Elamur Research⌕
☰
home / papers / 2026 / MedCPI: A Construct–Personalize–Integrate Framework for KG-enhanced Clinical Prediction

MedCPI: A Construct–Personalize–Integrate Framework for KG-enhanced Clinical Prediction

Paper · 2026 · Hang Wang, Hang Dong and Lu Liu

Summary

Clinical prediction from electronic health records (EHRs) suffers because the records are sparse, incomplete and heterogeneous. Medical knowledge graphs (MKGs) can help, but prior KG-enhanced methods draw on large multi-source graphs with redundant, overlapping relations and do not select knowledge for a specific task. Their per-patient graphs are also built loosely and fused poorly with visit histories. MedCPI addresses this in three stages:

  1. Construct: prompted by a task description, it induces a relation schema, maps or discards the original relation types, and extracts a compact task-specific Concept MKG.
  2. Personalize: it builds a personalized knowledge graph (PKG) for each patient through bounded local expansion and short path search, then encodes it with a task-aware R-GCN.
  3. Integrate: it fuses the PKG nodes with time-aware EHR encodings through cross-attention.

On MIMIC-III and MIMIC-IV, across four prediction tasks, MedCPI beats EHR-only, static-KG and PKG-reasoning baselines. Ablations show that each stage contributes. The work suggests that tailoring knowledge to the task, and aligning it with the patient's timeline, matters more than simply injecting more graph knowledge.

Key points

  • EHR codes (diagnoses, procedures, medications) are linked to UMLS concepts using SapBERT. Baselines include RETAIN, Dipole, HiTANet, GRAM, KAME, GAMENet, DKEC, KGxDP, KerPrint and GraphCare.
  • For in-hospital mortality, the full model reaches an AUROC of 0.713 on MIMIC-III and 0.809 on MIMIC-IV. Without Construct, the MIMIC-III score falls to 0.684, and the EHR-only variant falls to 0.655.
  • Allowing knowledge paths of up to 2 hops works best. Longer paths gradually hurt AUROC on both datasets, so 2 is used throughout.
  • Cross-attention patterns differ by task. Mortality favours has_complication and treated_by, while T2DM onset favours has_risk_factor and diagnosed_by, which gives some interpretability.
  • The stated limitations are fixed personalization constraints, structured data only (no clinical notes), and evaluation on MIMIC alone. The authors note the method is not meant to be a standalone clinical decision-making tool.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓University of ExeterExeter · United Kingdom12 works · 203 citations

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksMedicine · 241 worksHealth Informatics · 184 works

Cite

Cite as Hang Wang, Hang Dong and Lu Liu. 2026. MedCPI: A Construct–Personalize–Integrate Framework for KG-enhanced Clinical Prediction. Findings of the Association for Computational Linguistics: ACL 2026.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.findings-acl.1215

@inproceedings{wang2026medcpi,
  title = {{MedCPI: A Construct–Personalize–Integrate Framework for KG-enhanced Clinical Prediction}},
  author = {Hang Wang, Hang Dong and Lu Liu},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.1215},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.1215}
}

