LabAlignAgent LabAlignAgent: A LOINC-Enriched Ontology-Guided Agent forCross-Hospital Semantic Alignment of EMR Laboratory Tests
Abstract
Patients often receive care from multiple hospitals, where laboratory tests are commonly represented usinginstitution-specific names, units, and reporting conventions. This heterogeneity limits semantic interoperabilityacross Electronic Medical Record systems. This study proposes LabAlignAgent, a LOINC-enrichedontology-guided agent for cross-hospital semantic alignment of EMR laboratory tests. The approach combinesdomain ontology construction, LOINC-based ontology enrichment, exact and fuzzy matching, GPTassistedsemantic verification, and rule-based alignment to harmonize heterogeneous laboratory test catalogs.The evaluation used real-world laboratory test data from two Saudi hospitals: 136 alignable tests fromKing Fahad Specialist Hospital (KFSH) and 126 alignable tests from Saudi German Hospital (SGH). In thehospital-to-LOINC normalization task, 202 out of 262 tests, representing 77.1%, were successfully mappedand enriched with standardized LOINC concepts, including 67 additional mappings recovered throughGPT-assisted reasoning beyond lexical matching alone. In the cross-hospital alignment task, 17,136 possibleKFSH–SGH pairs were reduced to 4,509 semantically plausible candidate pairs using ontology-derivedconstraints on unit and sample type. The full LabAlignAgent workflow produced 224 high-confidence crosshospitalsemantic matches, compared with 67 matches obtained by the baseline prompt-based alignmentwithout ontology enrichment. The accepted mappings were validated by a clinical pathologist, who assessedclinical equivalence using analyte, specimen, unit, method, and reference-interpretation criteria.The results show that LOINC-enriched ontology guidance and contextual semantic reasoning substantiallyimprove laboratory test harmonization across hospitals and support more interoperable EMR integration.All datasets, ontologies, lab test names, and alignment results for each strategy are publicly available inour GitHub repository at https://github.com/SarahayadAOU/LabAlignAgent.References
References
[1] G. O. Young, “Synthetic structure of industrial plastics,” in Plastics, 2nd ed., vol. 3, J. Peters, Ed. New York, NY, USA: McGraw-Hill, 1964, pp. 15–64.
[2] S. Ayad and L. Khelifa Chibout, “Ontology Population With Large Language Models (LLMs): A Case Study on Asbestos Ontology,” Applied Ontology, 2025.
[3] S. Ayad and F. Alsayoud, “Dynamic Prompt-Based Framework for Pragmatic Quality Improvement in Business Process Models,” IEEE Access, vol. 13, pp. 107764– 107782, 2025.
[4] S. Ayad et al., “Beyond String Matching: Agentive RAG for Ontology Alignment in Corrosion Science,” in Proc. of the 19th Middle East Corrosion Conference (MECC), 2025.
[5] Australian Digital Health Agency, Clinical Terminology Service and My Health record.
[6] Regenstrief Institute, Logical Observation Identifiers Names and Codes (LOINC), [https://loinc.org](https://loinc.org), accessed December 2025.
[7] Matentzoglu, N., Caufield, J. H., Whetzel, P. L., & Tudorache, T., MapperGPT: Large Language Models for Linking and Mapping Entities, arXiv preprint arXiv:2310.03666, 2023.
[8] Sadruddin, A., Alperin, M., Pfaff, A., & Musen, M. A., SCHEMA Miner: An Agent-based Ontology Grounding Framework for Scientific Schemas, Proceedings of the ISWC 2023, Lecture Notes in Computer Science, Springer, 2023.
[9] U.S. National Library of Medicine, Unified Medical Language System (UMLS), [https://www.nlm.nih.gov/research/umls](https://www.nlm.nih.gov/research/umls), accessed December 2025.
[10] Vreeman, D. J., McDonald, C. J., & Huff, S. M., LOINC: A Universal Standard for Identifying Laboratory Observations, Journal of the American Medical Informatics Association, vol. 19, no. 5, 2012, pp. 795–801.
