Organisational readiness for the adoption of AI diagnostic tools in clinical workflows – a case study from the National Breast Screening Programme in the East Midlands of England. (Preprint)

BACKGROUND Application of artificial intelligence (AI) in healthcare is accelerating but relatively little is yet known about the real-world implementation of AI in clinical workflows. OBJECTIVE In this paper, we have focused on one application of AI as a second reader of breast mammograms in the context of a national breast screening programme. We look at the development and testing of an AI image reading tool for mammograms and the effect of organisational readiness for AI tool adoption. We focus on two aspects of organisational readiness as conceptualised by Weiner (2009) for AI technology specifically and answer the questions (1) what are the views of the technology adopters in a healthcare organisation to the use of AI technology in the case of breast screening? (2) What are some of the emerging organisation factors that are likely to effect adoption and spread and are any unique to AI technology? METHODS A prospective mixed methods study of the real-world development of AI tools for use in the National Breast Screening Programme in England. We recruited 67 radiologists and reporting radiographers in four breast screening services and 18 organisational leaders who were the AI project decision-makers. Data was collected using an online survey of breast screening staff (adopters), semi-structured interviews with organisational leaders, participant observation of project meetings and document review. Data regarding organisational and adopter readiness for technology adoption was analysed over the duration of the project. RESULTS Sixty-seven clinicians and eighteen organisational leaders participated the study. Commitment to adoption is positive but adopters want to see clinical evidence of AI safety and accuracy. Decision-makers and other organisational adopters do not yet have shared views on their resources, capacity and capability to adopt and spread the technology and significant challenges related to task demands and situational factors emerged during the project causing substantial delays to adoption. The nature of AI and ML technology surfaced novel complexities not encountered by traditional health technology related to explainability and meaningful decision-support. CONCLUSIONS The case study shows that adopter commitment in this case and AI technology in breast screening is growing but gaps remain in the collective capability of organisations to adopt these novel technologies. CLINICALTRIAL Not applicable