ARCH: Large-scale knowledge graph via aggregated narrative codified health records analysis

Ziming Gan; Doudou Zhou; Everett Rush; Vidul A. Panickan; Yuk Lam Ho; George Ostrouchovm; Zhiwei Xu; Shuting Shen; Xin Xiong; Kimberly F. Greco; Chuan Hong; Clara Lea Bonzel; Jun Wen; Lauren Costa; Tianrun Cai; Edmon Begoli; Zongqi Xia; J. Michael Gaziano; Katherine P. Liao; Kelly Cho; Tianxi Cai; Junwei Lu

doi:10.1016/j.jbi.2024.104761

ARCH: Large-scale knowledge graph via aggregated narrative codified health records analysis

Ziming Gan, Doudou Zhou, Everett Rush, Vidul A. Panickan, Yuk Lam Ho, George Ostrouchovm, Zhiwei Xu, Shuting Shen, Xin Xiong, Kimberly F. Greco, Chuan Hong, Clara Lea Bonzel, Jun Wen, Lauren Costa, Tianrun Cai, Edmon Begoli, Zongqi Xia, J. Michael Gaziano, Katherine P. Liao, Kelly ChoTianxi Cai, Junwei Lu

Cyber Resilience and Intelligence Divisi

Research output: Contribution to journal › Article › peer-review

Abstract

Objective: Electronic health record (EHR) systems contain a wealth of clinical data stored as both codified data and free-text narrative notes (NLP). The complexity of EHR presents challenges in feature representation, information extraction, and uncertainty quantification. To address these challenges, we proposed an efficient Aggregated naRrative Codified Health (ARCH) records analysis to generate a large-scale knowledge graph (KG) for a comprehensive set of EHR codified and narrative features. Methods: Using data from 12.5 million Veterans Affairs patients, ARCH first derives embedding vectors and generates similarities along with associated p-values to measure the strength of relatedness between clinical features with statistical certainty quantification. Next, ARCH performs a sparse embedding regression to remove indirect linkage between features to build a sparse KG. Finally, ARCH was validated on various clinical tasks, including detecting known relationships between entity pairs, predicting drug side effects, disease phenotyping, as well as sub-typing Alzheimer's disease patients. Results: ARCH produces high-quality clinical embeddings and KG for over 60,000 codified and narrative EHR concepts. The KG and embeddings are visualized in the R-shiny powered web-API. ARCH achieved high accuracy in detecting EHR concept relationships, with AUCs of 0.926 (codified) and 0.861 (NLP) for similar EHR concepts, and 0.810 (codified) and 0.843 (NLP) for related pairs. It detected drug side effects with a 0.723 AUC, which improved to 0.826 after fine-tuning. Using both codified and NLP features, the detection power increased significantly. Compared to other methods, ARCH has superior accuracy and enhances weakly supervised phenotyping algorithms’ performance. Notably, it successfully categorized Alzheimer's patients into two subgroups with varying mortality rates. Conclusion: The proposed ARCH algorithm generates large-scale high-quality semantic representations and knowledge graph for both codified and NLP EHR features, useful for a wide range of predictive modeling tasks.

Original language	English
Article number	104761
Journal	Journal of Biomedical Informatics
Volume	162
DOIs	https://doi.org/10.1016/j.jbi.2024.104761
State	Published - Feb 2025

Funding

We would like to acknowledge the invaluable contributions arising from the collaboration between Veterans Affairs (VA) and the Department of Energy (DOE) which provided the computing infrastructure essential to develop and test these approaches at scale with nationwide VA EHR data. This project was supported by the NIH grants 1OT2OD032581 , R01 HL089778 and R01 LM013614 , P30 AR072577 , and the Million Veteran Program, Department of Veterans Affairs, Office of Research and Development, Veterans Health Administration, and was supported by the award #MVP000. This research used resources from the Knowledge Discovery Infrastructure at Oak Ridge National Laboratory, which is supported by the Office of Science of the US Department of Energy under Contract No. DE-AC05-00OR22725. This publication does not represent the views of the Department of Veterans Affairs or the U.S. government. We would like to acknowledge the invaluable contributions arising from the collaboration between Veterans Affairs (VA) and the Department of Energy (DOE) which provided the computing infrastructure essential to develop and test these approaches at scale with nationwide VA EHR data. This project was supported by the NIH, United States grants 1OT2OD032581, R01 HL089778 and R01 LM013614, P30 AR072577, William F. Milton Fund, and the Million Veteran Program, Department of Veterans Affairs, Office of Research and Development, Veterans Health Administration, and was supported by the award #MVP000. This research used resources from the Knowledge Discovery Infrastructure at Oak Ridge National Laboratory, which is supported by the Office of Science of the US Department of Energy under Contract No. DE-AC05-00OR22725. This publication does not represent the views of the Department of Veterans Affairs or the U.S. government. Z Xia is supported in part by NINDS R01NS098023 and NINDS R01NS124882.

