Artificial Intelligence in Clinical Care: Perceptions of Retina Specialists and Patients Gathered Through a Multicenter Survey

Abstract

Purpose: To characterize retina specialist and patient attitudes toward artificial intelligence (AI) in retina clinical care. Methods: A parallel survey was concurrently administered to retina specialists electronically and to patients in 5 retina practices. Responses were based on a 5-point Likert scale. Data were analyzed using χ² and Mann-Whitney U tests. Results: A total of 291 (97%) of 300 patients approached participated, while 78 (17%) of 447 physicians responded electronically. Major differences (mean differences > 0.5) were seen in opinions on patients choosing whether AI is used in their care (Likert scale mean, 3.3 for physicians vs 4.2 for patients; P < .0001), desire for AI use in clinical care (3.7 for physicians vs 2.9 for patients; P < .0001), and physicians being responsible for harm caused by AI (3.4 for physicians vs 4.1 for patients; P < .0001). Both groups aligned on discomfort with AI taking clinical roles such as deciding on treatments (2.6 for both; P = .9) and answering patients’ questions about diagnosis and treatment (2.7 for both; P = .9). Conclusions: The present study demonstrated important differences in retina specialist and patient perceptions on the use of AI in clinical care, particularly pertaining to patient autonomy. Both groups expressed discomfort with AI taking direct clinical roles. As AI integration in clinical care progresses, physicians and developers should continue to understand and address patient concerns.