Stakeholder Perspectives of Implementation Barriers of Artificial Intelligence in Eye Care: A qualitative framework-based study.

Abstract

Despite the revolution of artificial intelligence (AI), its integration remains limited in healthcare. A comprehensive understanding of the barriers to implementation is crucial to enhance the utilisation of AI. This study applies a conceptual framework-based analysis, to explore stakeholder perspectives of implementation barriers of AI in digital diagnosis in eye care. Purposive sampling was used to identify key individuals across stakeholder groups, including technology developers, clinicians, patients and healthcare leaders. Semi-structured interviews were conducted with 37 stakeholders. Using the updated Consolidated Framework for Implementation Research (CFIR), responses to the question: 'What is the biggest barrier to digital diagnosis or AI for macular disease in Australia?' were analysed. Barriers identified by stakeholders were mapped to thematic constructs of the updated CFIR, and the prominence of each implementation barrier was measured. Data saturation was not assessed. For clinicians and developers, the 'innovation' domain was most frequently cited. Clinicians were most concerned with the costs involved, whereas for developers, a lack of evidence surrounding real-world application was the main challenge. For leaders and patients, 'individuals' domain was the most frequently cited. Leaders were focused on the innovation deliverers: expressing the potential risk of over-reliance on the innovation and the subsequent consequence of clinician deskilling. Patients were more concerned about innovation recipients: emphasising the perceived lack of human empathy with the implementation of AI. Differences were revealed in the identified barriers to the implementation of AI across stakeholder groups. A co-design approach to address the misalignment in key barriers may be essential to the successful implementation of AI in digital health innovations.