Economic, ethical, and regulatory dimensions of artificial intelligence in healthcare: an integrative review

Abstract

Background: Artificial Intelligence (AI) is revolutionizing healthcare by improving diagnostic precision, streamlining clinical workflows, and reducing operational costs. Yet, its integration into real-world settings remains fraught with challenges-including economic uncertainty, ethical complexities, fragmented regulatory landscapes, and practical implementation barriers. A growing body of literature highlights that many of AI's purported benefits are derived from idealized models, often failing to reflect the nuances of clinical practice. Objectives: This integrative review aims to critically evaluate the current evidence on the integration of artificial intelligence into healthcare, with a particular focus on its economic impact, ethical and regulatory challenges, and associated governance and implementation strategies. Methods: A comprehensive literature search was conducted across PubMed/MEDLINE, Embase, Web of Science, and the Cochrane Library. Data extraction followed a structured, pre-tested template, and thematic synthesis was employed. Study quality was assessed using an integrated framework combining PRISMA, AMSTAR 2, and the Drummond checklist. Results: Seventeen studies-including systematic reviews, scoping reviews, narrative syntheses, policy analyses, and quantitative case studies-met the inclusion criteria. Three core themes emerged from the analysis. First, while AI interventions-particularly in treatment optimization-are projected to generate significant cost savings and improve operational efficiency, most economic evaluations rely on theoretical models. Many lack transparency regarding key assumptions such as discount rates, sensitivity analyses, and real-world implementation costs, limiting their generalizability. Second, ethical and regulatory concerns persist, with widespread underrepresentation of marginalized populations in training datasets, limited safeguards for patient autonomy, and notable equity disparities across clinical domains. Regulatory frameworks remain fragmented globally, with marked variation in standards for cybersecurity, accountability, and innovation readiness. Third, effective governance and risk management are critical for ensuring safe and sustainable AI integration. Persistent implementation barriers-such as clinician trust deficits, cognitive overload, and data interoperability challenges-underscore the need for robust multidisciplinary collaboration. Recommendations: TF Framework-a theoretical model pending empirical validation. It is built on five pillars: co-design and problem definition, data standardization, real-world performance monitoring, ethical and regulatory integration, and multidisciplinary governance. This framework offers an actionable roadmap for fostering equitable, trustworthy, and scalable AI deployment across healthcare systems. Conclusion: TF Framework provides a foundation for ethically grounded, patient-centered, and financially sustainable AI integration.

Ähnliche Arbeiten