Using the Normalization Process Theory to study facilitators and barriers to sustainable implementation of artificial...

Area of Science:

• Ophthalmology
• Public Health
• Health Services Research

Background:

• Rising diabetes and diabetic retinopathy (DR) incidence in sub-Saharan Africa strains healthcare systems.
• Shortages of skilled professionals exacerbate the challenge of managing DR.
• Artificial intelligence (AI)-based screening offers a potential solution to improve DR detection and management.

Purpose of the Study:

• To evaluate the integration and sustainability of AI-based diabetic retinopathy screening in Rwanda.
• To understand user perceptions and identify barriers and facilitators to embedding AI screening into routine clinical practice.

Main Methods:

• Qualitative study using the Normalization Process Theory (NPT) framework.
• Semi-structured interviews and focus groups with 9 clinical staff and 67 patients in Kigali, Rwanda.
• Thematic analysis of interview data aligned with NPT constructs (coherence, cognitive participation, collective action, reflexive monitoring).

Main Results:

• Participants understood the AI screening program's value in addressing healthcare provider scarcity and DR burden.
• Key challenges included workload, labor division, initial patient distrust of AI, and organizational policies.
• The NPT framework effectively analyzed implementation, yielding actionable recommendations.

Conclusions:

• AI-based DR screening demonstrates potential for sustainable integration, supported by positive user perceptions.
• Addressing identified challenges is crucial for successful long-term implementation.
• Further research is needed to assess long-term health outcomes, including visual preservation and treatment success.