Abstract
The widespread adoption of electronic health records has created new opportunities for translational clinical research, yet this promise remains constrained by fragmented data across privacy-siloed institutions and substantial heterogeneity in local coding practices. While privacy-preserving collaborative learning allows institutions to work together without sharing patient-level data, it does not address inconsistencies in how clinical concepts are represented across sites. We introduce a graph-based framework that addresses this gap by treating data harmonization as a scalable representation learning problem. Rather than relying on fixed standards or manual mappings, the framework integrates institution-specific summary statistics from health records, curated biomedical knowledge graphs, and semantic information derived from large language models to learn a shared semantic space. This joint learning approach aligns diverse, site-specific vocabularies while preserving patient privacy. Evaluated across seven institutions and two languages, the framework provides a robust, data-centric foundation for training and deploying clinical models across heterogeneous healthcare systems.
Data availability
Source data are provided in the Source data file. The minimum dataset required to interpret and reproduce the main findings of this study consists of the pairwise cosine similarity data derived from the GAME embeddings, which are publicly available at https://shiny.parse-health.org/GAME/. Institution-level summary data are available under restricted access due to data use agreements (DUAs) with the participating healthcare institutions. Access to these data may be obtained by submitting a request and establishing a DUA with the relevant institution(s); requests should be directed to the corresponding author. Individual-level clinical data used for downstream analyses are not publicly available due to patient privacy protections, ethical approval constraints, and institutional regulations. These data were accessed under institution-specific DUAs and IRB approvals and cannot be shared beyond those agreements. Requests for restricted data are reviewed by the relevant institution(s), with an expected response time of approximately 4–8 weeks. The duration of access and permitted use are governed by the terms of the corresponding DUA. Source data are provided with this paper.
Code availability
The code used in this study is publicly available at https://github.com/celehs/GAME. The specific version used in this study has been archived on Zenodo at https://doi.org/10.5281/zenodo.18222787, ensuring long-term accessibility and reproducibility69. An interactive visualization of the knowledge graph derived from the GAME embeddings is available at https://shiny.parse-health.org/GAME/.
References
Liao, K. P. et al. Development of phenotype algorithms using electronic medical records and incorporating natural languageprocessing. BMJ 350, h1885 https://doi.org/10.1136/bmj.h1885 (2015).
Wang, L. et al. Stratification of Alzheimer’s disease patients using knowledge-guided unsupervised latent factor clustering with electronic health record data. Preprint at Dec 26 https://doi.org/10.1101/2024.12.23.24319588 (2024).
Doshi-Velez, F., Ge, Y. & Kohane, I. Comorbidity clusters in autism spectrum disorders: an electronic health record time-series analysis. Pediatrics 133, e54–e63 (2014).
Sheu, Y. -h. et al. An efficient landmark model for prediction of suicide attempts in multiple clinical settings. Psychiatry Res. 323, 115175 (2023).
Federico, P. et al. Gnaeus: Utilizing clinical guidelines for knowledge-assisted visualisation of EHR cohorts. In Roberts, J. C. & Bertini, E. (eds.) 6th International EuroVis Workshop on Visual Analytics, EuroVA@EuroVis 2015, Cagliari, Sardinia, Italy, May 25-26, 2015, 79–83 (Eurographics Association, 2015).
Ferté, T., Jouhet, V., Griffier, R., Hejblum, B. P. & Thiébaut, R. The benefit of augmenting open data with clinical data-warehouse EHR for forecasting SARS-CoV-2 hospitalizations in Bordeaux area, France. JAMIA Open 5, ooac086 (2022).
Wen, J. et al. Multimodal representation learning for predicting molecule–disease relations. Bioinformatics 39, btad085 (2023).
Cai, T., Xia, D., Zhang, L. & Zhou, D. Consensus knowledge graph learning via multi-view sparse low rank block model. Preprint at https://doi.org/10.48550/arXiv.2209.13762 (2022).
Hur, K. et al. Unifying heterogeneous electronic health records systems via text-based code embedding. In Proc. Conference on Health, Inference, and Learning, Vol. 174 of Proc. of Machine Learning Research, (eds. Flores, G., Chen, G. H., Pollard, T., Ho, J. C. & Naumann, T.) 183–203 (PMLR, 2022).
Molaei, S. et al. Federated learning for heterogeneous electronic health records utilising augmented temporal graph attention networks. In Proc. International Conference on Artificial Intelligence and Statistics, 1342–1350 (PMLR, 2024).
