Every winter, respiratory admissions to English hospitals double, topping 200,000 emergency cases in 2021–22 and overwhelming the health service. Existing clinical decision-support tools cannot handle the complex, incomplete, and fast-changing data needed to spot deterioration early. This project refines a proof-of-concept AI system called SLAIDER, which creates a "digital twin" of each patient—a model that uses both historical and real-time data to predict their future condition at any point during their stay. The team will test SLAIDER on data from multiple hospitals serving diverse populations, identify where it falls short, and make targeted improvements. They will then verify that the refined system gives accurate predictions across different patient backgrounds. If successful, SLAIDER could help clinicians prioritise patients more effectively and intervene earlier, especially for those initially admitted with low or moderate severity who later worsen. This would reduce preventable deaths, ease winter pressures, and cut the strain on emergency departments. The work is a necessary step before clinical validation and eventual hospital deployment, and it is funded under the MRC’s Artificial Intelligence highlight notice.
View original technical description
Respiratory-related hospital admissions remain a major concern in the UK. In England alone, there were over 200,000 emergency hospital admissions in 2021-22 – a number which has continued to rise. The effect of this is most apparent during winter, when respiratory-related admissions double due to 'winter pressures', leading to an overloaded health service and preventable deaths. There is an urgent need for better clinical decision support tools, which can help clinical staff make smarter decisions - by prioritising patients more effectively and support real-time identification of changes in patient condition and deterioration. Existing tools lack the adaptability and predictive power needed to cope with complex, multidimensional, and often incomplete data in a timely way. Our prior research has led to the development of a proof-of-concept clinical decision support tool (SLAIDER) based on a concept known as 'digital twins'. This innovative Artificial Intelligence (AI) based approach can predict a patient's future state at any point in the patient's stay using both historical and real-time data. This will allow earlier prioritisation and clinical interventions – leading to improved patient outcomes, particularly in patients who are initially admitted with low or moderate severity conditions. In this gap project, we aim to evaluate and improve SLAIDER using data from multiple hospitals. We will use this data to first test the proof-of-concept technology and identify where there are weaknesses, then use these findings to make targeted improvements to solve these problems. Finally, we will show that the refined technology works correctly and provides appropriate predictions for patients from across diverse backgrounds. This will provide key evidence for further development of the technology including feasibility studies, clinical validation and ultimately, implementation into hospitals. The project will collaborate with healthcare partners and hospitals serving diverse populations. We will work closely with a broad range of people including academics, clinical staff, healthcare informaticians, computer scientists and commercial/development experts to ensure project success. This regular feedback will help ensure SLAIDER meets the real-world needs of healthcare professionals. Led by a multidisciplinary team, this timely project will fill a key gap in the clinical implementation of SLAIDER, an innovation which has the potential to transform patient care, improve health outcomes and significantly reduce pressures on the healthcare system. This project has been submitted within the remit of the MRC Artificial Intelligence highlight notice.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know