Abstract
INTRODUCTION: Artificial Intelligence (AI) refers to a system that can take input data and subsequently provide a prediction-based output. In some instances, these systems can provide a recommendation for a clinical decision. Such systems are being developed to analyse retinal images and to determine if there is active retinal disease with a view to directly influencing treatment decisions. This study looks at whether AI-based decision making is acceptable to patients with macular disease. METHODS: The Macular Society has a monthly newsletter which it sends out to its members and subscribers in which views were canvassed. They were offered participation in a conjoint analysis exploring human or AI decision making, the error rate, the time to follow up and whether the scans were double read by a human or AI tool. Options were presented in a random order and participants were asked to rank the suitability of different scenarios in order of preference. RESULTS: The task was completed by 181 participants. The two most important factors were the error rate (p < 0.0001) and whether the results were being checked (p < 0.0001). Participants did not state a preference for the first and/or second reader being either human or AI and there was a non-significant trend for rapid turnaround. CONCLUSIONS: Patients with macular disease find AI to be acceptable in the assessment of retinal images. The most important factors to patients relate to the accuracy of the decision making.