How data from clinical care can be used in practice | 5 Questions for… Dr Philipp Breitfeld, Head of the SMITH Junior Research Group BENEFIT

Research using clinical data is essential for the further development of medical technologies, treatments and diagnostics. The first step, however, is to create the right conditions for healthcare data to be made available for different applications. The BENEFIT Junior Research Group at the University Medical Center Hamburg-Eppendorf (UKE) has dedicated itself to this goal: it is working on the question of how data need to be prepared for different application scenarios so that they can be made available in high quality and in compliance with data protection regulations. The focus is on clinical research, but also on applications by external partners. Dr Philipp Breitfeld, a specialist in anaesthesiology at the UKE, took over as head of the junior research group in April 2023. He studied both computer science and medicine and is therefore familiar with the requirements of both disciplines. His motivation as a doctor is to work at the “frontier between medicine and information technology”, he says.

This year, funding for the BENEFIT junior research group comes to an end. In an interview, Dr Philipp Breitfeld explains what practical application scenarios the group has developed so far and what it hopes to achieve by the end of the funding period.

You have been head of the BENEFIT Junior Research Group at the University Medical Center Hamburg-Eppendorf since April 2023. As an anaesthetist, how did you come to this position?

As I had already worked in this interface area at the Department of Anaesthesia and collaborated on many projects with the Institute of Medical Informatics, I was already involved in some of the research questions. In anaesthesia, most of the processes are digital, because you often need to make quick decisions based on a good database. This information needs to be available centrally in one place so that we can act quickly in emergency situations. The resulting pool of data offers many opportunities for research questions and applications. At the moment, however, no one really has access to this data.
Thanks to my dual qualification in computer science and medicine, I am also familiar with medical processes. This has made it easier for me to familiarise myself with the topic and to see what scientists, administrative processes or external parties need in terms of data.

The junior research group started its work in September 2021. One of the goals of BENEFIT is to identify application scenarios for the use of medical data by external partners. What application scenarios has BENEFIT developed so far?

We are currently working intensively in two areas. Firstly, there are always questions about performance data in the hospital system to show how well a department is working. The problem is that it is difficult to collate data from different source systems. We want to automate that process. We are trying to take data from different sources and put it into one system so that we can draw conclusions, for example, about the performance of departments. We are also interested in how the patient is doing after treatment. To do this, we need other data from the clinical information system, such as billing data from controlling or even external information from cost units. In the past, these data could only be analysed separately. Now we are bringing them together to develop, for example, a decision dashboard for clinical management.

A second use case looks at how to extract, process and make available live patient data from monitoring systems. Other projects are currently being evaluated. They focus on working with small and medium-sized enterprises, with a particular focus on analysing the legal issues involved. We are looking for ways to make patient data sufficiently anonymous to be shared externally. For this purpose, the Institute for Medical Informatics at the UKE has developed the so-called “Data Hotel”. This is an application that can be used to provide pseudonymised patient data in a protected space for research purposes. Our task is to identify and curate the scientifically relevant data in the source systems and to integrate them for provision.

How do current developments in Artificial Intelligence affect the application of medical data?

Artificial Intelligence is a very important topic, especially in the area of applications for the development of predictive algorithms or decision support systems. However, we are not concerned with the clinical application of this data and AI algorithms, but with the provision of high quality data to train these algorithms.
However, we also work with algorithms in the provision of data. This primarily involves automatic annotation of data to contextualise it, explain it clinically and link it to domain knowledge from the clinic. We are currently developing concepts that can be used to annotate data from the clinical side. If the data is later used for a medical device, it must be quality assured. To do this, different doctors evaluate the data from the systems and try to map it to standards and terminologies to achieve the highest possible data quality. Artificial Intelligence supports this process.

What would you like to achieve with the young researchers by the end of the funding period* in 2024?

Firstly, I would like to get all our use cases to prototype status. This will show the importance of making data available first, so that other projects can build on it. Of course, there is a lot of research going on with AI right now. Research groups repeatedly encounter the same problem: they don’t have enough real-world data to train models and then actually use them. We are working hard to pave the way for a platform where researchers can reliably access data. We are making good progress. The data in the Data Hotel Use Case is largely accessible and will soon be transferred to our Data Integration Center. We have also decided how to annotate the data. Now it’s time to carry out the annotation in the remaining time.

The BENEFIT Junior Research Group harnesses the untapped potential of patient data for medical research by…

… developing viable prototypes that can be used directly in everyday clinical practice. This is how we enable research for clinical applications and further research ideas.

*Addendum from February 15, 2024: This refers to the first funding period 2021 – 2024. In February 2024, further funding was approved until 2026.