Article Text

Download PDFPDF
089 Levers and limitations of artificial intelligence (AI) to support the assessment and implementation of shared decision making (SDM): perspectives of key stakeholders
  1. Anik Giguere1,2,
  2. Adrian Edwards3,4,
  3. Denitza Williams3,
  4. France Légaré1,2,
  5. Natalie Joseph-Williams4,
  6. Samira Abbasgholizadeh-Rahimi5,6,7,
  7. Karina Prévost8,
  8. Marie-Clare Hunter9,
  9. Anna Torrens-Burton3,
  10. Justine Laloux2,
  11. The AI-SDM Consortium10
  1. 1Department of Family and Emergency Medicine, Laval University, Quebec City, Canada
  2. 2VITAM Research Centre on Sustainable Health, Quebec City, Canada
  3. 3Division of Population Medicine, Cardiff University, Cardiff, UK
  4. 4PRIME Centre Wales, Division of Population Medicine, Cardiff University, Cardiff, UK
  5. 5Department of Family Medicine, McGill University, Montreal, Canada
  6. 6Lady Davis Institute for Medical Research, Jewish General Hospital, Montreal, Canada
  7. 7Mila-Quebec AI Institute, Montreal, Canada
  8. 8Patient partner, Montreal, Canada
  9. 9Patient partner, Cardiff, UK
  10. 10AI-SDM Consortium of the ISDM Society


Introduction Artificial intelligence (AI) is a promising avenue to advance the assessment and implementation of SDM. We explored stakeholder perspectives on the potential of AI to facilitate the implementation of SDM, focusing specifically on the assessment of SDM in healthcare professionals’ practice.

Methods Our environmental scan combined a web-based survey and targeted semi-structured interviews with key stakeholders. Participants included AI or SDM researchers and patient partners from our networks and from post-2015 studies using observer-rated instruments. The survey explored participants’ research experience, perceptions of levers and barriers for AI to support SDM assessment and implementation, and challenges in developing an AI system for consultation assessment. We then conducted semi-structured videoconference interviews with a subset of 6–10 participants who were purposively selected. Data analysis included descriptive statistics for quantitative results and thematic qualitative analyses for qualitative results, by two researchers based on the Consolidated Framework for Implementation Research.

Results (preliminary) To date, we have recruited 22 participants from seven countries - seven AI experts and 14 SDM experts, five of whom were practicing clinicians, and 10 owned observer-rated consultation datasets in different clinical settings. Levers in AI included developing decision aids or personalized information for patients to facilitate decision-making, and facilitating training in SDM through assessment in real or simulated consultations. Barriers cited included ethical concerns about privacy and confidentiality, reliability and transparency of AI systems, and reduced understanding of the complexity of human interaction. Barriers to implementation included cost, time, social acceptance of AI and data quality issues for AI training.

Discussion Most participants considered the use of AI for SDM assessment feasible. Half of the dataset owners indicated that they would make data available for AI if ethical and administrative concerns were addressed.

Conclusion(s) These findings will help inform future research agendas for AI-assisted SDM assessment and implementation.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.