An Agentic AI Framework for Master Data Synchronization, Validation and Self-Healing Across Integrated Healthcare Information Systems

Abstract

Background. Large healthcare organizations run many specialty applications, such as prescribing and dispensing, warehouse and inventory, laboratory, blood bank and billing, that depend on shared master data, chiefly the patient identifier (MRN) and the product identifier with its description. Standards such as HL7 v2 and FHIR can propagate a new or changed record from an origin system, but propagation failures (rejections, unmapped codes, missing mandatory fields, duplicates and timing dependencies) still rely on manual diagnosis and repair. The resulting gaps surface as held or failed transactions at the point of care and supply. Failures also arise without any interface error, when a record exists but its local build or configuration (for example location, charge or dispensing parameters) is incomplete. Objective. To specify a human-governed, closed-loop framework in which AI agents act as integration architect, builder, verifier and support engineer: designing field mappings, configuring receiving systems against an approved build standard, verifying cross-system data validity and local build completeness, and diagnosing and remediating propagation errors. Methods. The study followed a design-science approach. The synchronization gap is formalized, a seven-class failure taxonomy is derived, and a four-layer architecture (sense, reason, act, govern) is specified with tiered autonomy grounded in the Parasuraman-Sheridan-Wickens model of types and levels of automation. An explicit tier-decision rule and tool contracts are defined, verification checks are mapped to established electronic health record data quality frameworks, the closed loop is illustrated with a synthetic worked example, and a staged evaluation protocol with quantified safety gates is specified. Results. The framework specifies agent roles and tool contracts; a proactive build-completeness check for local configuration; a default and maximum autonomy tier for each of seven failure classes; a decision rule in which the permitted tier is the minimum of a policy tier, a confidence tier and a criticality cap; safeguards (patient identity actions reserved for humans, grounding in versioned build standards, deterministic pre-execution validation, and treatment of message content as untrusted data) mapped to a threat model; and evaluation metrics (time to consistency, held-transaction rate, false-fix rate, harmful-action rate) with the sample sizes needed to demonstrate safety bounds, for example 149 consecutive correct fixes to show with 95% confidence that the false-fix rate is below 2%. This is a conceptual and design paper; no empirical results are reported. The tier-decision rule was checked exhaustively against ten stated properties over 11,440 input combinations, with no violations, and six deliberately seeded faults were each detected (Section IV-I). Conclusions. The framework proposes that an AI agent layer can be added over existing integrated systems without replacing them, turning error queues from passive backlogs into a supervised remediation loop. Whether it achieves this is untested. The paper supplies testable hypotheses, pre-specified safety gates and a staged path from shadow mode to limited autonomy for future empirical validation.

Introduction

The text proposes an AI-driven framework for improving master data synchronization across integrated hospital information systems. Modern hospitals use multiple specialized applications for pharmacy, laboratory services, blood banks, inventory management, and billing. These systems depend on consistent shared information, particularly patient identifiers and product records. Although interoperability standards such as HL7 v2 and HL7 FHIR support data exchange, differences in local configurations, code tables, mandatory fields, and validation rules can prevent records from becoming usable across all systems.

The main problem is the synchronization gap, which occurs when a patient or product record is missing, invalid, incomplete, or inconsistent in one or more receiving systems. Such gaps can delay medication dispensing, laboratory testing, inventory transactions, and billing. Manual correction by IT staff is repetitive and time-consuming, while some configuration problems remain undetected until a transaction fails. The paper proposes using AI agents to identify, diagnose, correct, and verify these problems automatically within clearly defined safety limits.

The proposed framework introduces a seven-class failure taxonomy covering reference or code mismatches, missing mandatory fields, dependency and ordering problems, transport failures, identity conflicts, semantic inconsistencies, and incomplete local configurations. It also defines two important measures: the synchronization gap, which identifies systems where a record is not yet usable, and time to consistency (TTC), which measures how long it takes for all receiving systems to become consistent with the originating record.

The framework consists of four main architectural layers:

• Sense layer: Collects integration acknowledgments, message queues, system logs, origin-system updates, and configuration information.

• Reason layer: Uses AI agents to analyze failures and retrieve relevant specifications, approved build standards, and historical incident information.

• Act layer: Provides controlled tools for proposing configuration changes, testing fixes, applying authorized changes, and replaying failed messages.

• Governance layer: Enforces approval requirements, autonomy limits, audit logging, monitoring, and emergency-stop mechanisms.

Four specialized agents support this architecture. The Architect agent proposes field mappings and interface specifications; the Builder agent prepares and applies authorized configuration changes; the Verifier agent checks consistency and configuration completeness across systems; and the Support agent classifies failures, identifies their causes, coordinates remediation, and verifies resolution.

