Interview: Ceire Costelloe and Lisa Scerri on using health data to improve cancer outcomes at The Royal Marsden NHS Foundation Trust

We recently sat down to chat with Ceire Costelloe, professor of health informatics at The Institute of Cancer Research, London, and Lisa Scerri, business manager of BRIDgE (Biomedical Research Informatics Digital Environments) at The Royal Marsden NHS Foundation Trust. Ceire and Lisa discussed how health data is being used to improve outcomes across the cancer pathway at the trust, as well as the role of BRIDgE in enabling researchers to access real-world cancer data within secure, collaborative and cloud-based workspaces.

Hi Ceire and Lisa, thanks for joining us. Can you introduce yourselves and tell us a bit about your roles?

Ceire: I work across the Institute of Cancer Research (ICR) and The Royal Marsden. I’m a medical statistician by training, and much of the work from my research group focuses on trying to make the best use of routinely available clinical data. I lead a team of data scientists, epidemiologists and statisticians who are doing lots of work using highly granular electronic patient record data, as well as other types of information, including population health data and global data. Essentially, we are trying use this information to improve the cancer pathway, including boosting early diagnosis of cancer, optimise and inform treatment strategies, and to support long-term patient follow-up.

Lisa: I have an operational role at the Royal Marsden, encompassing both IT and research. A lot of my work is around helping with data-driven study design and set-up, often right from the kernel of an idea from a researcher, and supporting them to utilise the technology that we have at The Royal Marsden. For the past four years, I’ve been involved with a new programme of digital transformation at The Royal Marsden which has included a new EPR, and a cloud-based data warehouse where we bring in all clinical data, not only from the old EPR but also from the new. On top of that we introduced a trusted research environment: the Biomedical Research Informatics Digital Environment (BRIDgE), powered by an Aridhia platform, which I now manage for the trust.

BRIDgE: Biomedical Research Informatics Digital Environments

Ceire: I’ve been working in health informatics for about 12 years now. When I started off, I was using a lot of paper-based patient data that was then translated into datasets or Excel spreadsheets that we would use for statistical modelling. Over the past five to 10 years the big development for me has been really having the ability to access individual patient record data to do my research. The infrastructure that has been developed in this area, such as BRIDgE, has really facilitated this work.

Lisa: BRIDgE provides remotely accessible workspaces to which we can either provision data or allow users or sites to upload their own data for analysis. The platform does not allow data output without authorisation. It’s totally self-contained, which is what makes it so secure for patient data. One of the best things about the platform is that it’s collaborative – you can have multiple users working in a single workspace at the same time, from multiple organisations. Once researchers have approval for their study, they register with the BRIDgE platform using their chosen email address and we ensure that they get access to the workspace which is designated for their particular research project. Most of our researchers tend to work collaboratively with other centres; take Ceire’s work across the NHS and the ICR as an example. Previously, sharing NHS data could be a challenge between institutions but by using BRIDgE it’s relatively simple to do and the increased security makes setting up sharing easier.

Ceire: Through BRIDgE, I as a researcher can safely access anonymised data on patients throughout their healthcare journey, capturing every aspect of healthcare delivery, from diagnostic tests and medicines prescribed to the location of the patient and how they move within a hospital environment. That helps us to make decisions around diagnostic criteria, optimal treatments and new trials and therapies, in a much more detailed way. It’s been a real game-changer in terms of the kind of work that analysts can do.

Challenges, and tackling them

Ceire: One of the challenges that we face in the UK is that we’ve had such a well-established healthcare system for so many years. But it’s been predominantly paper-based, so we are now making the translation from paper into digital and there is a vast amount of data involved. The rate at which this digital translation is happening across the UK varies widely, with different levels of digital maturity across different NHS centres

A project we are leading at the ICR and The Royal Marsden, in collaboration with other specialist cancer hospitals, focuses on taking this wealth of EPR data, collating it in BRIDgE and using artificial intelligence and machine learning algorithms to build clinical decision support tools that can provide an early warning to clinicians that a patient might be at risk of neutropenic sepsis, a side effect of cancer treatment. We’ve been working with NHS trusts that have various levels of digital maturity. Some of them are using paper-based records, some are using bespoke EPR solutions that they have designed, some are using the highest quality systems. From a research perspective, we have been able to showcase what we can do with the model that we’ve set up. The culmination of this work will be to collaborate with the clinical and the informatics leads across all of the trusts so that we can implement the model into practice to ensure rapid diagnosis and timely treatment for neutropenic sepsis.

Lisa: BRIDgE overcomes many of the challenges associated with data-driven research, especially AI and machine learning, because data upload from multiple sites is easy and the collaborative workspaces provide tools for analysis and coding. Data interoperability can be another challenge, especially if you’re collecting retrospective data from patient medical records – Ceire’s team have had to do a lot of work to align datasets from different trusts. In the future, there will be more use of the standard data model – Observational Medical Outcomes Partnership (OMOP) – and transforming datasets into OMOP for federated analysis. That means that instead of asking those collaborating trusts to share their data with us, which could cause interoperability issues, queries could be sent to the data and the result returned, without any data having to move.

Looking back: advice and learnings

Ceire: If I could go back to the start of my career and give myself one piece of advice, I think I would stress the importance of taking a collaborative approach. Collaboration is key to making the best use of data. As an academic researcher, when I undertake or plan a project, I need to spend time getting the stakeholders and all of the people working on the data together. I need input from clinicians, data warehousing staff, informaticians, statisticians, with academics and NHS workers all operating together. Having a strong sense of collaboration has enabled us to be successful in many of our projects in recent years.

