Purpose: This systematic review evaluates the socio-technical integration of autonomous artificial intelligence (AI) in diabetic retinopathy (DR) screening. While the technical accuracy of deep learning algorithms is well-established, their successful deployment depends on a complex interplay of economic, psychological, and organizational factors. This study synthesizes evidence from 35 peer-reviewed sources to provide a comprehensive roadmap for AI implementation in diverse healthcare settings. Methods: A systematic analysis of 35 high-quality studies (2017–2026) was conducted, focusing on diagnostic performance, cost-effectiveness, and stakeholder acceptance. The findings were interpreted through the Consolidated Framework for Implementation Research (CFIR) to identify systemic barriers and facilitators. Results: The evidence confirms that autonomous AI reaches high sensitivity (95.7%–100%) in detecting referable DR, matching specialist performance. Economically, AI is highly cost-effective in resource-limited and rural areas by reducing travel costs and labor burdens. However, significant barriers remain, including the "biological divide" in elderly populations, "black box" anxiety among patients, and organizational disruption in primary care workflows. Trust is identified as a critical mediator, with acceptance increasing when AI is positioned as a supportive "safety net" rather than a human replacement. Conclusion: The transition to AI-driven screening is a transformative shift toward democratized healthcare. Success requires a move toward Human-Centered Design (HCD), localized staff education using tools like instructional videos, and inclusive governance to bridge the digital divide.
Donocik et al. (Thu,) studied this question.