Federated Data and Sepsis Management in the EHDS

Original Title: The next frontier in sepsis: connected ICU data for real-world clinical decision making

Journal: Intensive care medicine

DOI: 10.1007/s00134-025-08284-3

Overview

Sepsis is a major healthcare challenge, causing one in five deaths globally and affecting approximately 49 million individuals every year. In Europe, hospital treatment costs range from 16,000 euros in France to over 27,000 euros in Greece, while follow-up care for survivors in Germany costs about 6.8 billion euros annually. Despite these high stakes, clinical data remains fragmented across local silos, hindering the development of effective decision-support tools. The European Health Data Space (EHDS) proposes a federated infrastructure to connect intensive care units across borders. This framework allows for the secondary use of routinely collected health data while ensuring that patient records remain under the control of individual hospitals. By shifting from isolated registries to an interconnected ecosystem, clinicians can perform standardized analyses locally and share only anonymized results.

Novelty

This review introduces a multi-country ecosystem that moves beyond the single-center limitations of traditional databases like the Medical Information Mart for Intensive Care. While existing databases reflect the specific demographics and therapeutic cultures of a single institution, the EHDS framework addresses the significant variability in European critical care, where bed availability ranges from 4 per 100,000 people in Portugal to 30 in Germany. The novelty lies in the integration of federated learning with synthetic data generation within a unified regulatory framework. Federated systems allow for joint analysis without the transfer of sensitive patient-level information, effectively bypassing the logistical and ethical barriers of large-scale data retrieval. Furthermore, the use of common data models, such as the Observational Medical Outcomes Partnership, enables the harmonization of heterogeneous clinical terminologies and documentation practices across different languages and healthcare systems.

Potential Clinical / Research Applications

The implementation of these connected infrastructures has direct implications for both research and bedside practice. In research, federated queries can identify rare sepsis phenotypes or subgroups that are too small to be studied within a single hospital. For clinical education, federated grand rounds could allow healthcare professionals to discuss anonymized, aggregated trends in critical care management, situating individual cases within broader population benchmarks. Quality improvement activities would benefit from real-time dashboards that compare local performance against European-wide standards without compromising privacy. Furthermore, synthetic datasets can be used to train healthcare staff in decision-making under uncertainty by simulating rare but instructive scenarios, such as refractory septic shock. These applications collectively support a learning health system where clinical insights generated in the intensive care unit inform subsequent research, policy evaluation, and personalized treatment strategies across the continent.