References

35 references, 5 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Ali Amirahmadi, Mattias Ohlsson, and Kobra Etminani. 2023. Deep learning prediction models based on ehr trajectories: A systematic review. Journal of biomedical informatics, 144:104430. https://physionet.org/ ↗
  2. Olivier Bodenreider. 2004. The unified medical language system (umls): integrating biomedical terminology. Nucleic acids research, 32(suppl_1):D267– D270. and Roger G Mark. 2016. Mimic-iii, a freely accessible critical care database. Scientific data, 3(1):1–9.
  3. Kyunghyun Cho, Bart Van Merriënboer, Caglar Gulcehre, Dzmitry Bahdanau, Fethi Bougares, Holger Schwenk, and Yoshua Bengio. 2014. Learning phrase representations using rnn encoder-decoder for statistical machine translation. arXiv preprint arXiv:1406.1078.
  4. Zeljko Kraljevic, Dan Bean, Anthony Shek, Rebecca Bendayan, Harry Hemingway, Joshua Au Yeung, Alexander Deng, Alfred Balston, Jack Ross, Esther Idowu, and 1 others. 2024. Foresight—a generative pretrained transformer for modelling of patient timelines using electronic health records: a retrospective… find
  5. Edward Choi, Mohammad Taha Bahadori, Le Song, Walter F Stewart, and Jimeng Sun. 2017. Gram: graph-based attention model for healthcare representation learning. In Proceedings of the 23rd ACM SIGKDD international conference on knowledge discovery and data mining, pages 787–795. find
  6. Isotta Landi, Benjamin S Glicksberg, Hao-Chih Lee, Sarah Cherng, Giulia Landi, Matteo Danieletto, Joel T Dudley, Cesare Furlanello, and Riccardo Miotto. 2020. Deep representation learning of electronic health records to unlock patient stratification at scale. NPJ digital medicine, 3(1):96. find
  7. Edward Choi, Mohammad Taha Bahadori, Jimeng Sun, Joshua Kulas, Andy Schuetz, and Walter Stewart. 2016. Retain: An interpretable predictive model for healthcare using reverse time attention mechanism. Advances in neural information processing systems, 29. find
  8. Jin Li, Benjamin J Cairns, Jingsong Li, and Tingting Zhu. 2023. Generating synthetic mixed-type longitudinal electronic health records for artificial intelligent applications. NPJ digital medicine, 6(1):98. find
  9. Edward Choi, Cao Xiao, Walter Stewart, and Jimeng Sun. 2018. Mime: Multilevel medical embedding of electronic health records for predictive healthcare. Advances in neural information processing systems, 31. find
  10. Edward Choi, Zhen Xu, Yujia Li, Michael Dusenberry, Gerardo Flores, Emily Xue, and Andrew Dai. 2020. Learning the graphical structure of electronic health records with graph convolutional transformer. In Proceedings of the AAAI conference on artificial intelligence, volume 34, pages 606–613. find
  11. Junyoung Chung, Caglar Gulcehre, KyungHyun Cho, and Yoshua Bengio. 2014. Empirical evaluation of gated recurrent neural networks on sequence modeling. arXiv preprint arXiv:1412.3555. find
  12. Xueren Ge, Abhishek Satpathy, Ronald Dean Williams, John Stankovic, and Homa Alemzadeh. 2024. Dkec: domain knowledge enhanced multi-label classification for diagnosis prediction. In Proceedings of the 2024 conference on empirical methods in natural language processing, pages 12798–12813. find
  13. Pengcheng Jiang, Cao Xiao, Adam Cross, and Jimeng Sun. 2024. Graphcare: Enhancing healthcare predictions with personalized knowledge graphs. In The Twelfth International Conference on Learning Representations. find
  14. Alistair EW Johnson, Lucas Bulgarelli, Lu Shen, Alvin Gayles, Ayad Shammout, Steven Horng, Tom J Pollard, Sicheng Hao, Benjamin Moody, Brian Gow, and 1 others. 2023. Mimic-iv, a freely accessible electronic health record dataset. Scientific data, 10(1):1.
  15. Alistair EW Johnson, Tom J Pollard, Lu Shen, Li-wei H Lehman, Mengling Feng, Mohammad Ghassemi, Benjamin Moody, Peter Szolovits, Leo Anthony Celi, Yikuan Li, Mohammad Mamouei, Gholamreza Salimi-Khorshidi, Shishir Rao, Abdelaali Hassaine, Dexter Canoy, Thomas Lukasiewicz, and Kazem Rahimi. 2022. Hi-b… find
  16. Fangyu Liu, Ehsan Shareghi, Zaiqiao Meng, Marco Basaldella, and Nigel Collier. 2021. Self-alignment pretraining for biomedical entity representations. In Proceedings of the 2021 conference of the North American chapter of the association for computational linguistics: human language technologies, pa…
  17. Junyu Luo, Muchao Ye, Cao Xiao, and Fenglong Ma. 2020. Hitanet: Hierarchical time-aware attention networks for risk prediction on electronic health records. In Proceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data mining, pages 647–656. find
