Impact on education and learning needs

Introduction

Health and social care workers will need to adopt additional skills to work effectively with Artificial Intelligence (AI).

A recent survey of 210 doctors found that 92% reported insufficient AI training in their current curricula9.

This suggests that some of the time that was previously spent on memorizing medical/clinical/health information may need to be devoted to the development of new skills in order to safely and effectively use AI systems and tools in practice.

Proposals for incorporating AI into educational training included skills such as:

  • knowledge of mathematical concepts
  • fundamentals of AI and data science
  • related ethical and legal issues
  • data input skills
  • communication of the outputs of algorithms
  • ability to communicate AI derived results to patients

This has been adapted from Paranjape, K., et al. (2019) and Banerjee, M., et al. (2021).

In contrast to these skills and knowledge, emotional intelligence will also increase in importance when dealing with patients physical and emotional states12. More advanced capabilities surrounding domains like AI and robotics are in turn dependent on the use of patient data. This data and it's quality is integral to the success (or otherwise) of an AI system13.

The success of data-driven AI projects can be further increased through the application of a collaborative team based approach which combines interdisciplinary skills and knowledge with domain knowledge13. It is not possible for people to be experts in all areas given the complexity of the medical/health and technology domains, instead multidisciplinary team work and an understanding of roles with the development of a common shared language and understanding can help move us toward a successful digital future.

Case Study

Case Study: Assessing organisational readiness for embedding Artificial Intelligence (AI) tools in the plan of action phase of an orthopaedic pathway

Predicting outcomes is a key part of the patient care pathway before undertaking orthopaedic surgical procedures. Artificial Intelligence (AI) tools exist to enhance the accuracy of risk prediction for surgical outcomes by clinicians using data held within Electronic Health Records. But evidence of such tools being embedded in practice is limited. Health care organisations need further understanding on why AI tools might be difficult to embed in organisations including what costs and risks they may face.

We worked with Ramsay Health Care UK to perform a feasibility study to assess organisational readiness for implementing Artificial Intelligence/Machine Learning (AI/ML) tools in the pre-operative assessment phase of an orthopaedic service and understand the social and technical costs/risks involved in the transformation.

We undertook an agile-inspired approach utilising the LOAD model and data journey model to map the existing data landscape of the organisation and identify potential socio-technical barriers and risky areas when implementing the AI tool. Domain expertise was gathered from 6 subject matter experts through semi-structured interviews. Potential next steps that mitigate the highlighted costs/risks are suggested.

The data journey model clarified that the organisation does collect all relevant data required to produce an evidenced based AI/Machine Learning (ML) risk prediction tool. Significant boundaries to effective implementation were predicted such as limited access to structured data, actor interactions introducing errors, costs of changing the media of data and the costs of transferring data across organisational boundaries.

Results showed that whilst the organisation does collect the relevant data required to produce an evidenced based AI/ML risk prediction tool, significant social and technical boundaries to effective implementation were predicted. Limited access to structured data can risk ineffective deployment of the prediction tool. The need for the AI/ML tool to access the required type of data will result to significant process transformation costs and heavy reliance of actors to consume and transform the data types between existing data containers.

Organisations should understand their existing data landscapes, data journeys and associated social and technical costs and risks of implementing AI/ML based tools before embarking on such projects to prevent unforeseen costs or failure.

References

5 Paranjape, K., Schinkel, M., Panday, RN., Car, J., Nanayakkara, P. (2019) Introducing Artificial Intelligence Training in Medical Education. JMIR Medical Education 5(2):pp1-12

9 Banerjee, M., Chiew, D., Patel, KT., Johns, L., Chappell, D., Linton, N., Cole, GD., Francis, DP., Szram, J., Ross, J., Zaman, S. (2021) The impact of artificial intelligence on clinical education: perceptions of postgraduate trainee doctors in London (UK) and recommendations for trainers. BMC Medical Education 21:429

12 Wartman, SA., Combs, CD. (2018) Medical education must move from the information age to the age of artificial intelligence. Acad Med 93(8):pp1107-1109

13 Davies, AC., Davies, A., Wilson, A., Saeed, H., Pringle, C., Eleftheriou, I., Bromiley, P (2021) Chapter 17: Working as an AI Specialist in The Health Information Workforce. Switzerland: Springer

Page last reviewed: 14 February 2023
Next review due: 20 February 2024