P006 Evaluation of referral quality and barriers to artificial intelligence integration in a teledermatology 2-week-wait pathway

Abstract

Teledermatology plays a vital role in the assessment and triage of urgent skin cancer referrals, particularly within 2-week-wait (2WW) pathways. As artificial intelligence (AI) diagnostic tools are developed within dermatology, the quality of clinical information and imaging in referrals becomes increasingly critical. Suboptimal referral standards may compromise both the effectiveness of teledermatology and safe implementation of AI diagnostics. This service evaluation aimed to assess the quality of 2WW referrals received by a teledermatology service and identify practical barriers to the implementation of AI diagnostic pathways. A retrospective review of 292 consecutive 2WW referrals was conducted at a single centre; six patients were excluded due to withdrawal, leaving 286 for analysis (mean age 57.4 years). Data collected comprised referral clarity, lesion count, predefined exclusion criteria likely to compromise remote assessment (e.g. more than two lesions, and hair-bearing or anogenital sites), face-to-face general practitioner (GP) consultation, inclusion of GP-provided photographs, medical photographer concerns and triage outcomes. Referral clarity was graded as ‘excellent’, ‘acceptable’, ‘inadequate’ or ‘none’. Of the 286 referrals, 130 (45.5%) were graded ‘excellent’, 99 (34.6%) ‘acceptable’ and 48 (16.8%) ‘inadequate’, while 9 (3.1%) offered no descriptive detail. Most referrals (77.6%) documented a single lesion, while 26 (9.1%) documented more than two lesions. Overall, 57 referrals (19.9%) met at least one exclusion criterion. Even among referrals graded ‘excellent’, 20 (15.4%) met exclusion criteria, restricting potential AI applicability. Despite 91.3% of referrals involving face-to-face consultations, only 36.4% included clarifying photographs. Medical photographer concerns arose in 10 instances, and additional lesions were identified in 7.7% of photographed patients, prompting further referral. Twenty teledermatology referrals were deemed unsuitable and were converted to a face-to-face dermatology consultation. A significant proportion of teledermatology referrals lack sufficient detail or meet exclusion criteria that may preclude AI integration. Standardized referral templates, clearer documentation and greater adoption of high-quality imaging in primary care may enhance teledermatology services and their AI readiness.