Machine Learning for Paediatric Related Decision Support in Emergency Care – A UK and Ireland Network Survey Study of Emergency Staff
Abstract
There is great potential for artificial Intelligence (AI) and machine learning (ML) to support decision making in emergency departments (ED), however their implementation in routine clinical practice remains limited. The objective of this study was to assess the understanding, experience and perspectives of the wider paediatric ED workforce (nurses, doctors and support staff) in the United Kingdom and Ireland on the use of ML decision support tools. A voluntary and anonymised survey was carried out across 75 sites. The survey consisted of questions on self-reported knowledge of AI concepts (which included watching a short video), exposure, barriers to adoption, training and potential ML applications. Mostly quantitative analysis was performed. The survey had a 72.3% response rate (660 responses). Prior to viewing the video, understanding of AI concepts varied, with AI the most understood (60.3%) compared to deep learning at 19.1%. Post-video, 40.2% of respondents changed their answers. While many respondents had experience of decision rule-based systems (53.3%), only 7.7% reported using ML based tools. Key barriers to adoption included uncertainty about suitable clinical applications (42.8%), lack of skilled resources (example: data scientists or engineers) (37.0%), limited model explainability (31.5%) and poor data quality (28.8%). Many respondents believed these tools could integrate well into clinical workflows (60.7%), would trust these tools (58.2%), and had a strong interest in furthering their knowledge in ML (78.2%). Early warning systems and radiology applications ranked highest, and diagnosis of mental health conditions lowest, for doctors and nurses. These findings show knowledge gaps and limited exposure to ML tools, yet a strong interest in learning is evident. To realise the potential of ML in children’s emergency care, domain specific AI literacy, improved model transparency, investment in infrastructure and resources, and better integration into clinical workflows are essential.
Author Summary
In this study we set out to understand how the staff (nurses, doctors, and support staff) in children’s emergency care across the United Kingdom and Ireland perceive and engage with machine learning decision support tools. Whilst much research into these tools exists, few are used in clinical practice. To understand why, we conducted a multi-centre survey, asking participants about their understanding of artificial intelligence concepts, past use of machine learning tools, barriers to adoption, and views on the most useful clinical applications. We found that although familiarity with machine learning and its application was low, interest was high, with many believing these tools could be integrated well into clinical workflows. Early warning systems and radiology were viewed as the most promising uses. Barriers included uncertainty around where machine learning could add value, not understanding how these tools generated their output, lack of skilled resources to implement these tools, and concerns around poor data quality. Our findings on the lack of artificial intelligence literacy align with surveys internationally. Addressing these challenges, along with providing targeted training tailored to each role, will be key to supporting safe, confident, and effective use of machine learning tools for decision support in children’s emergency care.
Related articles
Related articles are currently not available for this article.