Thakur, A. et al. Knowledge abstraction and filtering based federated learning over heterogeneous data views in healthcare. NPJ Digit. Med. 7, 283 (2024).
Centre for Disease Control and Prevention et al. International classification of diseases, ninth revision (ICD-9). Cincinnati, Ohio: National Center for Health Statistics (1979).
McDonald, C. J. et al. Loinc, a universal standard for identifying laboratory observations: a 5-year update. Clin. Chem. 49, 624–633 (2003).
Chen, M. et al. Privacy protection and intrusion avoidance for cloudlet-based medical data sharing. IEEE Trans. Cloud Comput. 8, 1274–1283 (2016).
Sheller, M. et al. Federated learning in medicine: facilitating multi-institutional collaborations without sharing patient data. Sci. Rep. 10, (2020).
Mikolov, T., Sutskever, I., Chen, K., Corrado, G. S. & Dean, J. Distributed representations of words and phrases and their compositionality. Adv. Neural Inf. Process. Syst. 26, (2013).
Pennington, J., Socher, R. & Manning, C. D. Glove: global vectors for word representation. In Proc. 2014 Conference on Empirical Methods in Natural Language Processing, 1532–1543 (ACL, 2014).
Wang, Z., Zhang, J., Feng, J. & Chen, Z. Knowledge graph embedding by translating on hyperplanes. In Proc. AAAI Conference on Artificial Intelligence, Vol. 28 (AAAI, 2014).
Balažević, I., Allen, C. & Hospedales, T. Tucker: tensor factorization for knowledge graph completion. In Proc. 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), 5185–5194 (ACL, 2019).
Yuan, Z. et al. CODER: knowledge-infused cross-lingual medical term embedding for term normalization. J. Biomed. Inform. 126, 103983 (2022).
Lin, Y., Lu, K., Yu, S., Cai, T. & Zitnik, M. Multimodal learning on graphs for disease relation extraction. J. Biomed. Inform. 143, 104415 (2023).
Bodenreider, O. The Unified Medical Language System (UMLS): integrating biomedical terminology. Nucleic Acids Res. 32, D267–D270 (2004).
Liu, F., Shareghi, E., Meng, Z., Basaldella, M. & Collier, N. Self-alignment pretraining for biomedical entity representations. In Proc. 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, 4228–4238 (ACL, 2021).
Maldonado, R., Yetisgen, M. & Harabagiu, S. M. Adversarial learning of knowledge embeddings for the Unified Medical Language System. AMIA Summits Transl. Sci. Proc. 2019, 543 (2019).
Michalopoulos, G., Wang, Y., Kaka, H., Chen, H. & Wong, A. UmlsBERT: clinical domain knowledge augmentation of contextual embeddings using the Unified Medical Language System Metathesaurus. In Proc. 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, 1744–1753 (ACL, 2021).
Piya, F. L., Gupta, M. & Beheshti, R. HealthGAT: node classifications in electronic health records using graph attention networks. In Proc. 2024 IEEE/ACM Conference on Connected Health: Applications, Systems and Engineering Technologies (CHASE), 132–141 (IEEE, 2024).
Choi, E. et al. Multi-layer representation learning for medical concepts. In Proc. 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 1495–1504 (ACM, 2016).
Kartchner, D., Christensen, T., Humpherys, J. & Wade, S. Code2vec: embedding and clustering medical diagnosis data. In Proc. 2017 IEEE International Conference on Healthcare Informatics, 386–390 (IEEE, 2017).
Hong, C. et al. Clinical knowledge extraction via sparse embedding regression (KESER) with multi-center large scale electronic health record data. NPJ Digit. Med. 4, 151 (2021).
Zhou, D. et al. Multiview incomplete knowledge graph integration with application to cross-institutional EHR data harmonization. J. Biomed. Inform. 133, 104147 (2022).
Gan, Z. et al. ARCH: large-scale knowledge graph via aggregated narrative codified health records analysis. J. Biomed. Inform. 162, 104761 (2025).
Wang, K., Chen, N. & Chen, T. Joint medical ontology representation learning for healthcare predictions. In Proc. 2020 International Joint Conference on Neural Networks (IJCNN), 1–7 (IEEE, 2020).
Ying, H., Zhao, Z., Zhao, Y., Zeng, S. & Yu, S. CoRTEx: contrastive learning for representing terms via explanations with applications on constructing biomedical knowledge graphs. J. Am. Med. Inform. Assoc. 31, 1912–1920 (2024).
