Hospitals and emergency care facilities are struggling with a ‘tripledemic’ of influenza, COVID-19, and respiratory syncytial virus (RSV). This, as well as a backlog of hospital appointments, a cost-of-living crisis that could push more vulnerable people into falling ill, and a workforce shortage, has put the NHS under pressure like never before.

In England, some hospitals have seen waits of up to 40 hours for a bed. New approaches are needed to meet the increasing health needs of our population, especially in acute care, and the provision of unplanned medical care such as out of hours primary care, ambulance assessment, emergency medicine, surgery and intensive.

Professor Elizabeth Sapey, a medical consultant in acute medicine and Director of the Institute of Inflammation and Ageing, first became interested in understanding the challenges facing emergency services prior to COVID-19. She saw first-hand the pressure on the emergency care health system and the need for more research into alternative approaches.

“Healthcare services need to be effective and resilient, able to deliver treatments that benefit patients and manage times of high demand without dropping the quality of care.  Health care research can develop and test new treatments, care pathways and new models of healthcare delivery, to make sure that you're always at the cutting edge. The data that the NHS collects to provide healthcare services can be an important part of that research, with the appropriate ethical checks and approvals,” says Professor Sapey.

After talking with academic researchers and industry, Professor Sapey discovered a broad consensus of the need for more research but faced two common obstacles: a lack of access to health data to develop better ways of caring for people, and too few academic researchers working on acute care challenges with clinical staff. Although acute healthcare providers collect a huge amount of data to enable care of patients, more needs to be done to improve care through research. 

In 2019, Sapey secured funding from Health Data Research UK, a national institute for data science in healthcare, to develop a Digital Innovation Hub focused on understanding and improving unplanned health services. The goal was to weave together data to provide a more holistic picture of patients and of services. The Hub, called PIONEER, aimed to gather data from primary care (GPs, pharmacies and out of hours GP services), secondary care (NHS hospitals) and the ambulance service across the West Midlands, to follow an individual’s acute care journey across community and hospital healthcare providers. “This allows us to understand the medical problems that people have, how they might coexist with other health problems and the kinds of treatments which might help people most,” explains Sapey.

One benefit of highly detailed, anonymised health data is that is can be used to create better ways to diagnose diseases earlier. “A lot of chronic medical conditions are diagnosed late in the disease process which can limit treatment options,” explains Professor Sapey. For example, the blood cancer myeloma can present with tiredness, paleness, poor kidney function, repeated infections and even broken bones due to bone thinning. “Individually, all of these things could be caused by a wide range of medical conditions, but only when you bring that information together you can see that this is classical for myeloma. However, people can present with these problems over years before the pattern is spotted and the diagnosis is made.  Computer systems might be much better at identifying these patterns, and PIONEER is working with researchers to see if we can develop systems which automatically alert doctors to potential health problems, helping healthcare professionals identify disease earlier”.

All health data projects delivered by PIONEER are reviewed by members of the public and patients, to make sure we have their support for health data access. This is important, as health data is sensitive, and privacy needs to be protected. However, the real benefits can come from research that uses health data. For example, one project PIONEER is supporting is looking for better ways to treat people who live in care homes when they become unwell. Sometimes people admitted to hospital can experience negative events, such as falls, hospital acquired infections and confusion. PIONEER is working with doctors and university researchers to assess what kinds of treatments could be safely delivered with patients staying at home, to reduce the unwanted effects of hospital admission. Another project seeks to understand the reasons that pregnant women and new mothers attend Emergency Departments for reasons that are not pregnancy-related, and to determine whether they could benefit from being assessed differently to woman who are not pregnant. Currently, health services have no specific system for triaging pregnant and newly postnatal women and in some cases, staff may lack familiarity with the specificities needed. PIONEER data can also help to understand which patients are at most risk of developing serious complications after treatments.  One project is developing predictive tools for identifying people most at risk of severe post-surgical infection.

PIONEER is based in Birmingham, a highly diverse city in terms of its population and socio-economic status. During this cost of living crisis, PIONEER is supporting researchers to look at the relationship between food banks and malnutrition and body weight in patients being admitted to hospital.  It is hoped that this data will map areas of greatest nutritional need to new food bank initiatives.  

