Feasibility of mapping cross-country population coronavirus disease 2019 metrics in a federated design: learnings from a HealthData@EU Pilot use case

Abstract

The European Health Data Space aims to transform health data management across the EU, supporting both primary and secondary uses of health data while ensuring trust through General Data Protection Regulation compliance. As part of the HealthData@EU Pilot, this study investigates coronavirus disease 2019 (COVID-19) testing, vaccination, and hospitalization metrics across six European countries, with a focus on socioeconomic disparities and challenges in cross-border data access and standardization. This observational, retrospective cohort study used a federated analysis framework across Belgium, Croatia, Denmark, Finland, and France. Data were linked from administrative, social, health, and care records within each country's trusted research environment. A Common Data Model (CDM)-guided data harmonization, enabling nodes to perform independent analyses and share aggregated results. Key data processes (discovery, access, preparation, and analysis) were decentralized, with significant variability in data access procedures, security protocols, and available resources among nodes. The study revealed substantial differences in COVID-19 testing, vaccination, and hospitalization rates across countries. Denmark exhibited notably higher testing and infection rates. However, the study encountered key challenges: complex data access procedures, fragmented and incomplete socioeconomic data, and the need for extensive harmonization. Learnings from this pilot underscore the importance of streamlined, cross-country data access and standardization processes, which the European Health Data Space (EHDS) framework aims to address. The pilot demonstrates the feasibility of federated health data analysis across multiple countries while highlighting limitations in data access and interoperability. The EHDS framework offers a promising path to overcome these barriers, supporting efficient and standardized cross-border health research in the EU.

Figure 1.

Adapted methodological stepwise workflow for federated research. (This figure is an adaptation of the Figure 1 published in ‘PHIRI: Lessons for an extensive reuse of sensitive data in federated health research’ [10].) The research process began with the formulation of eligibility criteria and research questions, defining the study cohort, outcomes, and ensuring clear data requirements. A collaborative effort between the central node and the research nodes (Nodes, Table S1) translated these requirements into a CDM. The central node, which serves as the coordinating hub, was responsible for facilitating collaboration between research nodes, defining the research protocols, translating them into the CDM, and developing the analytical pipeline. The central node also prepared mockup data and scripts for data quality checks, analysis, and reporting, which were shared via an online repository for deployment across SPE at each research node. The research nodes, which represent individual entities (e.g. institutions, organizations, or countries), retained ownership of their data and operated autonomously within their SPEs. Each research node prepared and analysed its local data according to the CDM using the shared pipeline. They then submitted aggregated results, such as summary statistics or model updates, to the central node. Following data preparation and analysis at each node, the central node aggregated the results, performed meta-analysis, and disseminated the overall insights. Throughout the process, challenges and learnings were documented.

Figure 2.

COVID-19 PCR testing metrics across countries. The mean (▲) and median (●) values for the total number of PCR tests (top panel) and positive PCR tests (bottom panel) are shown for Denmark, Belgium, and Croatia. The denominator for all PCR tests includes all individuals who had at least one PCR test performed. For positive PCR tests, the denominator consists of all individuals who received at least one positive PCR test result. Horizontal lines represent the interquartile range for median values (●) and standard deviation (SD) for mean values (▲). The vertical dashed line at one is used as a reference point.

Figure 3.

COVID-19 testing, vaccination, and hospitalization metrics by country. This figure displays the percentage of the population in Denmark, Belgium, and Croatia that underwent PCR testing, tested positive for COVID-19, were fully vaccinated, were hospitalised due to COVID-19, and were hospitalised due to COVID-19 after vaccination. From left to right, the panels represent the percentage of individuals tested via PCR, the percentage with a positive PCR result, the percentage fully vaccinated, the percentage hospitalized due to COVID-19, and the percentage of fully vaccinated individuals hospitalized due to COVID-19.

Figure 4.

Proportion of COVID-19 vaccination uptake by socioeconomic indicators in Denmark and Belgium. The bar charts represent the percentage of vaccination uptake among various socioeconomic groups, categorised by education level, income level, migration background, and household type. Vaccination uptake percentages are shown on the left y-axis. Croatian data are not included due to missing socioeconomic data.