“It’s the future, come on!”: a think aloud study exploring clinicians’ use of knowledge-based AI decision support.

AI-based clinical decision support (CDS) is hailed as the solution to many healthcare capacity problems. However, there is a known implementation gap in AI CDS. Studies exploring barriers and enablers rely on abstract definitions or participants’ understanding of AI. Questionnaires often omit detail about specific settings or applications, relying upon participants’ prior knowledge to inform opinions. We address this research gap by giving an experience of AI CDS and exploring barriers and enablers through that experience. We used a phenomenological perspective with a think-aloud protocol and cognitive interview. We recruited fourteen participants (five doctors, nine physiotherapists) via social media and word of mouth. Participants used a prototype AI CDS designed to assess serious pathology in patients with low back pain. Analysis used a grounded theory approach to generate themes.Three main themes arose: Attitudes, Workflow and Patient Navigation. Whether attitudes were positive or negative, participants identified barriers and enablers in their work context “we ask all these questions anyway… why not input it to help us with our decision.” Workflow issues were identified, such as design, prompts and use of clinical reasoning alongside : “it is forcing you to ask … more evidence-based questions.” Facilitators identified were timesaving features, funding, and support from organisations. Forward navigation of the patient journey was of paramount importance. Exploring clinicians’ experience of AI CDS yielded new insights to practical application before implementation. Participants were solution-focused irrespective of initial biases. The study demonstrated how the CDS could prevent errors of judgement.