Abstract
AIMS: To explore attitudes of people living with diabetes (PLD) and healthcare professionals (HCP) towards the use of automated retinal image analysis systems using artificial intelligence (AI) in NHS Diabetic Eye Screening Programmes (DESP) and how these perceptions vary by sociodemographic subgroups. METHODS: Two anonymous online surveys (28 questions for PLD and 21 for HCP) were developed to assess attitudes towards AI. Data were collected from four English DESPs, diabetes charities and patient groups between September and December 2023. Likert-scale responses were analysed using regression to examine subgroup differences. RESULTS: A total of 1577 PLD and 262 HCP participated. Fifty-eight per cent of PLD believed AI would perform equally well in all subgroups, compared with 32% of HCP. Seventy-one per cent of HCP disagreed that AI could replace human grading, and 81% of PLD felt humans should remain responsible for screening outcomes. Both groups supported AI's efficiency but had concerns about data security, trust, job security and who would be responsible for AI errors. Linear regression of Likert scores showed women were less accepting of AI; PLD of Black and Asian ethnicities were more cautious of data security and impact on screening experience. HCP of Asian ethnicity generally held more negative views across themes. Those using more online applications had more positive views towards AI. CONCLUSIONS: While both PLD and HCP recognise AI's potential benefits, concerns regarding security, job impact and errors highlight the need for targeted outreach based on sociodemographic factors.