Type 2 diabetes is a public healthcare crisis in the UK with 80% of the NHS diabetes budget spent on treating the complications associated with poor diabetes control. For the individual, these complications mean heart attacks, strokes, kidney failure and amputations, most of which are avoidable with early intervention and proper diabetes control. Diabetes patients often have long clinical histories and notes from various specialties but time pressures on doctors and nurses mean that, in standard practice, they are most likely to read only the most recent notes and may miss important symptoms or high-risk individuals.

To help solve this issue, Red Star AI is developing a tool to analyse the discharge letters and clinical notes of 110,000 historic diabetes patients. The tool uses a combination of Natural Language Processing and Machine Learning. The former is AI software through which a computer programme can absorb, understand and translate transcriptions of speech or text written by people, while the latter allows the programme to learn from its own actions and improve its performance as well as learn patterns. Red Star hopes to improve the diagnosis and treatment of diabetes patients, predicting risk of hospitalisation and non-response to standard therapies by finding similar cases and understanding patterns in treatment response. Working with both diabetes specialists and GPs and nurses in primary care, Red Star AI will design a user interface to automatically highlight and recommend interventions for high-risk patients. The prevention or delay in complications will enable the patients to live longer lives in better health and have potentially huge savings for the NHS.

AI for Good funding is to assist Red Star AI in moving its work into a clinical setting, using the models they are creating to identify high-risk individuals and validating predictions. The clinical evaluation of this technology is a key step in moving AI into routine clinical practice. Red Star is also exploring how the technology can be scaled up.

Currently, all NHS resources are directed towards COVID-19 making it difficult to visit GP surgeries and clinicians, so additional effort is being put into non-clinical evaluation and exploring the option of speaking directly with patients, with both patients’ and GPs’ permission. Red Star AI is also exploring how this project could be applied to diabetes patients who test positive for Coronavirus and use the datasets to monitor how the virus specifically affects diabetes patients.

This project is led by Andrew Conkie of Red Star AI in partnership with Data and Innovation group, NHS Glasgow, Digital Health & Care Institute Scotland and the University of Strathclyde. The project has received funding from Nesta in Scotland’s AI for Good programme.