Exploring the future – artificial intelligence (AI) and dermatology
The principles of embracing new and innovative technologies to reduce unnecessary outpatient attendances and improve access to care, were key messages that remain very important within the NHS’s Long Term Plan.
As a clinical dermatologist for nearly forty years, how might I use new technologies to diagnose and manage skin conditions? We read a lot about the potential role of artificial intelligence (AI) in healthcare but is it realistic to expect artificial intelligence as a medical device (AIaMD) to help counter the challenges faced by specialist dermatology departments?
Specialist dermatologists see and manage the care of people with a range of skin conditions and suspected skin cancer, the latter making up an increasing proportion of our lists. Over the last 10 years there has been an escalation by 170% of the number of people referred urgently for assessment of suspicious skin lesions, with an 82% increase in the Referral to treatment dermatology waiting list between April 2021 and March 2024. Ongoing work to recruit more dermatologists will make an impact in the future, but we need to ensure that people who are experiencing the damaging effects of long-term skin conditions, such as psoriasis and eczema, have equal access to care.
Of those people referred with suspected skin cancer, only around 1 in 10 have urgent skin cancer, according to NHS England cancer data, and the majority of referrals are people with benign skin lesions that don’t require any treatment. A neat solution would be to use AIaMD to diagnose the benign skin lesions referred in urgent suspected skin cancer pathways. This would free up capacity in dermatology services which would be welcomed by all. But is it safe to do this, or would patient care be compromised?
Examining Edge Health independent report
A new NHS commissioned report ‘Evaluating Pathways for AI Dermatology in Skin Cancer Detection’ by Edge Health, has reviewed the evidence around use of AIaMD in dermatology services and the results are encouraging. AIaMD has been used in skin cancer pathways with human supervision in the NHS for 4 years. In the UK it can be used autonomously (without human supervision) if it is appropriately registered by the MHRA after careful evaluation. The report evaluates the effectiveness of this type of certified device in diagnosing benign lesions in NHS skin cancer pathways. It also considers ways in which on-going safety of the AIaMD can be assured and the cost-effectiveness of the new technology.
The results, which reviewed over 30,000 assessments of skin lesions using AIaMD in NHS practice, showed that the tool achieved a performance that was at least as good as dermatologists in face-to-face clinical settings. With regards to ongoing checks of the effectiveness of the tool, the team suggested ways in which the safety of the tool could be regularly checked to provide clinicians and patients with the confidence needed to implement the technology autonomously. The report concluded that there was also the potential for the AIaMD to save money. Their analysis shows that for every £1 invested in the implementation of the technology, there is a potential saving of up to £2.30. In addition, they suggest that the rapid processing of initial assessments may reduce waiting times and improve service delivery.
This is not simply an academic exercise. Some NHS Trusts are already using AIaMD to help diagnose benign skin lesions, thanks to investment by NHS England. The evidence from these Trusts which has been peer-reviewed and published in journals, has been used to inform the Edge Health report. What can’t be conveyed through data are the interviews with clinicians who have told us that they have been able to see more patients who would otherwise be waiting for the care of a specialist.
There is no doubt that increasing the use of AI in the NHS will require reassurance that an adequate safety-net is in place. It is essential that, where a skin lesion is assessed by AIaMD as benign, patients are given detailed information about checking their skin and receive an SMS message from the Trust, 6 months after discharge. This message is to prompt them to check the lesion and include clear instruction on how to come back into the system if required. It is also important that the safety is ensured for people of different ethnicities and different amounts of skin pigmentation. This includes ongoing data collection amongst different ethnic groups and ensuring that the recent recommendations within the ‘Equity in medical devices’ review are taken into account.
In summary, the report makes clear that the use of AI holds considerable promise for improving the efficiency and effectiveness of skin cancer pathways. Evidence of its deployment in the NHS has demonstrated that whilst the tool could be used autonomously to exclude benign skin, adequate safeguards, such as those described above, will need to be in place. This provides the potential to free up specialists to focus their expertise on the most urgent and complex cases.
The successful incorporation of AIaMD into routine dermatological practice will require careful implementation including regular checks of any AI tool. Employing essential safety-nets for patients will help to ensure patient safety and maintain the high standards of care that the public expects from the NHS.