Artificial intelligence - position statement

Advantages of AI in child healthcare

AI's potential impacts on the sector are many, varied and constantly evolving, but some of the key areas of benefit are:

• Genomics and multi-modal personalised medicine: When combined with advances in whole genome sequencing AI algorithms can analyse these highly complex data sets rapidly and to a high degree of accuracy, aiding clinicians in achieving confident and early disease detection and individual treatment planning.
• Population level clinical data integration and analytics: AI has the ability to connect and parse, within a reasonable time frame, huge population level health care data sets and provide analysis, insight and potential predictions important to the planning, resourcing and efficacy of health care programmes and services UK wide and globally.
• Advanced clinical imaging: AI approaches can enable significant increases in the speed and accuracy of clinical imaging, particularly in scenarios traditionally difficult, for very young babies and children for instance.
• Automated medical software development: AI enables non-technical individuals to create, augment and amend clinical software, usable in a range of devices and settings, rapidly and in a manner never previously possible.
• Operational efficiency: AI can significantly support streamlining of the majority of clinical administrative processes, allowing paediatricians and allied healthcare professionals to spend more time with their patients. At an individual clinician’s level the use of assistive AI approaches such as Ambient Voice Technology (AVT) offer significant improvements in the administrative burdens for paediatricians and their teams, be this in the taking of notes, the preparation of letters or through clinical decision support systems (when used appropriately). At departmental levels AI can significantly improve the alacrity, flexibility and efficiency of important organisational processes such as scheduling, triage, and workforce planning to help alleviate some of the known contributors to burnout, mental welfare challenges and organisational inefficiency.

In addition to the many other areas AI is impacting today, we can be certain its efficacy and capability will not only increase and broaden, but do so in increasingly rapid and impactful ways.

These opportunities, and their associated risks, need to be continually monitored for, assessed and, where proven appropriate and safe, taken up as quickly as the sector can in the pursuit of excellence in child health.

Challenges and ethical considerations

• Data privacy and security: Any use of AI must ensure the protection of sensitive patient information against unauthorised or inappropriate access. Vigilance must be exercised, for instance, when using AI for clinical purposes, that there is no confusion between personal and professional usage. Additionally, the increased availability of AI as a cyber security attack tool, and the high speed of change across AI systems in general, places even greater emphasis on taking a “secure-by-design” focused approach to all new digital healthcare capabilities, AI powered, and otherwise.
• Informed consent: As with all matter relating to child health data, the consent process for using AI in paediatrics must be both clear and appropriately adapted to the understanding of widest possible range of both parents and children.
• Bias and fairness: Both AI systems themselves, and any services powered by them, must be designed to avoid taking forward the inherent biases of the data upon which they are trained, and that might otherwise lead to unequal treatment of any patient groups.
• Summarisation and hallucination: AI systems must always be approached in the knowledge that they are prone, by design, to errors in both summarisation and generation. All responses provided by AI should be reviewed with a thorough and critical human eye before they are acted upon. In particular, sources provided should be manually verified as “citation hallucination” remains a pervasive and dangerous tendency across all large language models (LLMs) at this time.
• Transparency and explainability: Any AI leveraging decision-making processes must be fully transparent, with clear explanations provided to healthcare professionals and patients in relation to the factors, criteria and variables used to arrive at decisions and how these are processed.
• Regulatory compliance: AI applications in paediatrics and child health must continually comply with all existing, and relevant, future regulations, standards and best practice, ensuring child safety, treatment efficacy and value for money. Regulations themselves must also be continuously evolved to ensure they are always striking the correct balance between supporting innovation and upholding the rights of individuals.
• Social and environmental impacts: There is significant risk that uneven uptake of AI exacerbates existing social inequalities present within and without healthcare systems. Care and deliberate actions must be taken to ensure all stakeholders can take advantage of these opportunities in the fairest manner possible for all. Specific focus must also be applied to understanding, managing and mitigating the large environmental impact(s) of developing, training, deploying and maintaining AI systems.

Recommendations

• Child-centred design: In particular, across the paediatric sector, children, young people and their support networks must be at the very centre of all potential changes envisaged by the leveraging of AI and AI related technologies. Novel, child specific approaches should be explored and robust criteria applied to any considered re-purposing of adult focused techniques and / or technologies.
• Collaboration among stakeholders: Healthcare professionals, patients, AI developers, and policymakers should work together to guide the ethical development and deployment of AI in paediatrics and child health.
• Leading the way: Support of embedded clinical champions for AI across the healthcare system will be critical to increased rates of successful and safe adoption of AI in paediatrics. Resident doctors may well be particularly well suited to this role and should be engaged with pro-actively by all concerned to address their potential to act as immediate, and longer term, advocates for increased adoption here.
• Decision accountability: Safe use of AI requires of those involved a clear and unambiguous understanding of where the responsibility for clinical decisions influenced or supported by AI resides. As with all other tools which are used by clinicians, any associated decision must be made from the perspective that it is the trained human professionals that hold ultimate responsibility in all cases.
• Continuous education: The paediatric workforce must be enabled to educate themselves on the capabilities and limitations of AI, enabling them to integrate appropriate AI tools effectively into clinical practice as rapidly as possible.
• Robust evaluation and testing: AI systems must undergo particularly rigorous testing and validation to ensure they meet clinical needs, are free from errors and represent real world value for money.
• Public engagement: Patients and the public should be transparently informed about the benefits and risks of AI in healthcare, fostering a positive culture of trust between all parties.

Conclusion

RCPCH believes AI has the potential to revolutionise child health care, but its adoption must be managed responsibly to maximise the benefits while mitigating risks.

By adhering to these principles, we can ensure that AI serves as a valuable, and safe, tool in advancing the health and well-being of the children and young people we care for.

Footnote

This statement reflects a balanced view, acknowledging both the significant potential benefits and the many challenges that come with the adoption of AI in healthcare.

See all RCPCH position statements