Clinicians' perspectives on machine learning for obstructive sleep apnoea detection: a human factors study.

Obstructive sleep apnoea is a common sleep disorder affecting 5%-15% of the population, with many going undiagnosed due to accessibility issues and long waiting lists. This is due to the bottleneck method of diagnosis, which is polysomnography. The implementation of machine learning, particularly when applied to a reduced set of physiological signals, has the potential to enhance accessibility and to mitigate the demands associated with conventional polysomnography, a procedure that is inherently time-consuming and labour-intensive. In this preliminary study, 10 clinicians were interviewed to gather their perspectives on the use of machine learning in the domain. Using thematic analysis, key themes were identified, highlighting important considerations on the deployment of machine learning. Findings suggest the need for guidelines and approvals to regulate the use of machine learning, the importance of including medical experts in the process to ensure the best healthcare service is provided and that clinicians lead and not machine learning. The aim is to further the understanding of the complexities of machine learning and understand how to better tailor machine learning to allow for the adoption of machine learning and increase the trust in machine learning by clinicians.