ACL Findings2026
MedCPI: A Construct-Personalize-Integrate Framework for KG-enhanced Clinical Prediction
Abstract
Electronic health records (EHRs) provide longitudinal evidence for clinical prediction, but EHR data are sparse, incomplete, and heterogeneous, which can limit robustness. Medical knowledge graphs (MKGs) have therefore been incorporated to support KG-enhanced clinical prediction by linking heterogeneous EHR codes to shared medical concepts via structured relations. However, existing KG-enhanced approaches remain limited in two aspects: (i) taskspecific knowledge selection when extracting knowledge from a large multi-source MKG; and (ii) patient-level personalization and knowledge integration, where personalization is often weakly controlled and knowledge integration is not sufficiently aligned with longitudinal patient trajectories. To address these issues, we propose MedCPI, a unified Construct-Personalize-Integrate framework. Med-CPI first performs task-guided schema induction and KG normalization to build a taskspecific Concept MKG as a denoised knowledge pool, then constructs controlled patientlevel PKGs via local expansion and short path search, and finally integrates PKG representations with time-aware EHR representations via cross-attention for prediction. Experiments on MIMIC-III and MIMIC-IV across four clinical prediction tasks show consistent improvements over strong EHR-only and KG-enhanced baselines. Ablations and additional analyses further validate the contribution of each stage and illustrate how MedCPI utilizes structured medical knowledge.
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
Lineage
- GraphCare: Enhancing Healthcare Predictions with Personalized Knowledge Graphs cites ConCare: Personalized Clinical Feature Embedding via Capturing the Healthcare Context
- GraphCare: Enhancing Healthcare Predictions with Personalized Knowledge Graphs cites Learning the Graphical Structure of Electronic Health Records with Graph Convolutional Transformer
Builds on8
- Learning the Graphical Structure of Electronic Health Records with Graph Convolutional TransformerEdward Choi, Zhen Xu, Yujia Li, Michael Dusenberry et al.AAAI 2020 · 293 citations
- ConCare: Personalized Clinical Feature Embedding via Capturing the Healthcare ContextLiantao Ma, Chaohe Zhang, Yasha Wang, Wenjie Ruan et al.AAAI 2020 · 190 citations
- HiTANet: Hierarchical Time-Aware Attention Networks for Risk Prediction on Electronic Health RecordsJunyu Luo, Muchao Ye, Cao Xiao, Fenglong MaKDD 2020 · 187 citations
- GraphCare: Enhancing Healthcare Predictions with Personalized Knowledge GraphsPengcheng Jiang, Cao Xiao, Adam Cross, Jimeng SunICLR 2024 · 77 citations
- Hierarchical Attention Propagation for Healthcare Representation LearningMuhan Zhang, Christopher Ryan King, Michael Avidan, Yixin ChenKDD 2020 · 44 citations
Related papers
- Trajectory-Aware Clinical Risk Prediction via Severity-Grounded Knowledge Graphs and Retrieval-Augmented GenerationKyunghoon Jeon, Youmin Ko, Woohwan Jung, Hyunjoon KimKDD 2026
- MedPath: Augmenting Health Risk Prediction via Medical Knowledge PathsMuchao Ye, Suhan Cui, Yaqing Wang, Junyu Luo et al.WWW 2021 · 77 citations
- KerPrint: Local-Global Knowledge Graph Enhanced Diagnosis Prediction for Retrospective and Prospective InterpretationsKai Yang, Yongxin Xu, Peinie Zou, Hongxin Ding et al.AAAI 2023 · 29 citations
- Knowledge-Enhanced Explainable Hypergraph Convolution Network for Medication RecommendationZihan Zhang, Hongzhi Liu, Xiaoshuang Guo, Tianqi Sun et al.AAAI 2026
- Beyond Traditional Diagnostics: Transforming Patient-Side Information Into Predictive Insights with Knowledge Graphs and PrototypesYibowen Zhao, Yinan Zhang, Zhixiang Su, Li-Zhen Cui et al.ICDE 2026
