A Multi-Level Framework for AI Adoption in Health Care: Integrating Behavioural, Organisational and Ethical Perspectives

Purpose: Research on the adoption of artificial intelligence in health care has proceeded in four streams that seldom communicate: acceptance theorists study clinicians, organisational theorists study institutions, fairness theorists study algorithmic discrimination, and sustainability researchers study distributional consequences. This review integrates these four streams to build a conceptual framework for empirical work. Design / Methodology: A narrative literature review was conducted using Scopus, Web of Science, PubMed, and Google Scholar (January 2015–March 2026). Around 430 records were filtered to 130 papers through title-abstract-fulltext screening, then thematically reviewed along four theoretical axes: UTAUT2, the Technology-Organisation-Environment (TOE) framework, the AI Bias Taxonomy, and the Triple Bottom Line (TBL). Findings: The review reveals five gaps: fragmentation across individual and organisational levels; weak integration of bias into adoption theory; narrow outcome framings dominated by financial measures; scarce India-specific evidence; and under-use of mediation and moderation methods. The paper develops a multi-level conceptual model with 17 testable propositions linking acceptance, organisational context, bias, trust, and sustainability outcomes. Originality / Value: The framework proposed here is the first to recognise algorithmic bias as a first-class adoption construct—a direct barrier to intention, a constraint on trust, and a moderator of the equity benefits of adoption. It offers a systematic empirical research agenda for scholars and a common language for leaders and policymakers assessing institutional AI decisions.