The evolution of health data ecosystems: An international survey

Summary

This paper reports the findings of an international survey of health data ecosystems (HDEs) in 12 countries plus the H3 Africa project using live, structured interviews with senior project team members under the auspices of Canada’s All for One Precision Health Initiative. We note the high level of interest in HDEs around the world, as well as in Canada, despite the financial, jurisdictional, and other barriers that continue to hold back widespread data sharing. We present results detailing operational profiles for each of the 13 participants, including whether their healthcare systems are centralized (national) or decentralized (regional), project start date, funding, information technology (IT) infrastructure, and the extent to which participants have implemented a data-sharing mandate. We find no evidence to confirm common assumptions about features conferring an advantage on HDE development, such as early launch date or top-down government mandate. We also find no evidence of a reference model to explain what makes any HDE effective, valuable, or successful and conclude, on the basis of our interviews, that the diversity that makes each of these projects unique may undermine collective actions like data sharing. While participants provided useful cautions about pitfalls they encountered, more research on these issues is required, and we anticipate that advanced assessment tools like the maturity level model (MLM) developed by the European Union (EU) may help countries understand what stage of the HDE development process they have reached and what strategies will be most effective for them in later stages.

We report on an international survey of health data ecosystems (HDEs) based on live interviews with project leaders in a dozen countries. The remarkable diversity of HDEs works against collective actions like international data sharing while spurring much-needed research on how to measure the comparative success of individual HDEs.

Introduction

We present the results of an international survey of health data ecosystems (HDEs) based on our original report as released in May 2024 and available in the supplemental information accompanying this paper. The study was conceived and carried out under the auspices of Canada’s All for One (AFO) Precision Health Initiative (web resources), launched in 2018 by Genome Canada (web resources). The AFO is itself a growing pan-Canadian HDE that resembles several of the initiatives examined in this paper. The term HDE is very inclusive and covers a range of interpretations, reflecting the particular subject area of interest, such as clinical and research activities, data policy and governance, or enabling technologies. For the purposes of this paper, we define an HDE as a dynamic network of individuals, organizations, technologies, and policies that allow researchers to generate, collect, store, analyze, and share genomic and health-related data in order to improve healthcare delivery, research, and public health outcomes. The study covered 12 countries: Australia, Denmark, England, Finland, France, Japan, New Zealand, Saudi Arabia, Singapore, Sweden, Switzerland, and the United States (web resources), in addition to the transnational Human Heredity and Health in Africa, or H3A initiative, itself comprising about 30 countries (web resources). The purpose of the study was to hear firsthand what our colleagues in other countries have learned in developing their HDE projects and to explore new ideas of potential benefit to the Canadian genomics community.

This study comes at a time of greatly increased interest in HDEs around the world. HDE teams have increasingly turned their attention to one crucial long-term goal: the integration of genomics into routine clinical care. The efforts to achieve that goal are taking a number of different paths. As an example of one approach, England’s 100K Genomes Project conducted a large pilot study in 2021 to measure the increase in diagnostic yield provided by whole-genome sequencing (WGS). The investigators found a significant increase across a range of rare diseases, further evidence of the benefits genomic medicine brings to healthcare¹ and a finding confirmed in other studies.² A second type of activity involves developing robust standards for measuring the progress of integration as a proxy for success, the most advanced current example being the European Union’s (EU) maturity level model (MLM), developed under the auspices of the 1+Million Genomes (1+MG) and Beyond 1+Million Genomes (B1MG) initiatives.³ A third approach involves funding for the creation of genomics-enabled learning health systems (gLHSs), exemplified by the 2024 NIH funding of the gLHS Network, whose central purpose is to improve how genomic information is integrated into patient care (web resources). Finally, Stark et al. provide a roadmap and “call to action” whose practical approach is reflected in its 12 steps to scale up genomic data sharing.⁴

The 13 participants in this study represent only a portion of the larger group of countries reported in recent years to be developing HDEs. In a systematic review published in March 2021, Kovanda et al. estimated the number of then-current national projects worldwide to be 41.⁵ The greater number of projects identified by Kovanda, compared to our selection, reflects basic differences in research goals. Kovanda’s systematic review was a quantitative exercise that aimed to be comprehensive. Our survey was designed as a qualitative exercise aimed at an in-depth understanding of the HDE field based on open-ended questions discussed in real-time interviews. The findings from this study will benefit teams embarking on early phases of HDE development, and the real-world lessons offered will appeal to a broad range of other interested parties, including policymakers, technologists, health economists, hospital administrators, patient advocates, and researchers working in other areas of the biomedical sciences.