Essentially we need to understand the complexity of the data that we are working with, and operationalise, or curate the data . That requires skills that a computer scientist might have. There is a real need for highly multi-disciplinary teams.

Finally, understanding the research governance and ethics processes, and how they are evolving as local and national infrastructure changes, is key.

Lisa: Similarly, my advice would be to work with people from different disciplines as much as possible and don’t assume that your skills aren’t transferable to other careers. I’ve had quite a varied career, across research and IT. When I started working in computing at the Royal Marsden after having worked in research, I hadn’t realised how useful my research knowledge would continue to be; it is the combination of my work experience that has led me to the job I do today, managing a Trusted Research Environment.

Interdisciplinary working is so important, especially for data-driven research. For example, I was involved in setting up the Data Stewards Group, which advises researchers with issues around data sharing and governance. We found that by having research and development, information governance, legal and informatics colleagues all in one room advising researchers together, we could help people whilst learning from each other. Now, I think it’s fair to say that we save teams months of work by talking their challenges through and directing them clearly, rather than them bouncing between departments trying to find the knowledge that they need.

What do you expect to see from digital and data in 10 years’ time?

Ceire: I think we’ll see increased opportunities for data availability – that’s already started and it’s been transformational for the work that I do. I think we will also see improvements in integration of data. I do believe that The Royal Marsden and The Institute of Cancer Research are leading in this area, because they have access to clinical data from the hospital plus molecular and genetic information coming from the ICR research laboratories.

There are two big things I expect to see in the future which I’m quite excited about. The first is the opportunity to integrate not only laboratory, clinical data and genetic data but also patient-generated data. There’s a vast amount of information that is being collected by individuals outside the healthcare system. The utility of that being used to support early prediction of cancer diagnosis, for example, or better understanding side effects of treatment could really be harnessed to speed up diagnosis, and to personalise cancer treatment strategies. I think and hope in the future that this type of information will almost seamlessly be integrated into the data that I work with. This would facilitate ‘digital’ and decentralised trials, where information can be collected remotely, rather than patients having to come into clinic.

Another project is using an app ‘Wysa” to record mental health outcomes for patients who are currently awaiting NHS treatment. The Wysa app helps promote improvements in wellbeing and provides a chatbot function to guide users on self-help tools. Importantly, we are able to integrate the data from that app directly into the patient’s NHS electronic patient record. It facilitates communication about mental health, because the clinician has access to information on the patient’s mental health whilst they are outside the healthcare service. I think that type of integration of data is key and will become more common in the future.

The second thing is the idea of improving our trials by using patient data for long term follow up. If we can integrate data that is collected on a patient during a clinical trial, with the patient EPR, we can then ensure we have long-term follow-up of patients. I think that’s really important; as more people are surviving cancer, we’re placing a lot of focus on predicting and managing recurrence.

Lisa: Picking up on that point, expanding the digitisation of clinical trials is something that I expect to see in the future too. With around 800 clinical trials open at any one time at The Royal Marsden, it’s key to improving patient outcomes. We recently introduced an electronic trial master file system, ELECTRA, which was built here in partnership with a third party. Where we used to keep rooms full of folders of clinical trial data, this is now becoming electronic. We’re determined to go fully digital, and we’d like to expand our work with trial managers at the beginning of trials to build electronic data collection forms populated straight from EPR systems rather than relying on a manual transcription process.

The other big area that will grow in the future is the development of AI and machine learning. At the moment, I think most NHS trusts would face challenges in deploying such systems, as there are relatively few platforms around that allow you to do that easily. I think we will see a massive change over the next 10 years – not only in terms of trial design for these types of software development, but also with digital deployment. I believe we’ll see digital teams changing and adapting to work more closely with researchers so that these systems can be more easily tested, certified and adopted into clinical care.

What do you expect the benefits will be on your role?

Ceire: I’m already excited about the prospect of having linked healthcare data and, in time, patient generated data feeding into a platform whereby we can design really personalised algorithms, taking on board specific patient characteristics to inform treatment decisions. If I was hypothesising what my role would look like in 10 years’ time, we would be harnessing data science approaches and generating decision support tools that are embedded within our EPR systems. These tools should be fully validated and evaluated in terms of their outcomes on patients, registered as medical devices, and would be available for patients from initial diagnosis, informing treatment plans. In terms of making that happen, it comes back to an earlier point I made – this is a really multi-disciplinary field to work in. We need to make the most of collaborative working, incorporating many voices to achieve this future.

Lisa: I think in 10 years I will be still be working in the computing and research space, but with a greater emphasis on machine learning and large language models (LLMs). The national initiatives around NHS data access will make use of trusted research environments commonplace and use of emerging technologies will transform the way that data is provisioned and analysed, hopefully allowing much greater insights into the causes and treatments for cancer.

Ceire: There are a lot of national initiatives happening around informatics at the moment that are quite exciting and could change the landscape over the next five to 10 years – for example, national operations to link trusted research environments. So rather than working across a small number of NHS trusts, we could run algorithms nationally, in different health systems. That really improves the generalisability of the research that we are doing, because we can design, innovate and evaluate these tools on as wide a population as possible.

Many thanks to Ceire and Lisa for joining us.