The BMA recently published their principles for Artificial Intelligence (AI) and its application in healthcare, so we thought this provides a great opportunity for us to share how we have implemented them at Skin Analytics over the last 5 years.
Full BMA publication & image credit: https://www.bma.org.uk/media/njgfbmnn/bma-principles-for-artificial-intelligence-ai-and-its-application-in-healthcare.pdf
BMA Principles:
Principle 1: A robust assessment for safety and efficacy in clinical settings must be carried out
At Skin Analytics, we have reviewed over 120,000 patients, forming our real world evidence. We have robust post market surveillance methods in place including feedback from our partners and via the MHRA yellow card scheme; monitoring the input data to our AI is appropriate through image quality and lesion suitability audits; and ensuring the output of our AI output continues to meet the highest levels of safety standards at the level of specialists, through proactive monitoring of performance with quarterly reporting and case audits.
Principle 2: Governance and regulation to protect patient safety is vital
DERM is the only AI as a Medical Device for Dermatology approved for use autonomously in the NHS with its UKCA class IIa approval. This status is only awarded to those who have passed thorough regulatory audits carried out by a designated body. Our robust approach to post market surveillance encompasses high standards being met across all domains; from data collection and sharing to training and intended use monitoring, from algorithm validation and risk management to performance monitoring; ensuring patient safety is always at the forefront of our practices.
Principle 3: Staff and patients should be proactively involved through development and implementation processes
We are constantly gathering patient and staff feedback through surveys, studies and regular meetings with providers to ensure we are addressing their needs and supporting them best. Results from a study done at one partner site showed 98% of patients surveyed would recommend the service to family and friends (1). The Department of Health and Social care evaluation showed 85% of patients perceived AI-enabled teledermatology positively rating it good or very good, with the majority of patients recognising the value in helping them get an appointment sooner. Staff also reported the effect of the AI teledermatology service on capacity as “transformational” (2).
Principle 4: Staff must be appropriately trained on new technologies, and they must be integrated into existing workflows
We designed educational resources to train staff on best practices to capture high quality images of lesions suitable for AI assessment, and our transformation teams work closely with clinicians and managers on the ground to ensure a smooth and effective pathway is in place, integrated with current workflows.
Principle 5: AI requires a robust and functioning NHS to be effective
Skin Analytics provides the necessary equipment to providers for capturing images at an appropriate standard, and lesions can be reviewed on our own platform to support this. Our transformation team deep dive into the end to end pathway design with our NHS partners to ensure we are optimising patient journeys, allowing them to access the appropriate care at the right time and right place.
Principle 6: Existing IT infrastructure and data must be improved
We collaborate closely with our NHS partners, ensuring a detailed data driven understanding of our pathways and their impact within the service, whilst following all appropriate data sharing and governance standards highlighted above.
Principle: 7: Legal liability must be clarified
Skin Analytics’ pathways create a clear distinction of legal liability as discharge decisions made by DERM are made autonomously, with no clinician involvement. This removes the possibility for AI errors to be attributed to reviewing clinicians. Skin Analytics bears legal responsibility for decisions made by DERM in line with its intended use.
DERM is the only AI as a medical device that can be used autonomously for support with diagnosis of skin lesions, backed by robust regulatory approval and substantial real world clinical evidence. Our pathways have assessed over 120,000 lesions, and detected over 12,000 histologically-confirmed skin cancers nationwide. Skin Analytics remains committed to ensuring the highest standards across the above domains are consistently being met or exceeded.
References