MEMORI v2: Improving the performance of an AI-driven Clinical Decision Support (CDS) system that supports the early identification of hospital-acquired infections (HAIs)
Every year in the UK, around 800,000 patients pick up an infection while in hospital, leading to over 28,000 deaths and costing the NHS £2.7 billion. Current early-warning tools like NEWS2 are designed to spot general patient deterioration, not infection risk specifically, and they often miss the warning signs. MEMORI is an AI system that fills this gap by predicting a patient’s risk of developing a hospital-acquired infection within the next seven days, updating that risk in real time as new data arrives. This gives clinicians a “window of opportunity” to intervene before an infection takes hold. The project will enhance MEMORI by feeding it additional data—such as clinical notes and lab results—and testing more advanced machine-learning configurations. The improved system will then be deployed and validated at the Royal Devon University Healthcare NHS Foundation Trust. If successful, MEMORI could reduce the severity of hospital infections, shorten the average 9.8-day extended stay these infections cause, and cut care costs. The tool has already shown better prognostic accuracy than NEWS2 in real-world use since June 2024, and clinical partners want it to perform even better.
View original technical description
MEMORI, an AI-based infection risk prediction tool developed by Human Digital Twin Ltd(trading as Sanome), aims to address the significant issue of hospital-acquired infections(HAIs). This grant proposes the enhancement of MEMORI.v1 through the incorporation of new data modalities and advanced machine learning configurations, followed by its deployment and validation at the Royal Devon University Healthcare NHS Foundation Trust. HAIs affect approximately 800,000patients annually in the UK, causing over 28,000 deaths and costing the NHS £2.7 billion p.a. due to extended hospital stays(average 9.8 day increased length of stay) and increased cost of care. Current scoring systems (e.g NEWS2) predict all-cause deterioration but are not designed for infection risk and has limited accuracy. MEMORI bridges this gap with advanced machine learning (ML) algorithms predicting HAI risk within 7 days. It updates risk levels in real-time providing clinicians with a "Window of Opportunity"for interventions, potentially reducing infection severity and improving outcomes and ultimately shortening hospital stays and reducing care costs. AI risk prediction is a growing field but MEMORI is unique in combining multimodal AI and direct integration into hospital EPR workflows. MEMORI was developed through co-design with clinicians and nurses at the RHN. MEMORI has demonstrated significantly improved prognostic accuracy compared to NEWS2\. It has been deployed in a real-world setting since June 2024 and has demonstrated strong user adoption. Despite promising results, clinical partners involved in the co-design of MEMORI.v1 have expressed a desire to further enhance its performance and expanding the "Window of Opportunity"for treating and managing HAIs. To address this, the project aims to significantly improve MEMORI by incorporating new data modalities, such as clinical notes and lab results, and by evaluating advanced ML configurations. Once these enhancements are achieved, Sanome will leverage strong interest from clinical groups in the UK and internationally to drive adoption and implementation.
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