  18. Fenglong Ma, Radha Chitta, Jing Zhou, Quanzeng You, Tong Sun, and Jing Gao. 2017. Dipole: Diagnosis prediction in healthcare via attention-based bidirectional recurrent neural networks. In Proceedings of the 23rd ACM SIGKDD international conference on knowledge discovery and data mining, pages 1903–… find
  19. Fenglong Ma, Quanzeng You, Houping Xiao, Radha Chitta, Jing Zhou, and Jing Gao. 2018. Kame: Knowledge-based attention model for diagnosis prediction in healthcare. In Proceedings of the 27th ACM international conference on information and knowledge management, pages 743–752. find
  20. Liantao Ma, Chaohe Zhang, Yasha Wang, Wenjie Ruan, Jiangtao Wang, Wen Tang, Xinyu Ma, Xin Gao, and Junyi Gao. 2020. Concare: Personalized clinical feature embedding via capturing the healthcare context. In Proceedings of the AAAI conference on artificial intelligence, volume 34, pages 833–840. find
  21. Chengsheng Mao, Liang Yao, and Yuan Luo. 2022. Medgcn: Medication recommendation and lab test imputation via graph convolutional networks. Journal of Biomedical Informatics, 127:104000. find
  22. Simon McIntosh-Smith, S. R. Alam, and Christopher Woods. 2024. Isambard-ai: a leadership class supercomputer optimised specifically for artificial intelligence. arXiv preprint arXiv:2410.11199. find
  23. OpenAI. 2025. Gpt-5 system card. https://cdn.op enai.com/gpt-5-system-card.pdf. ↗
  24. Alvin Rajkomar, Eyal Oren, Kai Chen, Andrew M Dai, Nissan Hajaj, Michaela Hardt, Peter J Liu, Xiaobing Liu, Jake Marcus, Mimi Sun, and 1 others. 2018. Scalable and accurate deep learning with electronic health records. NPJ digital medicine, 1(1):18. find
  25. Michael Schlichtkrull, Thomas N Kipf, Peter Bloem, Rianne Van Den Berg, Ivan Titov, and Max Welling. 2018. Modeling relational data with graph convolutional networks. In European semantic web conference, pages 593–607. Springer. find
  26. Kai Yang, Yongxin Xu, Peinie Zou, Hongxin Ding, Junfeng Zhao, Yasha Wang, and Bing Xie. 2023a. Kerprint: local-global knowledge graph enhanced diagnosis prediction for retrospective and prospective interpretations. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 37, pages 53… find
  27. Zongbao Yang, Yuchen Lin, Yinxin Xu, Jinlong Hu, and Shoubin Dong. 2023b. Interpretable disease prediction via path reasoning over medical knowledge graphs and admission history. Knowledge-Based Systems, 281:111082. find
  28. Zhihao Yu, Chu Xu, Yujie Jin, Yasha Wang, and Junfeng Zhao. 2024. Smart: Towards pre-trained missing-aware model for patient health status prediction. Advances in Neural Information Processing Systems, 37:63986–64009. find
  29. Muhan Zhang, Christopher R King, Michael Avidan, and Yixin Chen. 2020. Hierarchical attention propagation for healthcare representation learning. In Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, pages 249–256. find
  30. Junyuan Shang, Tengfei Ma, Cao Xiao, and Jimeng Sun. 2019a. Pre-training of graph augmented transformers for medication recommendation. arXiv preprint arXiv:1906.00346. find
  31. Junyuan Shang, Cao Xiao, Tengfei Ma, Hongyan Li, and Jimeng Sun. 2019b. Gamenet: Graph augmented memory networks for recommending medication combination. In proceedings of the AAAI Conference on Artificial Intelligence, volume 33, pages 1126– 1133. find
  32. Nenad Tomašev, Xavier Glorot, Jack W Rae, Michal Zielinski, Harry Askham, Andre Saraiva, Anne Mottram, Clemens Meyer, Suman Ravuri, Ivan Protsyuk, and 1 others. 2019. A clinically applicable approach to continuous prediction of future acute kidney injury. Nature, 572(7767):116–119. find
  33. Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. Advances in neural information processing systems, 30.
  34. Cao Xiao, Edward Choi, and Jimeng Sun. 2018. Opportunities and challenges in developing deep learning models using electronic health records data: a systematic review. Journal of the American Medical Informatics Association, 25(10):1419–1428. find
  35. Yongxin Xu, Xu Chu, Kai Yang, Zhiyuan Wang, Peinie Zou, Hongxin Ding, Junfeng Zhao, Yasha Wang, and Bing Xie. 2023. Seqcare: Sequential training with external medical knowledge graph for diagnosis prediction in healthcare data. In Proceedings of the ACM Web Conference 2023, pages 2819–2830. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/wang-2026-medcpi-construct-personalize-integrate-framework/