Access to health data for population health planning, basic research, and health industry use is both important and problematic.
Laws intended to improve access to personal data across national borders, where complexity and consequences of misinterpretation have paradoxically led to Data accumulation. In this vision, we propose federated networks as a solution to enable access to diverse data sets and to address known and emerging health problems. We first define the concept of federated networks in the health care context, present the value they can bring to multiple stakeholders, and discuss their creation, operation, and implementation. The challenges of federated networks in healthcare are highlighted, as well as their need and value from an independent orchestrator for secure, sustainable, and scalable implementation.
Introduction
Healthcare institutions generate and store health-related data from patients under their care. Data have great potential to improve diagnostic accuracy and treatment outcomes for common and rare diseases, but access or sharing of these data outside the host institution is often very limited.
Problem: Health data is accumulated
Several factors contribute to the accumulation of health data, including unclear ownership, insufficient consent to share or use data, and terms of use in data sharing/use agreements. Privacy concerns play a major role, where organizational interpretation of national privacy laws and the General Data Protection Regulation (GDPR) in Europe have led many clinicians and researchers to question the legality of sharing or accessing patient data for analysis and Analyzes are not reliable and unclear if the data is considered anonymous data in many cases.
In addition, the customized system structures and infrastructures, data formats, standards, and cybersecurity protocols that healthcare institutions typically work with lead to poor collaboration between different healthcare institutions. For example, silos prevent researchers from accessing the most up-to-date information or the most diverse and comprehensive data sets. They can slow down the development of new treatments and, therefore, key discoveries that could lead to much-needed treatments or cures.
Breaking down silos of health data has long been a recognized need, and access to health data for clinical decision-making and the development of new therapies remains challenging. The implications of this rely in particular on the knowledge generated from patients and are related to diagnostic discovery and customization of treatment strategies, both for current and future patients.
Solution: Federated networks
Federated networks (FN)s or data systems can be proposed as a solution to address health data silos and current barriers to data sharing. A definition of a federated data system is as follows:
An Federated network is a collection of decentralized, interconnected nodes that allow data to be queried or otherwise analyzed by other nodes. in the network without data exit from the node where it is located. As opposed to sharing, transferring, or merging data, Federated networks facilitate data access or data visitation, meaning that queries and algorithms can typically be sent to and applied to anonymized data.
FNs have the following common features:
- Each node is semi-autonomous and can make its own decisions about data access, but the nodes are still governed by a common standard agreed upon by all member nodes.
- FNs are also supported by a common infrastructure with harmonized standards and interoperability tools.
- Each member node needs local computing capabilities to perform query or processing locally. This is especially relevant when training AI and ML models through FNs, i.e. federated learning, which may require high-performance computing.
Federated health data networks (FHDNs) can facilitate access to sensitive health data and also have the potential to enable analysis of large cohorts across healthcare institutions, regional and national boundaries. In precision medicine and for the development of clinical decision support software, FHDNs have the potential to facilitate the exchange of algorithms and queries between nodes to be executed on a group data set and query results returned to the requesting node and Or the algorithms are modified. Federated learning is indeed particularly relevant in healthcare by allowing the training of a shared global algorithm on distributed sets of health-sensitive data that typically do not leave their home nodes. Examples of existing clinical applications of federated learning networks include the federated tumor segmentation network of 30 healthcare institutions working to improve tumor boundary detection, the AI4VBH (AI for Value-Based Healthcare) project, focusing on Improving patient pathways in cancer, coronary artery disease, stroke and covid-19 using federated learning in 12 NHS trusts (UK hospitals) and the Kaapana project, working through a shared imaging platform in 36 German university hospitals with a focus on analysis and analysis of radiological imaging and radiotherapy data.
Several initiatives have been launched based on this concept, which aims to create a federated system based on common standards that connect centralized and decentralized infrastructures to make data and services available. The proposed European Health Data Space (EHDS) aims to promote better exchange and access to health data for primary health care delivery, research and health policy.
But how exactly will these initiatives scale and interact?
Establishing an FHDN
As part of the “Breaking Barriers to Health Data” initiative, the World Economic Forum (WEF) states in a white paper that the creation of the FHDN requires three main components. Economics, governance and technology, which is divided between eight consecutive stages. Of course, only two of these steps are focused on technical requirements and standards, namely data structure and API deployment.
However, this should not be underestimated, as the usability of FHDNs is limited if the interoperability of data and metadata (eg, harmonization of data concepts, structures, or ontologies) is not addressed. The previous six steps focus on relationship building, i.e. trust and politics. Alignment of incentives and recognition of available resources help define the scope of the FHDN, and a specific FHDN governance model addresses common operational standards related to data inclusion criteria, intellectual property, and responsibilities. Establishing a governance model may require significant resources for technical, legal, and leadership alignment within and between organizations. It should ensure patient confidence, ethical use of sensitive data and trust between members before the FHDN consortium becomes operational.
The establishment of FHDNs has been recognized to provide economic return on investment in terms of diagnostic, clinical trials and personal benefits. Although the motivations for the economic value of these items differ in different countries, their social value, as measurable through the improvement of the quality of life, productivity and lifestyle of individual citizens, which potentially leads to a reduction of the health care burden in all countries is common
sources:
https://www.frontiersin.org/articles/10.3389/fpubh.2021.712569/full