Gao, Y. et al. Leveraging medical knowledge graphs into large language models for diagnosis prediction: design and application study. JMIR AI 4, e58670 (2025).
Cai, T., Huang, F., Nakada, R., Zhang, L. & Zhou, D. Contrastive learning on multimodal analysis of electronic health records. Preprint at https://doi.org/10.48550/arXiv.2403.14926 (2024).
Levy, O. & Goldberg, Y. Neural word embedding as implicit matrix factorization. Adv. Neural Inf. Process. Syst. 27, (2014).
Gu, Y. et al. Domain-specific language model pretraining for biomedical natural language processing. ACM Trans. Comput. Healthc. 3, 1–23 (2021).
Lee, J. et al. Biobert: a pre-trained biomedical language representation model for biomedical text mining. Bioinformatics 36, 1234–1240 (2020).
Chen, J. et al. M3-embedding: Multi-linguality, multi-functionality, multi-granularity text embeddings through self-knowledge distillation. Findings of the Association for Computational Linguistics: ACL 2024, 2318–2335 (Association for Computational Linguistics, Bangkok, Thailand, 2024).
Cipriani, A. et al. Comparative efficacy and acceptability of antimanic drugs in acute mania: a multiple-treatments meta-analysis. Lancet 378, 1306–1315 (2011).
Arvanitis, L. A. & Miller, B. G. Multiple fixed doses of “Seroquel” (Quetiapine) in patients with acute exacerbation of schizophrenia: a comparison with Haloperidol and placebo. Biol. Psychiatry 42, 233–246 (1997).
Ismail, Z. et al. Psychosis in Alzheimer disease-mechanisms, genetics and therapeutic opportunities. Nat. Rev. Neurol. 18, 131–144 (2022).
Liu, J., Chang, L., Song, Y., Li, H. & Wu, Y. The role of NMDA receptors in Alzheimer’s disease. Front. Neurosci. 13, 43 (2019).
Tariot, P. N. et al. Memantine treatment in patients with moderate to severe Alzheimer disease already receiving donepezil: a randomized controlled trial. J. Am. Med. Inform. Assoc. 291, 317–324 (2004).
Anthropic. Introducing the next generation of claude https://www.anthropic.com/news/claude-3-family (2024).
Meta AI. The LLaMA 4 herd: The beginning of a new era of natively multimodal AI innovation (2025) https://ai.meta.com/blog/llama-4-multimodal-intelligence/ Accessed: 2025-Apr-07.
OMOP. Standardized data: The OMOP common data model (2021) https://www.ohdsi.org/data-standardization/. Accessed: Jun, 2025.
Wen, J. et al. DOME: directional medical embedding vectors from electronic health records. J. Biomed. Inform. 162, 104768 (2025).
Chen, L. et al. Graph optimal transport for cross-domain alignment. In Proc. International Conference on Machine Learning, 1542–1553 (PMLR, 2020).
Veličković, P. et al. Graph attention networks. In Proc. International Conference on Learning Representations (ICLR, 2018).
Gori, M., Monfardini, G. & Scarselli, F. A new model for learning in graph domains. In Proc. 2005 IEEE International Joint Conference on Neural Networks, Vol. 2, 729–734 (IEEE, 2005).
Johnson, A. et al. MIMIC-IV (version 0.4). PhysioNet. (2020) https://physionet.org/content/mimiciv/0.4/. Accessed: June, 2025.
Bousquet, C., Trombert, B., Souvignet, J., Sadou, E. & Rodrigues, J.-M. Evaluation of the CCAM hierarchy and semi structured code for retrieving relevant procedures in a hospital case mix database. In Proc.AMIA Annual Symposium Proceedings, Vol. 2010, 61 (AMIA, 2010).
Beam, A. L. et al. Clinical concept embeddings learned from massive sources of multimodal medical data. In Proc. Pacific Symposium on Biocomputing, Vol. 25, 295–306 (PSB, 2020).
Shin, H.-C. et al. BioMegatron: larger biomedical domain language model. In Proc. 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), 4700–4706 (ACL, 2020).
Wang, X., Han, X., Huang, W., Dong, D. & Scott, M. R. Multi-similarity loss with general pair weighting for deep metric learning. In Proc. IEEE/CVF Conference on Computer Vision and Pattern Recognition, 5022–5030 (IEEE Computer Society, 2019).