A key feature is the tiered-autonomy policy, which determines how much freedom an agent receives based on failure type, diagnostic confidence, and the criticality of the affected data. Low-risk, reversible technical corrections may be automated within approved limits, whereas sensitive changes involving patient identity or clinically significant product attributes require human authorization. Every action must be traceable, verifiable, and reversible.

The literature review identifies a research gap between established interoperability standards, data-quality frameworks, and emerging AI-agent technologies. Existing approaches help transport data and detect errors, but the paper argues that a unified, safety-governed framework is needed to connect failure detection with verified correction across healthcare systems.

Conclusion

Standards such as HL7 and FHIR move master data between systems, but the last step of making that data valid and usable in each receiving system often depends on manual diagnosis and repair. We have specified a governed, closed-loop framework in which AI agents act as integration architect, builder, verifier and support engineer, with autonomy tiered by failure class, verification built into every fix, and patient identity left to humans. The framework is designed to be overlaid on existing systems and to be evaluated in stages. The framework also checks for incomplete local build or configuration, which raises no interface error, so that a gap can be flagged before a patient is waiting. A worked example and a threat model show how the pieces fit together, and quantified safety gates define what evidence would justify moving from shadow mode to supervised and then limited autonomy. The principal limitation is the absence of empirical validation: the claims of this paper are design claims and hypotheses, and the staged protocol is intended to test them. Future work includes: a pilot following the protocol in Section VI, beginning with calibration of the confidence thresholds and a comparison against a deterministic rule-based baseline; a shared, de-identified benchmark of interface-error cases for comparing agents; implementation-level and formal verification of the policy engine, beyond the specification-level check reported in Section IV-I; study of multi-agent coordination and failure modes; extension to additional master entities and to FHIR-native validation and subscription mechanisms; and a systematic literature review of agentic AI in healthcare operations and interoperability.

