EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC 4.0

The ethics-driven data governance framework for health-based AI (EDG-AI4H): an integrated approach for empowering explainable AI in health insurance risk prediction in Iran

Rezvan Rahimi

Dialogues in Health · 2026-12 · 9 · 100351

DOI: 10.1016/j.dialog.2026.100351 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Dialogues in Health, 2026-12. Verfasst von Rezvan Rahimi. Beteiligte Einrichtung: Department of Medical Informatics, Faculty of Medical Sciences, Tarbiat Modares University, Tehran, Iran. Die Arbeit nennt 32 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0).

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Purpose

The digital transformation of health data has enabled predictive models in insurance. In Iran, the absence of an indigenous data governance framework integrating ethical, legal, and technical considerations poses serious risks to responsible AI development. This study aimed to design an ethics-driven data governance framework to empower Explainable AI (XAI) in health insurance risk prediction in Iran.

Methods

This qualitative study (2024) used thematic analysis and framework development. Data came from: (1) a narrative review of international articles and Iranian policy documents; (2) semi-structured interviews with five experts in insurance IT, health data science, IT law, epidemiology, and insurance management. Data were analyzed using Braun and Clarke's six-step thematic analysis in MAXQDA 2022. The framework was validated through a focus group with the same five experts to ensure continuity and in-depth engagement with the framework design. The study adheres to COREQ guidelines, and a completed COREQ checklist is provided as Supplementary File 1.

Results

The EDG-AI4H framework comprises four hierarchical and interactive layers: (1) Governance-Ethics (data ownership policies, dynamic consent management, Data Ethics Board, legal compliance); (2) Operational-Integration (metadata-driven catalog, standardized data exchange, data quality assurance, secure storage architecture); (3) Intelligence-Analytics (domain-driven feature engineering, mandatory XAI-by-design, active bias monitoring, model lifecycle governance); and (4) Value-Decision Making (multi-level decision support, preventive and personalized insurance, population health management, policyholder transparency and empowerment).

Conclusion

The EDG-AI4H framework addresses a critical gap in Iran's health insurance sector by linking ethics, technology, and business. It offers a roadmap for responsible, transparent, and trustworthy AI in health risk prediction, potentially reducing ethical and legal risks while fostering public trust. The framework's operational feasibility requires empirical validation through future pilot implementation studies.

Schlagwörter

Stichwörter der Autor:innen

Conceptual frameworkEthics-driven data governanceExplainable Artificial Intelligence (XAI)Health insuranceHealth risk predictionQualitative research

Autor:innen und Institutionen

  • Department of Medical Informatics, Faculty of Medical Sciences, Tarbiat Modares University, Tehran, Iran.

Zitieren

APA

Rahimi, R. (2026). The ethics-driven data governance framework for health-based AI (EDG-AI4H): an integrated approach for empowering explainable AI in health insurance risk prediction in Iran. Dialogues in Health, 9, 100351. https://doi.org/10.1016/j.dialog.2026.100351

Vancouver

Rahimi R. The ethics-driven data governance framework for health-based AI (EDG-AI4H): an integrated approach for empowering explainable AI in health insurance risk prediction in Iran. Dialogues in Health. 2026;9:100351. doi:10.1016/j.dialog.2026.100351

BibTeX
@article{Rahimi2026,
  author = {Rahimi, Rezvan},
  title = {The ethics-driven data governance framework for health-based AI (EDG-AI4H): an integrated approach for empowering explainable AI in health insurance risk prediction in Iran},
  journal = {Dialogues in Health},
  year = {2026},
  volume = {9},
  pages = {100351},
  doi = {10.1016/j.dialog.2026.100351}
}
Kostenlos starten →