Keywords

Electronic health records
Knowledge graph
Natural language processing
Representation learning

Access to Document

10.1016/j.jbi.2024.104761

Cite this

Gan, Z., Zhou, D., Rush, E., Panickan, V. A., Ho, Y. L., Ostrouchovm, G., Xu, Z., Shen, S., Xiong, X., Greco, K. F., Hong, C., Bonzel, C. L., Wen, J., Costa, L., Cai, T., Begoli, E., Xia, Z., Gaziano, J. M., Liao, K. P., ... Lu, J. (2025). ARCH: Large-scale knowledge graph via aggregated narrative codified health records analysis. Journal of Biomedical Informatics, 162, Article 104761. https://doi.org/10.1016/j.jbi.2024.104761

@article{4dcd0df853d941bda424224209cf0c87,

title = "ARCH: Large-scale knowledge graph via aggregated narrative codified health records analysis",

abstract = "Objective: Electronic health record (EHR) systems contain a wealth of clinical data stored as both codified data and free-text narrative notes (NLP). The complexity of EHR presents challenges in feature representation, information extraction, and uncertainty quantification. To address these challenges, we proposed an efficient Aggregated naRrative Codified Health (ARCH) records analysis to generate a large-scale knowledge graph (KG) for a comprehensive set of EHR codified and narrative features. Methods: Using data from 12.5 million Veterans Affairs patients, ARCH first derives embedding vectors and generates similarities along with associated p-values to measure the strength of relatedness between clinical features with statistical certainty quantification. Next, ARCH performs a sparse embedding regression to remove indirect linkage between features to build a sparse KG. Finally, ARCH was validated on various clinical tasks, including detecting known relationships between entity pairs, predicting drug side effects, disease phenotyping, as well as sub-typing Alzheimer's disease patients. Results: ARCH produces high-quality clinical embeddings and KG for over 60,000 codified and narrative EHR concepts. The KG and embeddings are visualized in the R-shiny powered web-API. ARCH achieved high accuracy in detecting EHR concept relationships, with AUCs of 0.926 (codified) and 0.861 (NLP) for similar EHR concepts, and 0.810 (codified) and 0.843 (NLP) for related pairs. It detected drug side effects with a 0.723 AUC, which improved to 0.826 after fine-tuning. Using both codified and NLP features, the detection power increased significantly. Compared to other methods, ARCH has superior accuracy and enhances weakly supervised phenotyping algorithms{\textquoteright} performance. Notably, it successfully categorized Alzheimer's patients into two subgroups with varying mortality rates. Conclusion: The proposed ARCH algorithm generates large-scale high-quality semantic representations and knowledge graph for both codified and NLP EHR features, useful for a wide range of predictive modeling tasks.",

keywords = "Electronic health records, Knowledge graph, Natural language processing, Representation learning",

author = "Ziming Gan and Doudou Zhou and Everett Rush and Panickan, {Vidul A.} and Ho, {Yuk Lam} and George Ostrouchovm and Zhiwei Xu and Shuting Shen and Xin Xiong and Greco, {Kimberly F.} and Chuan Hong and Bonzel, {Clara Lea} and Jun Wen and Lauren Costa and Tianrun Cai and Edmon Begoli and Zongqi Xia and Gaziano, {J. Michael} and Liao, {Katherine P.} and Kelly Cho and Tianxi Cai and Junwei Lu",

note = "Publisher Copyright: {\textcopyright} 2025 Elsevier Inc.",

year = "2025",

month = feb,

doi = "10.1016/j.jbi.2024.104761",

language = "English",

volume = "162",

journal = "Journal of Biomedical Informatics",

issn = "1532-0464",

publisher = "Elsevier",

}

Gan, Z, Zhou, D, Rush, E, Panickan, VA, Ho, YL, Ostrouchovm, G, Xu, Z, Shen, S, Xiong, X, Greco, KF, Hong, C, Bonzel, CL, Wen, J, Costa, L, Cai, T, Begoli, E, Xia, Z, Gaziano, JM, Liao, KP, Cho, K, Cai, T & Lu, J 2025, 'ARCH: Large-scale knowledge graph via aggregated narrative codified health records analysis', Journal of Biomedical Informatics, vol. 162, 104761. https://doi.org/10.1016/j.jbi.2024.104761

TY - JOUR

T1 - ARCH

T2 - Large-scale knowledge graph via aggregated narrative codified health records analysis

AU - Gan, Ziming

AU - Zhou, Doudou

AU - Rush, Everett

AU - Panickan, Vidul A.

