“Data must be accessible, accurate, and timely to support better outcomes”

“We opted for a digital health data platform architecture because it supports our vision of being truly data-centric and modular“, explained Patrik Georgii-Hemming, Chief Medical Information Officer at the Karolinska University Hospital, when we asked him about why they chose the platform architecture when designing the new data-driven strategy of Sweden’s largest university hospital. He also explained what role does openEHR play in this strategy, how they will leverage AI at the hospital, and what the future will look like. 

You say that data should be at the centre of everything we do and that every healthcare transformation must be centred around data. How do you implement this at Karolinska University Hospital?

At Karolinska, we believe that good data is the foundation of effective decision-making for clinicians, patients, and researchers alike. Our entire healthcare process revolves around data, from the initial patient consultation to treatment decisions and post-care analysis. We ensure that every step is documented, and every decision is data-driven. For us, the key is to make data accessible, accurate, and timely to support better outcomes. This focus on data drives our digitalisation efforts, making sure it is integrated and readily available across the care continuum. One of the principles we also want to follow with the data is the FAIR approach, which is well-known in research, meaning that the data should be Findable, Accessible, Interoperable, and Reusable. 

Can you tell us a bit more about why you chose health data platform architecture at Karolinska and how you see the benefits of DHP for clinicians?

We opted for a digital health data platform (DHP) architecture because it supports our vision of being truly data-centric and modular but also to “own the architecture”. Traditional systems often tie data to applications, which can limit flexibility and interoperability. With DHP, we prioritise data above the applications, meaning we can unlock valuable insights for clinicians, allowing them to make more informed decisions. With this architecture, the data flows seamlessly between different systems, from primary care to research, which ultimately enhances patient care, clinical decisions and clinical workflows. 

You have said several times that the application-centric approach is not suitable, as data is tied up in different applications. Instead, we should opt for a data-centric approach. How can we make sure this shift happens, and how are you doing it at Karolinska?

Shifting from an application-centric to a data-centric approach requires rethinking our IT systems and processes. At Karolinska, we focus on decoupling data from applications. This is where open standards like openEHR play a crucial role, allowing us to standardise how data is stored and shared, independent of specific applications. By focusing on data first, we make the data accessible and reusable across various clinical and research systems, helping clinicians access the right information when they need it – the FAIR approach I have mentioned before.

Why did you put openEHR as the standard for managing all the data in your strategy at Karolinska? Where do you see the benefits of openEHR?

We chose openEHR because it offers a flexible, vendor-neutral way to manage healthcare data. I have mentioned that for us it is critical to also “own the architecture” and the most critical part of it is the data. openEHR enables us to build tailored clinical applications while ensuring that all data is standardised and interoperable, or to be more precise, “intraoperable”, meaning that the “data exchange” comes from within the data itself. Once you solve the persistence layer, meaning it is open and vendor-neutral, you completely eliminate the need for data transformation and exchange at the application level. Also, the ability to separate data from applications means we can adapt to evolving clinical needs without being locked into proprietary systems, and can change the front-end applications without conflicting with the underlying architecture.  

Can you share with us some lessons learned in the process, and what would you advise similar hospital centres going in the direction of an open data approach instead of a monolithic system?

One of the key lessons we have learned is the importance of collaboration – both within the hospital and with external partners. Moving to a data-centric, open architecture requires a lot of alignment across departments. We have also realised that the transition requires careful planning to avoid disruptions in clinical workflows. For other hospitals, I would recommend starting with clear goals and focusing on current interoperability issues and legacy applications, the first use case that needs adapting or a completely new one that needs to be built. Think big, but start small – the implementation of one-for-all systems can last for years and will never fulfil their needs. Focus on what matters and deliver value fast, then build upon that with the other use cases. The platform approach allows you that. 

Obviously, Karolinska is one of the best hospitals in Europe and in the world, but still, do you have any role models when it comes to modernising and digitising clinical processes?

We certainly look to leading hospitals globally that have successfully implemented digital transformations, particularly those embracing data-driven approaches. In terms of best practices, we are inspired by institutions that have pioneered person-centered care through the use of technology and those that focus on making data usable for both clinical and research purposes. We continuously learn from others and adapt those lessons to fit
our specific context. 

The emergence of AI presents both opportunities and challenges for healthcare. As we can see in the annual report, Karolinska is taking an active role in the development of AI. Can you share with us some examples of how AI is used at Karolinska?

AI is an exciting area for us at
Karolinska. One example is in oncology, where we use AI to help assess patient readiness for the next round of cancer treatment. Patients input data through a mobile app, and the system integrates this with clinical data from previous treatments and lab results, allowing nurses to make informed decisions about whether to proceed with treatment or adjust the care plan. We also use AI in pathology reporting, improving efficiency and accuracy by automating certain processes and enabling quicker turnaround times for crucial clinical reports.