease concepts.
• Figure 2: Acc@20 of diagnoses prediction on MIMIC and eICU, size of training data is varied from 20% to 80%.
• Figure 3: Annotations of MIPO Diagnosis Embedding
• Figure 4: t-SNE Scatterplots of Medical Codes Learned by Predictive Models on the MIMIC dataset.