BT06 Using artificial intelligence to triage skin cancer referrals: outcomes from a pilot study

Abstract

Artificial intelligence (AI) is a rapidly emerging field in dermatology, aimed at delivering efficient and effective patient care. To date, there is a lack of substantial evidence for the use of this technology in a clinical setting. However, the pandemic era has placed significant pressures on dermatology services, specifically the skin cancer clinics at our trust, which has experienced a 30% increase in demand compared with prepandemic levels. A novel solution for urgent referrals was to test the use of AI teledermatology software using a UK Conformity Assessed marked Class IIa Medical Device (and therefore regulated by the Medicines and Healthcare products Regulatory Agency). Appropriate cases are selected and booked for a teledermatology appointment at one of our four community-based hubs. Clinical and dermoscopic images, together with a short questionnaire, are uploaded to the software by a healthcare assistant. The AI immediately assesses the dermoscopic images of the lesion, and those cases identified as benign are discharged back to the general practitioner (GP), following a safety check by a consultant dermatologist employed by the AI provider. The trust dermatology consultants have chosen to further triage all cases suspected to be nonbenign. Patients with suitable small lesions at low-risk sites receive a telephone appointment to discuss surgery and obtain consent. Those with precancerous lesions can be discharged back to the GP with management advice. Low-risk basal cell carcinomas case are transferred to a routine pathway. Selected cases may be referred directly to other teams. The pilot started in March 2022 and now processes 5–600 cases per month at four photography hubs. To date, 19% of cases have been discharged back to the GP with a benign diagnosis. A further 15% of cases have been discharged back to the GP by the trust dermatologists, giving an overall discharge rate of 34%. Ten per cent of patients have received a telephone call to take consent for biopsy, and 2% have been referred directly to another speciality. Thirteen per cent of cases have been triaged to a routine pathway, and 38% of patients have required an urgent face-to-face appointment with a dermatologist. The trust has seen > 90% 2-week wait referrals within 2 weeks in the last two quarters. Since the introduction of this AI teledermatology service, we have seen a 62% reduction in the number of patients requiring an urgent face-to-face appointment with a dermatologist. Ongoing audit work is focused on the safety, acceptability and cost-effectiveness of this pathway, and will shape decision-making in the ongoing use of this technology both locally and across the UK.