Borderless data, boundless hurdles: navigating the challenges of international federated sharing

18 Sep 2025

Federated health data sharing is no easy feat, but innovation and collaboration are guiding the path forward.

By Lindsay Smith, International Precision Child Health Partnership (IPCHiP)

This is not a success story. It is not a story of failure, either. It is a call for the global genomics community to share experiences openly.

In the world of international data sharing, there is a daunting chasm between promising innovations and their real-world application — a space where data sharing initiatives often falter, not for lack of vision, but due to the complex web of challenges that arise in implementation.  

These challenges extend beyond technical infrastructure limitations, like sufficient storage capabilities and heterogeneous data formats, to include legal, ethical, and cultural barriers that can slow or even halt progress. Strict national regulatory frameworks control the movement of patient data and restrict its transfer across borders. These restrictions are major obstacles that researchers and healthcare professionals must navigate when seeking access to international datasets. Furthermore, the datasets available have limited genetic variability within the populations represented in their makeup, which constrains the clinical relevance of genetic discoveries, as insufficient sample sizes reduce the statistical power needed to detect disease-associated variants within specific populations. 

Yet, these challenges are not insurmountable. With collaboration, adaptability, and a shared commitment to overcoming these hurdles, the potential to bridge this divide remains firmly within reach. 

The International Precision Child Health Partnership (IPCHiP) is a multinational collaborative bringing together four leading paediatric institutes: Boston Children’s Hospital (United States); University College London Great Ormond Street Institute for Child Health and Great Ormond Street Hospital (United Kingdom); the Murdoch Children’s Research Institute with The Royal Children’s Hospital and the University of Melbourne Department of Paediatrics (Melbourne Children’s Campus) (Australia); and The Hospital for Sick Children (SickKids) (Canada). 

IPCHiP aims to improve access to genomic diagnoses and novel treatments in order to optimise outcomes for children with rare diseases. The partnership facilitates collaborative scientific investigation of rare diseases, creates innovative diagnostic and therapeutic solutions, and develops methods for acquiring, sharing, and analysing genomic and phenotypic data across institutions. As a new Driver Project for the Global Alliance for Genomics and Health (GA4GH), IPCHiP is able to share experiences with a network of institutions with similar goals, while bringing a clinical perspective to standards development to shape and adopt global best practices for precision child health. 

At the core of IPCHiP is the design, institutional support, and launch of multi-site cohort studies. Our flagship study, Gene-STEPS (Shortening Time of Evaluation in Paediatric Epilepsy Services), is demonstrating the feasibility of implementing rapid clinical genome sequencing on a disease-specific cohort in inpatient and outpatient settings to improve diagnostic yield. While each IPCHiP institute has generated robust and comprehensive datasets through Gene-STEPS, this data remains siloed, and there is no easy mechanism for investigators to explore that data. Links between diagnoses are only made if one investigator directly asks another, and it is unclear what connections, including both positive and negative results, are missed through this “phone a friend” approach. We need a better and more efficient way to pool and explore the data generated from each institution. 

As we considered how to approach data sharing, federation — a data-sharing model in which requesting parties bring their analysis software to the data — seemed like a logical solution to navigate differing institutional cultures, jurisdictional policy, and regulation, while accommodating the sensitive nature of health data. Despite feeling prepared to navigate setting up a federated system, we soon encountered many unexpected challenges. In keeping with our aim to share experiences openly, I want to highlight a sampling of issues we faced during this project.

On-premise vs cloud considerations

Institutions had different requirements for hosting data sharing software in a cloud vs an on-premise (on-prem) environment. It is possible to accommodate these differing requirements, but it would have been helpful to have the ability to easily compare notes. On-prem environments require greater isolation features than cloud, and while a cloud approach may be able to mitigate isolation concerns, it can be expensive to maintain even with a minimal set up. 

Data security concerns

In order to run workflows in another institution’s environment, we need open ports to the internet and significant infrastructure updates to support isolation or security management features. Open ports are not always feasible in an institution hosting health data, and these updates can be cost prohibitive. 

If the cloud supporting data sharing software is set up in an environment isolated from the main data store, data needs to move to be accessible. Is this really a federative approach, if we have to move the data? 

Querying connected datasets

User friendly interfaces and no-code solutions are needed to enable use by clinicians and researchers who are not trained in query methods. No-code solutions have limitations, and if a researcher wishes to query beyond simple filters, technical knowledge is required. 

Data access

Once the datasets of interest are located, administration of data access requests was an unanticipated complication for some sites. For sites without a central data access committee or system, it was unclear how to triage requests appropriately. 

Workflow languages

Preferences for particular workflow languages vary across bioinformaticians and sites.  

Workflow cost management and computing resources

If a researcher triggers a workflow in another site’s environment, it is unclear who pays for the workflow costs and how. To support cost management, there also needs to be agreement on what the workflow is intended to do at each site. 

Results management

The ability to combine datasets and queries from across our sites is key to letting go of our “phone a friend” process. With four sites, and the potential for multiple variables of interest, it is unclear if data is stored or harmonised in a way that allows for complex queries.  

The challenges listed here are just a snapshot of what we have faced during our pilot project. The independent nature of each participating institution and the diversity of our existing systems really underscore the complexity of aligning infrastructure, policies, and expectations. Success may take a different path at each site. Solutions will come as the result of trust, transparent communication, and a willingness to adapt. 

My hope is that by sharing experiences openly (both the good and the bad), engaging in collaborative pilot projects, and growing from the lessons learned so far, we are gradually building bridges over the chasm between innovation to application, one brick or beam at a time. I would call on other global data sharing collaborations to continue engaging in a collective dialogue about lessons learned, challenges faced, and opportunities to drive progress together. The potential rewards — better variant interpretation, improved analytical and diagnostic power, and the realisation of precision health — are well worth the journey.