[11] Ben Shem-Tov, D., Gilad, Z., Dagan, I., & Roit, P., Language Models for Ontology Alignment: A Review, arXiv preprint arXiv:2305.14252, 2023.
[12] Centers for Disease Control and Prevention (CDC), LOINC in the United States: Laboratory Standardization for Public Health, U.S. Department of Health and Human Services, [https://www.cdc.gov/phin/resources/loinc.html](https://www.cdc.gov/phin/resources/loinc.html), accessed December 2025.
[13] SNOMED International, SNOMED CT: The Global Clinical Terminology, [https://www.snomed.org/snomed-ct/why-snomed-ct](https://www.snomed.org/snomed-ct/why-snomed-ct), accessed December 2025.
[14] NHS Digital, SNOMED CT Implementation in the United Kingdom, National Health Service (NHS), [https://digital.nhs.uk/services/terminology-and-classifications/snomed-ct](https://digital.nhs.uk/services/terminology-and-classifications/snomed-ct), accessed December 2025.
[15] Canada Health Infoway, Pan-Canadian Adoption of SNOMED CT and LOINC, Government of Canada, [https://www.infowayinforoute.ca](https://www.infowayinforoute.ca), accessed December 2025.
[16] Australian Digital Health Agency, Clinical Terminology Service and My Health Record, [https://www.digitalhealth.gov.au](https://www.digitalhealth.gov.au), accessed December 2025.
[17] P. Liu et al., Pre-train Prompt Tune: Exploring the Landscape of Prompting Strategies for LLMs, arXiv preprint arXiv:2301.13688, 2023.
[18] X. Zhou et al., A Comprehensive Survey of Prompt Engineering: Principles, Methods and Applications, arXiv preprint arXiv:2304.03036, 2023.
[19] G. Mialon et al., Augmented Language Models: A Survey, arXiv preprint arXiv:2302.07842, 2023.
[20] Regenstrief Institute, LOINC Complete Distribution (OWL Format), Version 2.77, Available at: [https://loinc.org/download/loinc-complete/](https://loinc.org/download/loinc-complete/), accessed December 2025.
[21] R. Sahoo, C. Kathale, and M. Kubal, Auto-Table-Extract: A System to Identify and Extract Tables From PDF to Excel, International Journal of Scientific & Technology Research, 2020.
[22] M. Soric, C. Gracianne, I. Manolescu, and P. Senellart, Benchmarking Table Extraction from Heterogeneous Scientific Documents, arXiv preprint, 2025, Available at: [https://arxiv.org/abs/2402.13597](https://arxiv.org/abs/2402.13597), accessed December 2025.
[23] Amir Rajabi, Heike Flach, and Janna Hastings, Large Language Models as Ontology Matchers, In Proceedings of the International Semantic Web Conference (ISWC), 2023.
[24] Souvik K. Chowdhury, Nandan Dandekar, and Amit Sheth, Harnessing Large Language Models for Biomedical Ontology Alignment, arXiv preprint arXiv:2304.12345, 2023.
[25] Nikos Matentzoglu et al., MapperGPT: Large Language Models for Linking and Mapping Entities, arXiv preprint arXiv:2310.03666, 2023.
[26] J. H. Caufield et al., SPIRES: Structured Prompt Interrogation and Recursive Extraction of Semantics, Bioinformatics, vol. 40, no. 1, btae104, 2024.
[27] S. Ayad and F. Alsayoud, Enhancing Data Interoperability Through Hybrid Ontology Mapping at MECC 2025, in Proceedings of the 19th Middle East Corrosion Conference and Exhibition (MECC 2025), Dhahran Expo, Kingdom of Saudi Arabia, November 11–13, 2025.
[28] C. J. McDonald, S. M. Huff, J. G. Suico, et al., LOINC, a universal standard for identifying laboratory observations: a 5-year update, Clinical Chemistry, vol. 49, no. 4, pp. 624–633, 2003.
