What do primary healthcare professionals think of using artificial intelligence to make decisions with patients with multiple long-term health conditions?

Problem

Living with multiple long-term conditions (MLTC) is now the norm for over 50 year olds in the UK. Combining the richness of GP healthcare records data with artificial intelligence (AI) technologies may identify new strategies to improve lives for such patients who are typically systematically excluded from clinical trials. AI technologies are being rapidly developed for use in healthcare settings. We are developing an AI tool to support prescribing decisions in patients with MLTCs. However, the perspectives of the GPs, practice nurses, and pharmacists who may be expected to use these technologies have not previously been explored.

Approach

We conducted 20 1:1 half-hour interviews with GPs, practice nurses, and clinical pharmacists. Interviews were conducted using a topic guide to explore perspectives on the challenges of managing complex multimorbidity, current understandings of AI, and the principles of using AI in an example consultation involving prescribing a new medication to a patient who already has four long term conditions. Transcripts were analysed using Framework analysis.

Findings

Clinicians find multimorbidity challenging. That complexity is driven not simply by the number of conditions a patient has but by patients' health literacy, lifestyle factors, and the challenge of applying multiple rigid guidelines to individual people's lives. Time pressures and lack of continuity are key organisational factors that will influence the potential success or failure of AI tools in clinical practice. Clinicians are very interested but sceptical about how AI may be used to support decision-making with patients with multimorbidity. Potential advantages include standardising good quality care and reducing human error. However, clinicians are concerned that use of AI in the consultation could impact on the doctor-patient interaction, and have unforeseen medicolegal and ethical implications.

Consequences

Overall, clinicians feel AI tools could support but not replace expert clinical judgement in managing MLTC. An increasingly complex patient population combined with declining numbers of GPs means new strategies are welcome. However, AI tools built using healthcare data need to be developed in collaboration with clinicians and patients to ensure that they have a positive impact and do not worsen existing inequalities.