Implementing artificial intelligence in clinical practice: a mixed-method study of barriers and facilitators

Background: Though artificial intelligence (AI) in healthcare has great potential, medicine has been slow to adopt AI tools. Barriers and facilitators to clinical AI implementation among healthcare professionals (the end-users) are ill defined, nor have appropriate implementation strategies to overcome them been suggested. Therefore, we aim to study these barriers and facilitators, and find general insights that could be applicable to a wide variety of AI-tool implementations in clinical practice. Methods: We conducted a mixed-methods study encompassing individual interviews, a focus group, and a nationwide survey. End-users of AI in healthcare (physicians) from various medical specialties were included. We performed deductive direct content analysis, using the Consolidated Framework for Implementation Research (CFIR) for coding. CFIR constructs were entered into the Expert Recommendations for Implementing Change (ERIC) to find suitable implementation strategies. Quantitative survey data was descriptively analyzed. Results: We performed ten individual interviews, and one focus group with five physicians. The most prominent constructs identified during the qualitative interim analyses were incorporated in the nationwide survey, which had 106 survey respondents. We found nine CFIR constructs important to AI implementation: evidence strength, relative advantage, adaptability, trialability, structural characteristics, tension for change, compatibility, access to knowledge and information, and knowledge and beliefs about the intervention. Consequently, the ERIC tool displayed the following strategies: identify and prepare champions, conduct educational meetings, promote adaptability, develop educational materials, and distribute educational materials. Conclusions: The potential value of AI in healthcare is acknowledged by end-users, however, the current tension for change needs to be sparked to facilitate sustainable implementation. Strategies that should be used are: increasing the access to knowledge and information through educational meetings and materials with committed local leaders. A trial phase for end-users to test and compare AI algorithms. Lastly, algorithms should be tailored to be adaptable to the local context and existing workflows. Applying these implementation strategies will bring us one step closer to realizing the value of AI in healthcare.