From data request to publication: the lifespan of a real-world evidence (RWE) health research project

By Sophie Young, Health Research Lead 

Real World Evidence (RWE) is an important cog in a complex wheel when designing an effective health technology project. As a Health Research Lead at Discover-NOW, a hub for real world evidence and gateway to the deidentified data of a North West London population of over 2.5 million, I’ve seen first-hand how real-world data can support, shape and influence a project throughout its lifespan, supporting clinicians, researchers and scientists to accelerate the development of new treatments, devices and apps. An example of this is our project on predicting complications in Diabetes patients; the Risk Algorithms for Decision Support and Adverse Outcomes Reduction (RADAR) project, which was recently featured as a case study in the HDRUK and Health Foundation developing learning health systems report.   

Helping people in North West London (NWL) manage their Type 2 Diabetes (T2D) is a priority for NWL Integrated Care System (ICS), where 1 in 10 people have diabetes or non-diabetic hyperglycaemia (NDH), compared to 1 in 16 nationally. As the prevalence of T2D grows, the burden on healthcare is increasing in the form of additional consultations, hospital outpatient visits, death, and disability. In NWL alone, the cost burden of T2D healthcare is £600m annually, with complications of diabetes representing a significant factor in the morbidity and cost associated with this chronic disease.  

The HDRUK and Health Foundation funded RADAR project used our hub capabilities to work with our partners, NWL Health and Care Partnership, The Institute for Global Health Innovation (IGHI), Imperial College London (ICL), Imperial College Healthcare Trust (ICHT), MyWay Digital Health and Astra Zeneca. The collaboration used real world data to revalidate risk prediction models, to recruit patients to the project through use of the NWL Health Research Register. This was then used to facilitate development of proactive interventions through AI and Machine Learning, ultimately to deliver better information and visualisation decision support directly to front-line clinicians and patients.  

In NWL alone, the cost burden of T2D healthcare is £600m annually, with complications of diabetes representing a significant factor in the morbidity and cost associated with this chronic disease.  

This cross-industry collaboration between NHS organisations, academia and technology and research partners enabled the use of academic and commercial expertise, and key knowledge around evaluation, implementation, and scaling. The team included a number of healthcare professionals, which enabled us to link with a network of frontline staff and patients to support co-design throughout, as well as people living with diabetes who ensured that the tools developed were assessed and improved with direct input and feedback from people with lived experience. 

The project aimed to revalidate models which predict risk of complications in patients who have T2D. These models were first validated in the Scottish dataset (Scottish Care Information Diabetes) by project partners MyWay Digital Health. The aim was to see whether these models were applicable in the more ethnically diverse NWL Discover dataset. This model revalidation, and specifically use of metamodels which take advantage of the potential for models designed to predict one outcome (e.g., Type 2 Diabetes associated renal impairment) to predict another outcome (e.g., risk of incident amputation), resulted in improved prediction accuracy to provide calculated risk information (*AUROC = Area Under the Receiver Operating Characteristic Curve Median AUROC for the metamodels (0.79) was significantly greater than median AUROC of the base models (0.68)). 

These models were then linked to user interfaces for both clinicians and patients, which were tested in focus group sessions. Valuable feedback from these sessions led to changes to the tools to improve their usability and acceptability. The tools were finally assessed using ‘think aloud methodology’ as part of an evaluation led by IGHI, which used the NWL Health Research Register to contact patients to take part, in addition to NWL clinicians.

Patient and clinician feedback showed a real appetite for the use of this type of tool and the desire for patients to engage more with their own data. Feedback from focus groups suggested that enabling patients to view their own levels of risk and make changes accordingly, was also a driver for patient behaviour change, potentially impacting on short-term markers such as weight, glucose control, medication adherence, foot care and ultimately reducing complications. 

One of the reasons I’ve found this project so valuable is seeing the potential for impact of these tools for patients in NWL. This project provides a strong basis for further scaled evaluation of risk algorithm integration into clinical practice and patient self-management.  Though this project focussed on the usability and acceptability of these tools, further work is planned and underway by MyWay to assess the longer-term clinical impact of these tools, for example changes to HbA1c, Blood Pressure, Cholesterol, and national care process measures (e.g., eye/ foot screening, BP, weight measured in last 12 months). This further work to obtain more data is required to refine the implementation for highest impact and value for clinicians and patients alike.

This project demonstrates Discover-NOW at its best, utilising RWE to address local system needs through collaboration with industry, academic partners, and people with lived experience. Most importantly, the success of this project is just the beginning. Using it as an exemplar will provide the opportunity to use real-world data for the improvement of outcomes in patients with other long-term conditions too.  

Are you interested in working with us? Get in touch with