Using artificial intelligence in diabetic eye screening programmes: perspectives from people with diabetes and screening staff in Northern Ireland

Background: Artificial Intelligence (AI) is increasingly used in everyday life including within hospitals and medical appointments. Scotland have been using AI in their diabetic eye screening programme

Methods: A questionnaire was distributed to all NIDESP staff and PwDM (through patient events). The questionnaire included questions on knowledge of AI, expectations of using AI in DESP and main concerns. Likert Scale questions were scored from 1- 10, 10 being the highest. Likert Scale and thematic analysis was undertaken.

Results: In total, 13 NIDESP staff and 13 PwDM responded. NIDESP staff felt implementation of AI would be helpful averaging 7 on the Likert scale. Despite this, NIDESP staff were moderately concerned about AI implementation (average 6). Nearly all (92%), expected AI to help with grading including removing R0M0 grades, allowing more focus on other gradings and taking pressure off grading queues and staff. Staff felt AI would be most useful in primary grading, removing R0M0s and identifying potentially urgent cases in turn reducing staff pressure, prioritising urgent patient care and standardising grading further. Despite this, 46% stated patient safety concerns, 31% process efficiency concerns and several concerned with losses of personnel.

Eighty-five percent of PwDM knew AI could be used in healthcare. They felt AI could speed up results, take pressure off staff, streamline appointments, and identify eye complications. PwDM also felt it could speed up waiting times and prioritise those in need. Nearly all, 92% were happy for AI to be implemented into DESP but 1 patient stated they preferred a ‘human touch’. Four patients expressed concerns surrounding safeguarding their private information and overreliance on AI.

Conclusions: PwDM were very positive about implementing AI in DESP, many stating it should be embraced. Staff were overall happy with AI for grading prioritisation and efficiency but had several concerns around patient safety and loss of personnel. It is important when implementing AI into the NIDESP that all opinions are considered, and all stakeholders remain informed.