Collaboration is key to good and relevant register-based research

Projects that combine clinical and methodological expertise progress more quickly and achieve higher scientific quality.

Traditional register-based research is evolving rapidly. This is according to Johan Askling, Professor of Rheumatology at Karolinska Institutet, who has been working with register-based studies for over 20 years. As a clinical researcher, he sees a need for increasingly detailed data in order to address clinically and medically relevant research questions. This is particularly true in precision medicine, where diagnosis, treatment and follow-up are based on the unique characteristics of each individual. This places even greater demands on detailed data and, somewhat paradoxically, on even larger study populations.

“National health data registers, such as those maintained by the National Board of Health and Welfare or Statistics Sweden, are internationally competitive data sources that have been developed over many years. In my experience, they provide a strong foundation, for example for research on treatment effects and drug safety. At the same tim, medical advances are progressing rapidly, and today my own and many other areas of medical research require additional and even more detailed data than are available in the national health data registers,“ he says.

Divergent interpretations of legislation and non-interoperable information systems

Data extraction from quality registers and from various healthcare data repositories can provide access to detailed data on disease and its progression, which are needed in clinical epidemiological research. However, the challenges associated with handling such detailed data sources are considerable.

“Medical record data are stored in different systems across the country that do not interoperate. In addition, there is no common standard for how variables are defined. Extracting and then combining data on administered drugs from different medical record systems has proven to be quite complicated, largely due to the lack om a common underlying standard,“ he says.

The absence of national electronic health record registers means that researchers must contact each healthcare provider separately. This can result in long lead times and considerable manual effort.

“Since each local or regional healthcare authority has a say, there may be different interpretations of current legislation, to the frustration of all parties. It is not easy for even the most experienced medical researcher to navigate these waters, which are really more about logistics and legal matters than research,” he says.

Where a project requires more comprehensive and detailed datasets, privacy concerns also become more pressing.

“Considerable time and effort must be devoted to how the data should be handled, who should have access to them and for what purpose. This raises the question of whether it is reasonable for individual researchers themselves to build and maintain their own secure data management and analysis environments, or whether there is already a need for collaboration with organisations that can provide such environments.“

Collaboration across multiple areas of expertise

According to Johan Askling, it is generally advisable to draw on the expertise of others with greater experience in the field of register-based research. Researchers undertaking a comprehensive register-based study face a range of issues, such as application procedures, what data are needed to address the research question, and how data should be stored in a technically and legally appropriate manner.

His advice is therefore to seek out well-established research environments and establish collaborations with researchers or teams experienced in register studies.

“The need for collaboration between different areas of expertise has never been greater than it is now! Comprehensive and detailed register-based studies require expertise across multiple domains. I believe that the most effective approach is for researchers and clinicians involved in such projects to collaborate with data managers and experts in register-based research. Otherwise, we risk many researchers becoming occasional users of highly complex systems.“

Nordic collaboration – rewarding but complex

Johan Askling has extensive experience collaborating with Nordic research teams, which use data from other Nordic countries to adress similar questions. It is an advantage that Nordic countries often have comparable data from similar data sources and healthcare systems, as well as broad similarities in their populations.

“In rheumatology research, for example when we want to assess the safety and effectiveness of new drugs, the data is insufficient if we focus only on Sweden. In such cases, it is advisable to combine data across the Nordic countries. This provides access to larger datasets and allows us to divide larger groups into subgroups, for example based on specific patient characteristics,’ he says.

Register management and cross-border data sharing present many opportunities. At the same time, they also involve significant challenges, as each country has its own decision-making processes, legal systems and timelines. Johan Askling's experience is that the need for agreements can quickly become a project in its own right.

“I recently participated in a collaboration between four Nordic countries and two companies specialising in biomarker analysis, where we needed to draw up more than 20 agreements or data-sharing arrangements, including to ensure compliance with the GDPR. Each such agreement had to be reviewed by a legal expert at each participating organisation, which was a rather frustrating process. If we were to undertake this project again today, we would have placed greater emphasis on developing some form of overarching agreement, linked to trust in each party’s ability to handle and protect data, and devoted more time and effort to bringing all legal experts together in the same room at the same time,” he says.

In collaborative studies involving many parties, he observes a tendency to prioritise avoiding errors over achieving the intended outcomes.

“One way to address this tension between objetives and processes is to involve all experts, including legal experts, early in the project in order to establish a shared understanding of the purpose of the collaboration or project,” he says.

Multiple opportunities to statistical analysis in Nordic collaboration

Johan Askling notes that there are often subtle differences in data between Nordic countries. This means that a thorough comparison of raw data is required before statistical analysis can begin. He describes this as an iterative process: generating descriptive data, comparing across countries, making adjustments, and repeating the comparison. When it comes to clinical research questions and data, it is also important to ensure that clinical competence is available in each country, according to Askling.

“I wouldn’t feel confident analysing Danish data myself, but would leave that to our Danish colleagues,” he says.

The statistical analysis can be done in different ways, he points out.

“The simplest approach is to analyse the data locally in each country following harmonisation and in accordance with a common statistical analysis plan, and then combine – ‘pool’ – the results. This usually works well, but can be statistically inefficient,” he says.

A second approach is to send the data to a party that performs all analyses on the pooled data. However, it is important to remember that the Swedish Ethical Review Act applies to research conducted in Sweden, rather than specifically to ‘Swedish’ data.

“This important point is easy to overlook, but it means that a Swedish researcher who gains access to Nordic data for joint analysis needs to obtain ethical approval in order to analyse these data in Sweden,” he says.

Johan Askling also observes a growing trend towards the national “lock-in” of data within secure analysis environments from which data cannot be exported.

“This is understandable to some extent, but if more than one country does this, it means that collaboration based on pooled data may become impossible,” he says.

A third approach is so-called federated analysis. This makes it possible to conduct statistical analyses across multiple data sources with identically structured individual data, without having to physically merge the data.

"This means that advanced program code analyses data that remain stored locally, but as if they were combined into a single dataset. Federated analyses are developing rapidly and could reduce the burden of physically transferring data. Unfortunately, in my experience, considerable effort is required to ensure that the systems function effectively, for example due to firewalls that restrict access or analysis sessions that are time- limited,” he says.

Register-based research in development

In medical research, there is currently a strong focus on the EU, but in register-based research, the scientific potential is in many cases greater at the Nordic level, according to Johan Askling. He emphasises that the Nordic potential could be realised through a few relatively straightforward changes – at least in terms of their substance – to policies governing ethics approval and data storage.

“I think this is an obvious area for a joint Nordic initiative. One example is that an ethics application approved in Denmark would also be valid in Sweden. Common templates and harmonized processes for accessing health data registers across the Nordic countries would also be beneficial, or that all Nordic countries agree to make register data available within secure Nordic analysis environments for joint analysis,“ he says.

Looking ahead, Johan Askling anticipates and advocates greater consolidation within register-based research, for example through Sweden establishing a national healthcare data repository. He hopes that the Nordic countries will implement a number of policy and legislative reformes, and points out that the new European Health Data Space (EHDS) regulation could help facilitate this development.

Johan Askling also believes that other analysis methods, based on machine learning and AI, will transform the field of register-based research in the coming years.

However, in the present context, he returns to his earlier conclusion that collaboration is a key factor in successfully conducting register-based research.

“Good collaboration, both across areas of expertise and between the Nordic countries, yields clear benefits in terms of both efficiency and quality,” he says.

Text: Ulrika Ernström