AU - Ho, Yuk Lam

AU - Ostrouchovm, George

AU - Xu, Zhiwei

AU - Shen, Shuting

AU - Xiong, Xin

AU - Greco, Kimberly F.

AU - Hong, Chuan

AU - Bonzel, Clara Lea

AU - Wen, Jun

AU - Costa, Lauren

AU - Cai, Tianrun

AU - Begoli, Edmon

AU - Xia, Zongqi

AU - Gaziano, J. Michael

AU - Liao, Katherine P.

AU - Cho, Kelly

AU - Cai, Tianxi

AU - Lu, Junwei

PY - 2025/2

Y1 - 2025/2

N2 - Objective: Electronic health record (EHR) systems contain a wealth of clinical data stored as both codified data and free-text narrative notes (NLP). The complexity of EHR presents challenges in feature representation, information extraction, and uncertainty quantification. To address these challenges, we proposed an efficient Aggregated naRrative Codified Health (ARCH) records analysis to generate a large-scale knowledge graph (KG) for a comprehensive set of EHR codified and narrative features. Methods: Using data from 12.5 million Veterans Affairs patients, ARCH first derives embedding vectors and generates similarities along with associated p-values to measure the strength of relatedness between clinical features with statistical certainty quantification. Next, ARCH performs a sparse embedding regression to remove indirect linkage between features to build a sparse KG. Finally, ARCH was validated on various clinical tasks, including detecting known relationships between entity pairs, predicting drug side effects, disease phenotyping, as well as sub-typing Alzheimer's disease patients. Results: ARCH produces high-quality clinical embeddings and KG for over 60,000 codified and narrative EHR concepts. The KG and embeddings are visualized in the R-shiny powered web-API. ARCH achieved high accuracy in detecting EHR concept relationships, with AUCs of 0.926 (codified) and 0.861 (NLP) for similar EHR concepts, and 0.810 (codified) and 0.843 (NLP) for related pairs. It detected drug side effects with a 0.723 AUC, which improved to 0.826 after fine-tuning. Using both codified and NLP features, the detection power increased significantly. Compared to other methods, ARCH has superior accuracy and enhances weakly supervised phenotyping algorithms’ performance. Notably, it successfully categorized Alzheimer's patients into two subgroups with varying mortality rates. Conclusion: The proposed ARCH algorithm generates large-scale high-quality semantic representations and knowledge graph for both codified and NLP EHR features, useful for a wide range of predictive modeling tasks.

AB - Objective: Electronic health record (EHR) systems contain a wealth of clinical data stored as both codified data and free-text narrative notes (NLP). The complexity of EHR presents challenges in feature representation, information extraction, and uncertainty quantification. To address these challenges, we proposed an efficient Aggregated naRrative Codified Health (ARCH) records analysis to generate a large-scale knowledge graph (KG) for a comprehensive set of EHR codified and narrative features. Methods: Using data from 12.5 million Veterans Affairs patients, ARCH first derives embedding vectors and generates similarities along with associated p-values to measure the strength of relatedness between clinical features with statistical certainty quantification. Next, ARCH performs a sparse embedding regression to remove indirect linkage between features to build a sparse KG. Finally, ARCH was validated on various clinical tasks, including detecting known relationships between entity pairs, predicting drug side effects, disease phenotyping, as well as sub-typing Alzheimer's disease patients. Results: ARCH produces high-quality clinical embeddings and KG for over 60,000 codified and narrative EHR concepts. The KG and embeddings are visualized in the R-shiny powered web-API. ARCH achieved high accuracy in detecting EHR concept relationships, with AUCs of 0.926 (codified) and 0.861 (NLP) for similar EHR concepts, and 0.810 (codified) and 0.843 (NLP) for related pairs. It detected drug side effects with a 0.723 AUC, which improved to 0.826 after fine-tuning. Using both codified and NLP features, the detection power increased significantly. Compared to other methods, ARCH has superior accuracy and enhances weakly supervised phenotyping algorithms’ performance. Notably, it successfully categorized Alzheimer's patients into two subgroups with varying mortality rates. Conclusion: The proposed ARCH algorithm generates large-scale high-quality semantic representations and knowledge graph for both codified and NLP EHR features, useful for a wide range of predictive modeling tasks.

KW - Electronic health records

KW - Knowledge graph

KW - Natural language processing

KW - Representation learning

UR - http://www.scopus.com/inward/record.url?scp=85216854899&partnerID=8YFLogxK

U2 - 10.1016/j.jbi.2024.104761

DO - 10.1016/j.jbi.2024.104761

M3 - Article

C2 - 39863245

AN - SCOPUS:85216854899

SN - 1532-0464

VL - 162

JO - Journal of Biomedical Informatics

JF - Journal of Biomedical Informatics

M1 - 104761

ER -

ARCH: Large-scale knowledge graph via aggregated narrative codified health records analysis

Abstract

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this