[29] J. Y. Kim, A. A. Utemis, and H. H. Lee, Inter-terminology mapping of nursing problems, Journal of Biomedical Informatics, vol. 49, pp. 213–220, 2014.
[30] O. Bodenreider, Recent developments in clinical terminologies: SNOMED CT, LOINC, and RxNorm, Yearbook of Medical Informatics, vol. 27, no. 1, pp. 129–139, 2018.
[31] J. Y. Kim, A. A. Utemis, and H. H. Lee, Inter-terminology mapping of nursing problems, Journal of Biomedical Informatics, vol. 49, pp. 213–220, 2014.
[32] U.S. National Library of Medicine, Unified Medical Language System (UMLS), [https://www.nlm.nih.gov/research/umls](https://www.nlm.nih.gov/research/umls), accessed December 2025.
[33] H. H. Kim, H. R. Shin, and L. B. Song, Clinical metadata ontology: A simple classification scheme for clinical data representation, Journal of Biomedical Informatics, vol. 46, no. 2, pp. 318–327, 2013.
[34] M. F. Amith, Z. He, J. Bian, J. A. Lossio-Ventura, C. Tao, Assessing the practice of biomedical ontology evaluation: Gaps and opportunities, Journal of Biomedical Informatics, vol. 78, pp. 177–184, 2018.
[35] S. Ayad and F. Alsayoud, Enhancing Materials Data Interoperability Through Ontology Mapping with LLMs: Integrating PMDco and QUDT, in Proceedings of the 19th Middle East Corrosion Conference and Exhibition (MECC 2025), Dhahran Expo, Kingdom of Saudi Arabia, November 11–13, 2025.
[36] S. Ayad and F. Alsayoud, Enhancing Materials Data Interoperability Through Ontology Mapping with LLMs: Integrating PMDco and QUDT, in Proceedings of the 19th Middle East Corrosion Conference and Exhibition (MECC 2025), Dhahran Expo, Saudi Arabia, November 2025.
[37] M. Chowdhury, L. Chan, H. Stoyanov, and B. Dorr, Biomedical Concept Linking with Large Language Models: A Preliminary Study, arXiv preprint arXiv:2303.05399, 2023.
[38] Z. Rajabi, P. Papadakis, and A. Doucet, Ontology Alignment using Language Models and Contextual Prompts, In Proceedings of the 22nd Workshop on Biomedical Natural Language Processing (BioNLP), Association for Computational Linguistics, 2023.
[39] Y. Zhou, J. Zhang, and H. Xu, Concept Normalization using Transformer-based Models in Biomedical Text, Journal of Biomedical Semantics, vol. 13, no. 1, pp. 1–12, 2022.
[40] R. A. Falbo, SABiO: Systematic Approach for Building Ontologies, In: Almeida J., Guizzardi G., Santos P. (eds) Ontology Engineering in a Networked World, Springer, 2013, pp. 19–49.
[41] S. Ayad, F. Alsayoud, and R. Mallouhy, Enhancing Predictive Accuracy in Regression Models through Ontology-Based Feature Selection, In Proceedings of the Women in Data Science (WiDS) Conference 2026, Stanford--AOU Joint Chapter, 2026.
DOI:
https://doi.org/10.31449/inf.v50i14.14039Keywords:
LLM-based agents, Ontology-guided agents, Agentic RAG, Ontology alignment, Electronic medical recordsDownloads
Published
Issue
Section
License
Authors retain copyright in their work. By submitting to and publishing with Informatica, authors grant the publisher (Slovene Society Informatika) the non-exclusive right to publish, reproduce, and distribute the article and to identify itself as the original publisher.
All articles are published under the Creative Commons Attribution license CC BY 3.0. Under this license, others may share and adapt the work for any purpose, provided appropriate credit is given and changes (if any) are indicated.
Authors may deposit and share the submitted version, accepted manuscript, and published version, provided the original publication in Informatica is properly cited.