Li, L. et al. Identification of type 2 diabetes subgroups through topological analysis of patient similarity. Sci. Transl. Med. 7, 311ra174–311ra174 (2015).
Landi, I. et al. Deep representation learning of electronic health records to unlock patient stratification at scale. NPJ Digit. Med. 3, 96 (2020).
Van der Maaten, L. & Hinton, G. Visualizing data using t-SNE. J. Mach. Learn. Res. 9, (2008).
Garst, S. & Reinders, M. Federated k-means clustering. In Proc.International Conference on Pattern Recognition, 107–122 (Springer, 2024).
Armstrong, M. J., Song, S., Kurasz, A. M. & Li, Z. Predictors of mortality in individuals with dementia in the National Alzheimer’s Coordinating Center. J. Alzheimer’s. Dis. 86, 1935–1946 (2022).
Zheng, X., Wang, S., Huang, J., Li, C. & Shang, H. Predictors for survival in patients with Alzheimer’s disease: a large comprehensive meta-analysis. Transl. Psychiatry 14, 184 (2024).
Abdelnour, C. et al. Perspectives and challenges in patient stratification in Alzheimer’s disease. Alzheimer’s. Res. Ther. 14, 112 (2022).
Han, E., Kharrazi, H., Shi, L. et al. Identifying predictors of nursing home admission by using electronic health records and administrative data: scoping review. JMIR Aging 6, e42437 (2023).
Favril, L., Yu, R., Uyar, A., Sharpe, M. & Fazel, S. Risk factors for suicide in adults: systematic review and meta-analysis of psychological autopsy studies. BMJ Ment. Health 25, 148–155 (2022).
Sutar, R., Kumar, A. & Yadav, V. Suicide and prevalence of mental disorders: a systematic review and meta-analysis of world data on case-control psychological autopsy studies. Psychiatry Res. 329, 115492 (2023).
Fazel, S. & Runeson, B. Suicide. N. Engl. J. Med. 382, 266–274 (2020).
Lee, D., Jiang, X. & Yu, H. Harmonized representation learning on dynamic EHR graphs. J. Biomed. Inform. 106, 103426 (2020).
Panickan, V. A., CELEHS & Tong, H. celehs/game: representation learning to advance multi-institutional studies with electronic health record data https://github.com/celehs/GAME (2026).
Acknowledgements
This research was supported by the Office of Research and Development, Veterans Health Administration, under award MVP000. This work also used resources of the Knowledge Discovery Infrastructure (KDI) at Oak Ridge National Laboratory, supported by the Office of Science of the U.S. Department of Energy under contract no. DE-AC05-00OR22725. The contents of this publication do not represent the views of the U.S. Department of Veterans Affairs or the United States Government. D.Z. was supported by the MOE AcRF Tier 1 grant A-8003569-00-00 and the NUS Start-up grant A-0009985-00-00. Z.X. was supported by NIH grant 5R01NS098023. K.L. was supported by NIH grants P30 AR072577 and K24 AR085342. T.C. was supported by NIH grants R01 LM013614, R01 HL089778, P30 AR072577, and P50 MH129699.
Author information
Authors and Affiliations
Contributions
D.Z., H.T., L.W., and S.L. conceived the study. D.Z., H.T. contributed to the methodology and model design. D.Z. conceptualized the study and H.T. led the implementation. H.T., L. W. conducted data analysis and validation experiments. S. L., X.X., and Z.G. contributed to data preprocessing and experimental evaluation. R.G., B.H., V.J., and R.T. contributed to the French institutional data and clinical interpretation. Y.-C.L., C.H. contributed to data extraction and site-specific implementation at Duke University. C.-L.B., V.A.P., K.P., and Z.X. assisted with data harmonization and result interpretation. T. R.C., Y.-L.H., L.C., J.M.G., and K.C. contributed to study design, clinical interpretation, and application development. K.M. contributed to informatics design and integration strategy. K.L., T.C. jointly supervised the study. All authors contributed to manuscript writing and approved the final version.
Corresponding authors
Ethics declarations
Competing interests
The authors declare no competing interests.
Peer review
Peer review information
Nature Communications thanks Majid Afshar and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Source data
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
About this article
Cite this article
Zhou, D., Tong, H., Wang, L. et al. Representation learning to advance multi-institutional studies with electronic health record data from US and France. Nat Commun (2026). https://doi.org/10.1038/s41467-026-71152-1
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41467-026-71152-1