References

[1] Bender D, Sartipi K. HL7 FHIR: an agile and RESTful approach to healthcare information exchange. In: Proceedings of the 26th IEEE International Symposium on Computer-Based Medical Systems (CBMS). IEEE; 2013. p. 326-31. doi:10.1109/CBMS.2013.6627810 [2] Mandel JC, Kreda DA, Mandl KD, Kohane IS, Ramoni RB. SMART on FHIR: a standards-based, interoperable apps platform for electronic health records. J Am Med Inform Assoc. 2016;23(5):899-908. doi:10.1093/jamia/ocv189 [3] HL7 International. FHIR Release 4 (v4.0.1). Available from: https://hl7.org/fhir/R4/ [accessed 22 October 2026]. [4] Hohpe G, Woolf B. Enterprise Integration Patterns: Designing, Building, and Deploying Messaging Solutions. Boston: Addison-Wesley; 2003. [5] Wang RY, Strong DM. Beyond accuracy: what data quality means to data consumers. J Manag Inf Syst. 1996;12(4):5-33. [6] Weiskopf NG, Weng C. Methods and dimensions of electronic health record data quality assessment: enabling reuse for clinical research. J Am Med Inform Assoc. 2013;20(1):144-51. [7] Kahn MG, Callahan TJ, Barnard J, Bauck AE, Brown J, Davidson BN, et al. A harmonized data quality assessment terminology and framework for the secondary use of electronic health record data. eGEMs. 2016;4(1):1244. doi:10.13063/2327-9214.1244 [8] Fellegi IP, Sunter AB. A theory for record linkage. J Am Stat Assoc. 1969;64(328):1183-210. [9] Reason J. Human error: models and management. BMJ. 2000;320(7237):768-70. [10] Yao S, Zhao J, Yu D, Du N, Shafran I, Narasimhan K, Cao Y. ReAct: synergizing reasoning and acting in language models. In: International Conference on Learning Representations (ICLR); 2023. arXiv:2210.03629. [11] Wang L, Ma C, Feng X, Zhang Z, Yang H, Zhang J, et al. A survey on large language model based autonomous agents. Front Comput Sci. 2024;18(6):186345. doi:10.1007/s11704-024-40231-1 [12] Ji Z, Lee N, Frieske R, Yu T, Su D, Xu Y, et al. Survey of hallucination in natural language generation. ACM Comput Surv. 2023;55(12):Article 248. [13] Singhal K, Azizi S, Tu T, Mahdavi SS, Wei J, Chung HW, et al. Large language models encode clinical knowledge. Nature. 2023;620(7972):172-80. [14] Rajkomar A, Dean J, Kohane I. Machine learning in medicine. N Engl J Med. 2019;380(14):1347-58. [15] Parasuraman R, Sheridan TB, Wickens CD. A model for types and levels of human interaction with automation. IEEE Trans Syst Man Cybern A Syst Hum. 2000;30(3):286-97. doi:10.1109/3468.844354 [16] Onnasch L, Wickens CD, Li H, Manzey D. Human performance consequences of stages and levels of automation: an integrated meta-analysis. Hum Factors. 2014;56(3):476-88. [17] Sittig DF, Singh H. A new sociotechnical model for studying health information technology in complex adaptive healthcare systems. Qual Saf Health Care. 2010;19(Suppl 3):i68-74. [18] Hevner AR, March ST, Park J, Ram S. Design science in information systems research. MIS Q. 2004;28(1):75-105. [19] Peffers K, Tuunanen T, Rothenberger MA, Chatterjee S. A design science research methodology for information systems research. J Manag Inf Syst. 2007;24(3):45-77. [20] National Institute of Standards and Technology. Artificial Intelligence Risk Management Framework (AI RMF 1.0). NIST AI 100-1. Gaithersburg (MD): NIST; 2023. [21] Shingo S. Zero Quality Control: Source Inspection and the Poka-Yoke System. Cambridge (MA): Productivity Press; 1986. [22] Samiyullaha S. A configurable, mistake-proofed function for adding and removing medications across hospital pharmacy inventory locations. Int J Novel Res Dev. 2026;11(9):b392-b400. [23] HL7 International. HL7 Version 2.5.1 Messaging Standard (ANSI/HL7 V2.5.1-2007). Ann Arbor (MI): Health Level Seven International; 2007. [24] Ayaz M, Pasha MF, Alzahrani MY, Budiarto R, Stiawan D. The Fast Health Interoperability Resources (FHIR) standard: systematic literature review of implementations, applications, challenges and opportunities. JMIR Med Inform. 2021;9(7):e21929. doi:10.2196/21929 [25] Christen P. Data Matching: Concepts and Techniques for Record Linkage, Entity Resolution, and Duplicate Detection. Berlin: Springer; 2012. [26] Narayan A, Chami I, Orr L, Arora S, Ré C. Can foundation models wrangle your data? Proc VLDB Endow. 2022;16(4):738-46. arXiv:2205.09911 [27] Notaro P, Cardoso J, Gerndt M. A survey of AIOps methods for failure management. ACM Trans Intell Syst Technol. 2021;12(6):Article 81. doi:10.1145/3483424 [28] Bedi S, Liu Y, Orr-Ewing L, Dash D, Koyejo S, Callahan A, et al. Testing and evaluation of health care applications of large language models: a systematic review. JAMA. 2025;333(4):319-28. doi:10.1001/jama.2024.21700 [29] Greshake K, Abdelnabi S, Mishra S, Endres C, Holz T, Fritz M. Not what you\'ve signed up for: compromising real-world LLM-integrated applications with indirect prompt injection. In: Proceedings of the 16th ACM Workshop on Artificial Intelligence and Security (AISec \'23). ACM; 2023. p. 79-90. arXiv:2302.12173 [30] OWASP Foundation. OWASP Top 10 for Large Language Model Applications (2025). Available from: https://genai.owasp.org/llm-top-10/ [31] National Institute of Standards and Technology. Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile. NIST AI 600-1. Gaithersburg (MD): NIST; 2024. doi:10.6028/NIST.AI.600-1 [32] Venable J, Pries-Heje J, Baskerville R. FEDS: a framework for evaluation in design science research. Eur J Inf Syst. 2016;25(1):77-89. doi:10.1057/ejis.2014.36 [33] Vasey B, Nagendran M, Campbell B, Clifton DA, Collins GS, Denaxas S, et al. Reporting guideline for the early-stage clinical evaluation of decision support systems driven by artificial intelligence: DECIDE-AI. Nat Med. 2022;28(5):924-33. doi:10.1038/s41591-022-01772-9 [34] Samiyullaha S. Persistent stock discrepancies in hospital pharmacy supply chains: why robotics, system integration, and artificial intelligence have not closed the reconciliation gap. Int J Sci Res Technol. 2026;3(10):80-7. doi:10.5281/zenodo.23116496 [35] Fong A, Kazi S, Kazanas M, Ratwani RM, Hettinger AZ. Estimating the frequency of automated dispensing cabinet discrepancy safety events using Markov models. Drugs Ther Perspect. 2022;38(3):146-55. doi:10.1007/s40267-022-00900-2 [36] Dhar M, Singh R, Manikonda SCK. From single chatbots to governed agent ecosystems: an agentic AI pattern catalogue and orchestration framework for mission-critical hospital information management systems. Int J Innov Sci Res Technol. 2026;11(7):1851-84. doi:10.38124/ijisrt/26jul830

Copyright

Copyright © 2026 Samiyullaha Sayyed. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.