Amitava Banerjee is professor of clinical data science at the Institute of Health Informatics (IHI), University College London. He is also honorary consultant cardiologist at University College London Hospitals and Barts Health NHS Trust.
Whereas his research once primarily focused on big data in healthcare, Professor Banerjee says it is now almost 100% based on nationally linked electronic health records.
Over the past three years, his work shone a light on the prevalence, incidence and outcomes of cardiovascular diseases in homeless individuals. He also provided early modelling to inform the initial Covid-19 response.
Professor Banerjee, who is the chief investigator of the STIMULATE-ICP long-Covid study, shares more about his research projects and why the trend of using data from electronic patient records to predict outcomes and prognosis is here to stay.
How did you become involved in long-Covid research?
As a cardiovascular doctor and researcher, I’ve got no business working in infectious diseases. However, early in the pandemic we found out there was a high risk of dying from Covid-19 or being admitted to a hospital or intensive care with Covid-19 if you’re older or if you have a chronic disease such as cardiovascular disease or diabetes. I became interested because my patients were asking me, ‘why am I more at risk and how much is that risk?’.
What work did you initially undertake?
The IHI and UCL are part of Health Data Research UK – the UK’s national institute for health data science. We were able to work together to look at national level health records for certain diseases.
We did some very quick modelling at the beginning of the pandemic using electronic health record data on the risk of underlying conditions. Our analyses were done before the first UK lockdown and informed Government decision-making.
We showed that if the infection rate reached 10% in the first pandemic, we’d have 70,000 deaths. In fact, we underestimated this, as there were nearly 120,000 deaths in the first year of the pandemic.
Since then, we’ve carried on doing lots of different analyses around Covid, whether it’s on vaccinations, inequalities or underlying cardiovascular disease.
Please can you tell us more about the STIMULATE-ICP long-Covid study?
STIMULATE-ICP (Symptoms, Trajectory, Inequalities and Management: Understanding Long-Covid to Address and Transform Existing Integrated Care Pathways) is a multi-pronged approach to finding out more about the disease and is the largest clinical study of long-Covid to date over two years.
In the UK, an estimated two million people have experienced long-COVID symptoms lasting four weeks or more. Initially, scientists and politicians focused on the short-term impact of the virus. This research aims to improve care for chronically ill patients and deliver knowledge and evidence to clinicians, scientists and policymakers while collecting real-world data at scale. More than 50 researchers, health professionals, patients and industry partners from over 30 organisations are involved. The team spans primary care and specialist services, epidemiology, mental health and health economics. It also includes four patient groups who helped develop the research proposals.
What’s the nature of the work and when are the results expected?
We’re examining how people with long-Covid progress and recover and how healthcare resources such as investigations and rehabilitation are being used. We’re also exploring health inequalities and comparing them with other long-term conditions. We aim to recruit 4,500 individuals with long-Covid.
Among the care pathways being trialled are a multi-organ MRI scan and a digitally enhanced rehabilitation programme. This means that individuals can access information and rehabilitation via a purpose-built app ‘Living with Covid Recovery’. We’re also looking at how drugs can be potentially repurposed to treat long-Covid.
We expect preliminary results late in 2023.
Cardiovascular disease is the most common cause of death in the homeless population. Tell us more about your work in this area
The idea for this research arose because the existing published literature tended to focus on practical clinical crisis management for when this cohort became acutely sick, as well as drugs, alcohol and hypothermia. But there was not enough thought about chronic disease in this population.
Our study used primary care data collected between 1998 and 2019 to compare 8,482 homeless individuals with 32,134 housed people. They were matched by age and gender and lived in the same general practice area.
We found that homeless adults were 1.8 times more likely to have pre-existing cardiovascular diseases compared to other adults, putting them at higher risk of severe Covid-19 and early death.
You set up a pilot CV screening service for the homeless population as a result of this research. Can you tell us more?
We were piggybacking on a mobile screening service originally set up for tuberculosis screening and increasing rates of flu vaccination. Essentially, a mobile van drives around hostels and street locations in University College London Hospitals NHS Trust’s Find and Treat Service. We introduce clinical cardiovascular checks for cholesterol and blood sugar, and we’ve shown there’s a high rate of undiagnosed risk factors. We’re now looking to scale this up.
How are you using national electronic patient data to inform research?
When I started my career, most of my work used large datasets. These were often research cohorts or registries and not from routine care at large scale. Now we work with hospitals and local primary care where research is based almost 100% on electronic health records.
I’m currently working on a heart failure project examining large scale electronic health records to see if there are different ways to subtype people who have different types of heart failure that are identifying by machine learning. Traditionally, images show how a particular part of the heart is functioning, which is a way of telling people what type of heart failure they have. I’m looking to see if there are other links that we haven’t found and subtyping people to see if that will predict their outcome or prognosis.
What are the advantages of using electronic patient data for research projects, and is this the future of healthcare research?
There are many advantages to using electronic patient data. First, data are more generalisable to the actual population and therefore findings are more likely to be relevant to the whole population.
Second, the coding of diseases and their risk factors can lead to standardised methods across studies.
Third, this type of data is more likely to create an ‘open science’ culture where our health data is seen as a public good.
Fourth, if more electronic health record research is done, there is less need for developing bespoke research cohorts, which are not efficient across individual diseases, settings and countries.
Amitava Banerjee is an advisory board member for HHE Clinical Excellence in Cardiovascular Care. He will be chairing two sessions at the event on 10 May 2023. Find out more and register for free here.