AI for Paediatricians

Addressing unmet needs in education

Abstract

Objectives

To describe a novel training approach ('AI for Paediatricians') developed and run by the London School of Paediatrics Innovation Group.

Background

AI is a rapidly evolving subject that has gained increasing attention in the last five years. There is great excitement about clinical applicability of AI tools, but also significant caution.

Clinicians, many of whom have had no training in the basics of AI, will be at the forefront of using, appraising, and overseeing the implementation of these tools. Without a firm understanding of the topics, they may struggle, resulting in deployment of unsafe or ineffective solutions, or undue hesitancy in adopting novel technologies that may have great potential for benefitting our patients.

Paediatrics faces particular challenges in adopting AI. Therefore, there is an even greater need to support paediatricians in understanding this field.

Methods

We developed a whole-day training programme, designed to educate clinicians about what AI is, how it works, and how it can go wrong. Our primary audience was resident paediatric doctors, although the day was open to participants from a variety of backgrounds.

Topics included the basics of AI; neural networks and large language models; how AI models are created; evaluation and appraisal of AI tools; how AI goes wrong; and post-marketing monitoring of AI tools. All the presenters were paediatric resident doctors with significant practical experience of clinical AI. We gathered pre-course feedback to characterise participants’ confidence and understanding around the topic of AI as applied to clinical medicine, and again after the course to assess the impact of the training.

Results

We found that participants, despite being self-selected as a group with an interest in AI, had low levels of confidence and understanding prior to the course. The training day significantly improved confidence and understanding of AI topics. Participants enjoyed the day, with a majority strongly agreeing that they would recommend the course to a friend/ colleague.

Conclusions

Training clinicians to understand and critically appraise novel AI technologies will be key to safe and timely use of such tools. Our training day was highly effective in addressing this need. Following feedback from participants, we will further refine the programme for the day, and we aim to run the next session in late 2025.