Data-driven healthcare requires a federative data model

Prof. dr. Jan Hazelzet
Published on


Whilst the adoption of digital solutions appears to increase, prompted in part by the current Covid-19 pandemic, the question is how we can utilise the added value of digital healthcare to the fullest extent. The future of good and affordable care starts with a good exchange of digital information in the sector, also called interoperability. But we soon run into a big challenge in healthcare. Because we are still speaking the wrong language: literally and figuratively. The solution lies in a federative data model.

We are still speaking the wrong language for a data-driven healthcare

What I mean by figuratively is that the medical world and IT world are still acting as separate worlds. Too little is being developed in conjunction, which leads to a mismatch between that which is delivered and that which is in demand. In late March, for example, there was a conflict about a new software system that had to map the IC beds capacity. The Landelijk Coördinatiecentrum Patiënten Spreiding (National Coordination Centre for Patient Dispersion) eagerly insisted on using this system, which was supposed to aid in a better dispersion of corona patients. At the same time, several medical IT specialists advised against working with the system, because an empty bed doesn’t automatically mean a patient can be admitted[1]. Meanwhile, the software is still not being used in all hospitals.

In a literal sense, my experience is that the current practice of healthcare documentation is still not in order. Many existing information systems in healthcare, such as electronic health records (EHRs), still have their own data model. In other words; an own way in which data is structured in the information system. There is no adequate, uniform documentation and many of the outdated systems in healthcare are also not equipped for this. Exchanging and combining data from multiple sources requires substantial and complex processing of the separate datasets, so these can tie into each other. Healthcare information is thus not patient-oriented, but system-driven. This was one of the reasons to start the national program ‘registration at the source’ in 2014. In it, all healthcare information is recorded once and in a uniform way. The transfer of information thus becomes more uniform and complete, and the quality of that transfer increases. Reusing uniformly recorded data simplifies both scientific research and monitoring the quality of healthcare. Additionally, uniform and one-time recording of healthcare information also ensures less high administrative and registry loads.

The future is federative

I believe in a standardised set of base data of every Dutch person, such as the formulated Basisgegevensset Zorg (Base Dataset Healthcare). At the same time, much more information on patients is collected, often saved in fragments in different systems that are located in different places. New technologies are made available at a rapid pace. Smarter systems, more data and sensory technology allow healthcare professionals to collect and use more information. But to really benefit from the enormous growth of data in healthcare, this data needs to be sharable. ‘The solution lies in working federatively.’ It was one of the recommendations of the Van der Zande committee[2].

In a federative data model, healthcare information is stored in a decentralised manner (for example in an EHR) with the healthcare professional providing treatment. If another healthcare professional, or the patient, needs information, this information isn’t stored an additional time – instead, authorised access is granted to the decentralised, stored data. Compounded or derivative data can be generated upon request. This way, it is guaranteed that the data complies with all (legal) prerequisites when it is obtained. To ensure that data stays with the source, it is necessary to create an authorisation protocol. This shall detail which data of which patient can be analysed, from which patient records, after the patient’s approval, and to which extent  . The advantages of a federative data model are, amongst other things, that there are no copies, the healthcare professional always has access to the most recent and accurate data, administrative burdens are lighter, and there is less risk of information loss and data leaks. In addition, the patient can rest assured that their data is not passed on to others. The OMOP Common Data Model pursues such a model, and is currently being shaped further by the European Health Data & Evidence Network (EHDEN)[3].

For other researchers, a federative model is interesting. The Personal Health Train concept builds upon this development[4]. In short: data is no longer brought to the analysis, instead the analysis is brought to the data. It is, as it were, a kind of train that passes by: the data stays at the source and its owner presents it to the passing train for analysis. In using separate sources of data, the train passes through several stations in order to analyse the data anonymously. Every time, the data stays with the source – the train only brings conclusions back to the analyst. We are witnessing a surge in large IT multinationals that have obtained a powerful knowledge-position in a relatively short amount of time, by focussing on data and disruptive business models that revolve around knowledge and innovation. Their strength lies in the immense scale of data available to them. However, their data is generally not open. By combining powers federatively, and collecting and sharing data with each other, researchers can also find such large datasets at their disposal.

Acceleration is vital

Of course, over the past few years, several steps have been taken towards standardising data exchange. Think not only of the ‘Registratie man de bron’ programme, but of the Informatieberaad Zorg (Healthcare Information Council) and incentive programmes for standard digital data exchange such as VIPP and MedMij. It is estimated that from 2021 onwards, regulation will be implemented in order to make standard digital data exchange mandatory for every insurance provider. However, everything is slow to get off the ground, takes a lot of time, and is still too non-committal. The dependency on the often dominant and closed-off data systems of IT suppliers is also still too high. This endangers the safety of patients because the data, relevant or not, is made available too late, or only partially. This requires top-down guidance, and the government should play a part. Requirements should be mandatory in order to exchange data in a uniform way. ‘If not willingly, then compulsorily.’ We can learn a lot from other countries when it comes to this. If we want to innovate healthcare, we shouldn’t work within old, big data systems.

The future of good and affordable care starts with a good exchange of digital information in the sector. The challenge is in finding the right language. If that succeeds, it allows the sector to work faster, more efficiently, with more focus on the quality and the patient. Filled with hope, I watch developments in, for example, personal healthcare environments (persoonlijke gezondheidsomgevingen, PGOs), in which the patient has control over their data. However, as long as not yet everyone has access to a PGO, we will continue to look for a way to have the most up-to-date data for value-driven healthcare at our disposal, in an accessible and clear way. I expect the Netherlands will end up with such a hybrid format. The Dutch Hospital Data (DHD) already collects, manages, and edits hospital and UMC data as a central organisation. Of course we must still overcome some hurdles together, but modern technology could really boost the quality and affordability of healthcare.


  1. H. Modderkolk. Conflict over nieuw systeem dat bedden telt brengt minister De Jonge in lastige positie. Volkskrant. 30 maart 2020
  2. A. Van der Zande et al. Advies commissie governance van kwaliteitsregistraties. 29 maart 2019
  3. European Health Data Evidence Network – [Internet]. [cited 2021 Nov 30]. Available from:
  4. Personal Health Train – Dutch Techcentre for Life Sciences [Internet]. [cited 2021 Nov 30]. Available from:

Prof. dr. Jan Hazelzet

Prof. dr. Jan Hazelzet has a lot of clinical experience as a paediatric intensivist and university professor of paediatric medicine. He gradually switched to data and quality, first as CMIO at the Erasmus MC, later as a professor of ‘Quality and outcomes of healthcare’. As the clinical head of the Value Based Health Care programme at the Erasmus MC, he advocates for a shift towards a more patient-oriented healthcare. He is also active in several national (NFU) and international consortia (EUHA).