Enhancing Type 2 Diabetes treatment through peer-learning AI: Creating a support system for healthcare professionals in primary care to source and share optimal self-care interventions and best practice in order to improve care and efficiency

点击翻译按钮获取中文摘要

**NEED:** The management of Type2 Diabetes costs the NHS £8.8bn per year with patients estimated to account for 15-25% of all appointments at a local surgery level.The 'General Practice Forward View' identifies that all patients with long-term conditions should have a personalised plan of care that includes self-care, social prescribing and active signposting as part of its 10 high impact actions within 5 years.Innovations such as Artificial Intelligence (AI) can play a key role in this - however at present there is no market-ready AI-system to assist Healthcare Professionals (HCP) to do this effectively within their time pressures, in part due to the difficulties associated with deriving, machine taught (machine learning), clinically-safe recommendations from the huge amounts of patient data available to digital health companies.**APPROACH:** The project brings together Healum Ltd (digital health platform developer), Vernova CIC Foundation (NHS), Manchester University (Subcontractor) to develop a peer2peer Learning AI and AI platform to support the self-management of T2D patients through personalised plans of care, support, behaviour change and education.Working with 11 GP Practices and with SBRI support (Project Number:8625456) we have already established feasibility of approach, developed an a prototype collaborative self-care platform and innovative peer-learning algorithm.**FOCUS:** The project's focuses on the refinement of peer2peer algorithms and development of AI, machine learning (ML) algorthims, SnoMed code integration, content classifier and collective intelligence recommendation engine as well as aligning and evaluating the platform in the treatment of Type2 Diabetes (T2D).By harnessing the collective peer2peer intelligence of HCP inputting into the software, combined with aggregate anonymised clinical audit data,we are able to reduce the critical mass of data inputs required to train ML algorithms and provide automatic clinical recommendations with a greater level of data confidence.**IMPACT:** The platform will be integrated with and make use of new SnoMed codes and historical Read codes to act as a classifier - supporting the recommendation of the right content, service and plan to the right patient at the right time - and trialled with 21 GP practices during the project to to form a peer2peer community for providing personalised plans of care, support, behaviour change and education to patients with or at risk of T2D.By improving the management of T2D care workload, we estimate potential for cost saving per practice per annum of £11,304, increasing to £61,254 by Y5 once reductions in microvascular complications are factored in.