P044 Multi-centre service evaluation of the rheumatology advice line with suggestions for developing a national solution using artificial intelligence

Abstract

Abstract Background/Aims Advice lines are essential for the delivery of rheumatology services. The lowering of disease thresholds for biologic and advanced therapies, together with the ageing population, means a greater complexity of patient-care and increase in the number of adviceline queries. However, adviceline logistics vary considerably, including modes of contact. Timely response is imperative and the pressures resulting from this are significant. Service redesign since the Covid-19 pandemic has led to an increase in virtual approaches and its acceptance. Therefore, a team of clinicians from across the UK met to devise an artificial intelligence (AI)-assisted solution to support advice line delivery. Methods A task force was set up to explore current advice line provision. This task force analysed the various advice line types and mapped potential points where AI could support delivery. Additionally, a retrospective one-month audit (August or September 2025) of pooled data from eight centres in the UK (Berkshire, Chesterfield, Hertfordshire, Luton, Portsmouth, Poole, Southampton and Stoke) was conducted to establish the most frequently asked questions (FAQs) to advice lines, including the number and type of calls. These information sources are being used to build and develop an AI agent to handle online queries. A second workshop is planned, together with third-party representatives from patient charities and the British Society for Rheumatology, to test the AI agent and establish how this can be piloted at scale. Results All rheumatology service websites from across the four nations were reviewed. Of the 146 services, 21 didn’t advertise an advice line number and only 39 offered an FAQ page, ranging from one to 48 questions. The consolidated audit data from the eight centres showed the total number of calls in a month was 5327. The average call per centre was 666 (range 113 to 1423). The methods of contact to the advice line were direct calls 1656 (42.4%), online portal 1231 (31.5%) and answering phone 843 (21.6%). The types of calls were clinical routine (1494, 28.1%), administrative (1188, 22.3%), clinical urgent (712, 13.4%), non-rheumatology (180, 3.4%) and red flag (103, 1.9%). Over 40% of calls to the advice line were clinically-related: routine, urgent and red flags. Around 1 in 8 (13%) calls were urgent and 1 in 50 (1.9%) were red flags. This highlights the triage importance of advice lines. Administrative calls (22%) remain a major burden. Non-rheumatology calls remain a small (3.4%) but noticeable fraction. Conclusion A nationally deployed AI agent that addresses the most common frequently asked questions has the potential to reduce the burden of advice line queries. Centres with large call volume, high acuity and heavy administrative queries would benefit from AI triage and structured pre-screening. Disclosure S. Ryan: None. J. Begum: None. A. Eden: Honoraria; speakers fees for UCB Novartis and Fresenius. P. Cornell: None. A. Chan: Honoraria; Abbvie, Novartis, UCB. Member of speakers’ bureau; abbvie, Novartis, UCB, Medac. K. Fairman: None. N.R. Fuggle: None. L. Weizi: None. L. Sammut: None. M. Yates: None.

Ähnliche Arbeiten