Pivot to COVID

PIONEER started in November 2019 and quickly pivoted to COVID after the onset of the pandemic a few months later. Data from PIONEER revealed risks related to multimorbidity, and the heightened vulnerability of certain ethnic minority groups. They also worked with the Birmingham biomedical research centre to highlight inflammatory biomarkers; proteins in the blood that form targets in clinical trials and have now become part of the way we treat COVID.

“We built a COVID dashboard that's still in use today across the region,” explains Professor Sapey. “It was the first to highlight possible outbreaks of COVID in the community.  We shared that information with West Midlands public health to learn how the disease was spreading, and where personal protective equipment was needed”.

PIONEER provided insight on COVID ‘virtual wards’, where people admitted to hospital that meet certain parameters, can be managed at home with hospital care providers phoning in or collecting information about them.  “In some instances, this might be a really good idea, but we were able to show that in a really busy hospital with lots of COVID patients, a COVID virtual ward removed crucial resources from hospital wards and provided no improvements in patient outcomes, suggesting that deploying that kind of model wouldn't work if you're a hospital that had lots of patients coming through,” says Sapey.

“We are able to evaluate new service ideas with clinical staff at pace, to provide evidence as to whether a new care pathway is likely to work. That kind of information helps organise health services in a much better way”.

PIONEER was one of several high impact pandemic-related data initiatives at the University of Birmingham. Jean-Baptiste Cazier, Director of the Centre for Computational Biology, and a mathematician and modeller by background, played an instrumental role in the UK Coronavirus Cancer Monitoring Project, the first large-scale COVID-19 pandemic population-scale projects for cancer patients in the world.

It was a response to uncertainty at the time about how and whether to modify treatment approaches during the pandemic. “The question was, ‘if you have cancer, should you stop your cancer treatment, otherwise you might die of COVID?’ That was a rumour three years ago amongst patients and clinicians. Unfortunately, such a potentially dramatic advice was based on very little data,” recalls Professor Cazier. 

In three days, a team composed of data scientists and clinicians set up a project collecting data from cancer patients across the UK to understand the interactions between COVID and cancer treatment. They found no evidence that cancer treatments increased significantly the risk of COVID mortality, contradicting UK guidance at the time that anti-cancer treatment increased the risk of severe disease. The project was the first to perform an analysis by cancer subtype, identifying that only blood cancer patients were at significantly increased risk from COVID-19. And the team showed that vaccines were safe and effective in cancer patients, who were not involved in the original clinical trials.

Professor Cazier says artificial intelligence and modelling techniques can identify risk factors in new ways. “You can use AI to study the same data in a different way, completely blinded. And when we did that, we figured out age, sex and cancer subtypes are very important [to COVID mortality risk], but more important than sex and the disease type, is actually longitude and latitude. So, it is quite perplexing that out of the three big risks, the far more important one is where you are treated. We haven't managed yet to disentangle what it is linked to, but it hints that it could be ethnic diversity, socio-economics, or even access to care. It could be linked to anything. AI allows us to identify new risks or new factors which are critical”.

Professor Cazier says AI techniques can move researchers from an era of being “driven by clinical knowledge, to exploring the data to see what we can find. By just looking at the data with no a priori, the UK CCMP figured out that where people come from is very important. When we did the classic statistics, we couldn't figure that out because with classic statistics we ask specific questions as a normal way”.  Beyond COVID-19 and cancer, the integration of diverse data via AI can be used across the population, complementary to PIONEER. However, it is critical to gather the relevant data that will inform clinical decisions.

Data governance

Despite the benefits of a more seamless data ecosystem in the NHS, Professor Sapey acknowledges the need for public engagement given disquiet about health data privacy.  In the development of PIONEER, the team conducted interviews, webinars, and workshops with over 800 people in the local community to understand fears and concerns and how they could be mitigated. “Some of the things they highlighted were ‘making sure the data provided was enough to answer the question needed, but no more than that’; to ‘make sure the data was adequately de-identified to protect privacy’, and also to ‘make sure that we weren't providing data in a way that would exacerbate health inequalities’. They also wanted transparency and oversight”. 

The team set up a novel governance system, called a Data Trust committee, a group of 15 diverse members of the public, reflective of the city of Birmingham, who observe every request for data and 80% of whom have to approve data requests. The team have developed a number of additional safeguards to avoid re-identification of anonymised data, such as clustering some information together for individuals. They also conducted cyber-attack attempts to ensure the resilience of the system. “It is of fundamental importance to make sure that the public and patients trust what we're doing,” observes Professor Sapey.