The Artificial Intelligence Simulated-Patient Tool for Clinical Training: a Mixed-Method Evaluation

Şeyma Handan Akyön, Fatih Cagatay Akyon, Ikbal Humay Arman, Gulsah Onur, Rita Abecasis, Michael Harris

Keywords: Artificial Intelligence, Medical Education, Primary Health Care

Background:

Family medicine requires skills that include communication, decision-making and managing undifferentiated patients. Traditional teaching methods are resource-limited and vary in effectiveness. We designed an Artificial Intelligence (AI) simulated-patient tool that overcomes these challenges by simulating realistic patient interactions, providing personalised feedback and creating accessible learning opportunities.

Research questions:

How do General Practitioners (GPs) perceive the AI simulated-patient tool in terms of ease of use, benefits, and realism?
What are their opinions regarding the advantages, disadvantages, and features that could be improved in the AI simulated-patient tool?

Method:

This mixed-methods study recruited European GPs via the European Young Family Doctors Movement (EYFDM), networks and snowball sampling. Participants used the AI tool and completed a survey with demographic questions, nine 5-point Likert scale items (ranging from 1=Strongly Disagree to 5=Strongly Agree), and three open-ended questions on strengths, weaknesses, and improvements needed. Quantitative results were analysed using median and interquartile range (IQR), and triangulated with thematic analysis of qualitative responses.

Results:

We had 21 participants with a mean of 6.3 years of work experience. The tool was rated as easy to use (median 4, IQR 2), enjoyable, beneficial for clinical practice and for giving evidence-based feedback (all median 4, IQR 1). Its ability to make an appropriate response and the conversational flow were positively rated (both median 4, IQR 2). Participants supported its use in education (median 4, IQR 2) and strongly recommended it (median 5, IQR 1). Strengths included its supportive role, usability, educational value, and communication skill assessment, while weaknesses involved interaction limitations, realism and cultural challenges. Suggested improvements included expanding content, adding multilingual and cultural adaptations, and enhancing interactiveness.

Conclusions:

GPs perceived the AI simulated-patient tool as user-friendly, useful and evidence-based, highlighting its supportive, assistive role and educational value while suggesting improvements in realism, functionality, and cultural adaptability.

Points for discussion:

How useful would an AI simulated-patient tool be for teaching in your own health system?

What do you think the main advantages and disadvantages would be?

What topics in the medical curriculum might an AI simulated-patient tool be most useful for?

#121