Modern Data Management
The term “FAIR data” has been floating around healthcare IT circles for some time. There are no social IT projects behind this, yet the association with the topic of “sustainability” applies. This is because FAIR stands for “Find-able, Accessible, Interoperable, Re-usable” and thus for a form of data management which is designed to generate a benefit beyond the place where the data was created. In this way, FAIR data is fully in step with current times in which healthcare institutions deal with topics such as electronic health card or patient file every day.
Prof. Martin Dugas, director of the Institute for Medical Informatics at the Westphalian Wilhelms University in Münster, explains in an interview whether the data model is suited for medicine and which requirements for FAIR data have to be met. Prof. Dugas, you have already been working for a long time with the principle of FAIR data, however with a view towards research data infrastructures.
Does this principle apply at all to care-based medicine?
Research, in fact, has a different view of the topic, especially of findability and access. Nowadays, when I perform research on a new therapy, I would like to be able to compare my results with those of earlier research. For this purpose, it is ideal if I can access all of the data and metadata. Outside of research, this kind of findability is not at all desirable because patient data must of course be protected. Healthcare institutions cannot comply with the principle at all when it comes to specific patient data.
Director of the Institute for Medical Informatics at the Westphalian Wilhelms University in Münster
Prof. Martin Dugas
What is crazy, however, is the fact that in medicine, structural metadata is not even accessible, that is, the information about which information should even be collected at all. To use a non-technical term, we are referring to the forms. My personal theory is that healthcare institutions, because of patient data protection, have begun keeping everything confidential across the board, so that it is not even public which data was collected at all.
Would it then help us to make these structural data FAIR?
Absolutely. I always say: patient data is confidential, blank forms are not. If, in an initial step, we at least standardized metadata, that is, the information about which data on patients are collected, this would lay the foundation for an exchange of data. This is why this information must be publicly available so that every hospital involved has access.
The point is that in today’s IT systems, data is either completely unstructured or not uniformly structured – especially across institutions. If every institution keeps its structuring confidential, every institution will then build its own different structure. The actual need, in particular to find medical information across institutions, cannot be met in this way. To do so would require data to be stored in a uniformly structured way beforehand. A real chicken-and-egg problem.
What can providers such as VISUS do to solve this dilemma and make data FAIR?
In all honesty, the industry with its solutions currently represents another problem. This is because the solutions are tailored very individually to the needs of the hospitals. In principle, each hospital has its own software. This likewise makes it difficult to send data from hospital A to hospital B.
A conflict arises from this: on the one hand, there is a desire to adapt the software as individually as possible to the customers’ needs. On the other hand, the various hospitals and players have to collaborate and recognize certain standards.
This conflict can be solved only by the customers if they say: we would like to make our data findable and reusable, because this is important for us. Because it is good for our patients. Because the law will require this of us in the near future. Because it complies with modern care.
However, there is still the notorious problem of terminologies. In SNOMED, there are 350,000 terms; by contrast, the language of Goethe has only 75,000 terms. This makes the dilemma very clear: in medicine, there are too many possibilities for describing the same issue in a different way. Consequently, the data from various institutions does not match.
Can standards such as DICOM or HL7 help make data FAIR?
A little bit. The DICOM standard helps develop compatible systems with which a data exchange is fundamentally possible. However, we will not achieve more than a minimum standardization in this way. Because when it comes to what can be seen on an image and what it means, there is no standardization. For this there is only free text which leaves the usual room for interpretation and which, moreover, cannot be reliably assessed. It is sad that, in this regard, we are still in the Stone Age.
Even the FHIR standard cannot currently solve the terminology problem. We have been looking further into this. Here as well, medical data is not yet defined precisely enough.
What advice do you have for users and the industry?
Everyone has to pull together. It does not help anyone if different players work on different solutions for a standardization of metadata – whether industry, the regulatory authorities, or the KBV (National Association of Statutory Health Insurance Physicians) within the scope of their mission from gematik (Association for Telematics Applications for the Electronic Health Card). Healthcare institutions can make a significant contribution by publishing their forms and thus give other institutions the chance to adapt structures. This could result in a meaningful discussion for national standardization. Anyone interested in this can find further information on the MDM portal, Europe’s largest collection of medical forms.