Somerset NHS Foundation Trust has unveiled a comprehensive Artificial Intelligence (AI) policy, shared publicly by Andy Mayne, the Trust's chief scientist for data, operational research, and AI. The policy aims to steer the safe integration of AI technologies within the Trust, while maintaining a balance between innovation and ethical as well as legal responsibilities.
Published on LinkedIn, the policy underscores the importance of anticipating future advancements in AI technology, despite current models like Siri, Alexa, and Chat-GPT not surpassing narrow AI capabilities. The policy’s foresight ensures that protective measures are in place to accommodate future developments.
The policy is applicable to all staff within the Trust, including subsidiary company employees and contractors. It stipulates that processes involving AI, data meant for AI use, and technical systems integrating AI abide by principles of equality and inclusion. There is an express commitment to eradicating bias and discrimination within AI systems, promoting fairness, and ensuring transparency in decision-making. Design and implementation strategies are advised to prevent any discrimination based on protected characteristics. Furthermore, AI systems are to be rigorously tested with third-party suppliers required to provide datasets representative of Somerset's diverse population.
A focus on equal access to opportunities presented by AI systems, transparency in AI utilisation, and the need for registering new assets with the Trust’s information asset register is emphasised. Compliance with the DCB 0160 framework for clinical risk management is required to manage risks through the lifecycle of AI products.
To address responsibility and liability, the policy mandates clear accountability lines. Every AI model should include human oversight in its decision-making processes, with provisions for human intervention should a model generate incorrect or misleading information. It includes illustrative scenarios, such as liability lying with an employee misusing patient data through a large language model, and organisation responsibility if a third-party AI product is utilised with awareness of its inherent bias.
The document also highlights cybersecurity, enforcing evaluations of AI capabilities to ensure compliance and mitigate risks such as accidental information disclosures. It recommends excluding identifiable information from training datasets and encourages sourcing from companies that demonstrate transparent operational practices.
The policy further outlines the need for regular logging, review, and audit of AI systems, maintaining transparency by publicly disclosing AI systems using Trust data. A procurement checklist annexed to the policy includes criteria like AI ownership, bias checks, and human oversight.
In tandem with this policy, Mayne shared a condensed version for Trust staff to assist them in safely leveraging AI. Highlighted uses include the DERM system in dermatology for cancer detection, Microsoft's Copilot for workflow automation, and decision support in radiology. Predictive models for patient admissions and simulations for operational flows are also in use.
This initiative is part of broader trends within the NHS and healthcare sector, reflecting ongoing discussions around AI deployment, management of bias, and regulation in healthcare data. Regulatory concerns topped a recent LinkedIn poll regarding AI in healthcare.
The National Institute for Health and Care Excellence (NICE) recently launched new standards for improving transparency and quality in AI cost-effectiveness studies. These efforts coincide with wider exploration of ways AI is transforming healthcare, from enhancing diagnostics to personalising treatment, further evidenced by collaborations and acquisitions such as DeepHealth's recent incorporation of Kheiron Medical Technologies.
The development and adoption of AI in the NHS continue to advance with initiatives like HTN’s AI and Data Awards, recognising impactful AI technologies and innovations in health and care.
Source: Noah Wire Services