Methods

In the fall of 2023, our study team consulted the AFO steering committee and leadership of the Global Alliance for Genomics and Health (GA4GH)⁶ to identify suitable participants for a proposed survey. Recommendations were made on the basis of involvement in GA4GH activities, with consideration given to national structural factors, especially the nature of each country’s public healthcare system, which provides the medical, financial, and social contexts in which HDEs are developed. We also solicited suggestions for participants from multi-nation regions, such as East Asia and the Nordic countries. Once we identified potential participants, our canvassing drew positive responses from 12 countries: Australia, Denmark, England, Finland, France, Japan, New Zealand, Saudi Arabia, Singapore, Sweden, Switzerland, and the United States. In order to draw contrasts with these large-scale national efforts, we added the transnational H3Africa initiative (itself comprising about 30 countries), which depends on pooled resources and external funding to achieve its goals, which are centered on advancing the quality of both healthcare and biomedical research across sub-Saharan Africa.

Our methodology was anchored by the use of a structured, 27-item questionnaire in live, hour-long interviews with senior project personnel (see supplemental methods: participant acknowledgments and HDE participant questionnaire), supported by analysis of the website material published by each project (web resources) and references to the literature where appropriate. The questionnaire, which was circulated in draft form for quality control, was designed to elicit information on a range of factors that affect or typify HDEs, organized into four sections: project features, infrastructure and budget, user authorization and roles, and challenges and lessons (see supplemental methods: HDE participant questionnaire). In the post-interview phase, verbatim transcripts were prepared from the recorded interviews, and excerpts were captured in a series of spreadsheets, including a cross-national comparison matrix based on the topics addressed in the questionnaire, along with status tracking and interviewer comments. It was agreed with the interviewees that they would not be quoted for attribution in any of our public documents. The individual country profiles in the original report were all vetted, corrected, and updated by each country’s project leaders in at least two rounds, and their edits were all accounted for in our final report (see supplemental methods: country profiles).

Results

Brief profiles of each of the 13 study projects are provided below, followed by selected comparative data to highlight some of their similarities and differences (see Table 1). We divided the 13 projects into two sub-groups based on their public healthcare systems, categorized as either centralized (national) or decentralized (regional), to see what differences, if any, the broader system makes to the kind of HDE being developed in each country. Note that we use the centralized-decentralized healthcare distinction as one of degree: no bright line separates the two categories. Each of the two healthcare categories accounts for half of the 12 countries, i.e., six centralized and six decentralized. Because H3Africa comprises over 30 participating countries, they were reported collectively as being “mostly regional.” The regional category comprises Australia, Denmark, Finland, Sweden, Switzerland, and the United States, although in all six cases, the central government retains some degree of policy-, regulatory-, or standards-setting authority, in addition to a role in financing. The national or centralized healthcare category comprises England, France, Japan, New Zealand, Saudi Arabia, and Singapore. Eleven of the 12 countries provide their entire population coverage for core services on the single-payer system, with policy exclusions for certain expenses, like those for dentistry or prescription drugs (Switzerland’s system does not offer services free at the point of care). The exception to the single-payer model is the United States, which has a mixed, non-universal system that offers public support for senior citizens, some low-income citizens, and a few other special populations. All funding amounts herein are approximate, most having been converted from other currencies into United States dollars, without adjustments for purchasing power parity.

Table 1.

Selected results from the All for One HDE survey

Country	Healthcare system	Project launch date	Private sector funding	IT infrastructure	Data sharing mandate
Australia	regional	2016	yes	in planning	no: proof of concept
Denmark	regional	2019	yes	HPC private cloud	yes: national strategy
England	national	2013	yes	AWS cloud + local HPC	NHS mandate
Finland	regional	2017	yes	Google cloud storage	yes: national mandate
France	national	2016	no	HPC private cloud	no: preliminary work
Japan	national	2018	no	commercial cloud	yes
New Zealand	national	2017	no	HPC	no
Saudi Arabia	national	2018	no	private cloud	yes
Singapore	national	2017	no	AWS via government cloud	no
Sweden	regional	2018	no	HPC private cloud	no: preliminary work
Switzerland	regional	2017	yes	HPC private cloud	no
United States	regional	2018	yes	Google cloud + AnVIL	yes: NIH GREGoR
H3 Africa	mostly regional	2012	yes	HPC private cloud	yes: NIH mandate
Results	regional = 7 national = 6	–	no = 6 yes = 7	HPC = 7 cloud = 4 mixed = 1	yes = 6

Columns include the survey results for participating country, healthcare system, project launch date, private sector funding, IT infrastructure, and data-sharing mandate. dash indicated no data; H3 Africa, Human Heredity and Health in Africa; NHS, National Health Service (UK); GREGoR, Genomics Research to Elucidate the Genetics of Rare Diseases (https://gregorconsortium.org/); AnVIL, Genomic Data Science Analysis, Visualization, and Informatics Lab-Space (https://anvilproject.org/overview); HPC, high-performance computing; AWS, Amazon Web Services.

Countries with decentralized healthcare

Under Australia’s regional system, the federal government shares healthcare responsibilities with its six states and two of its territorial governments in a federated model that resembles Canada’s. While Australia’s system is based on the single-payer model, many of its services are provided through private-sector resources. Australian Genomics (AG) began operation in 2016 with a mission to promote genomic research and its translation into clinical care on a 10-year budget of $30 million. In 2021, the government tasked AG with developing recommendations for a national approach to genomic information management (NAGIM). NAGIM would use a federated model combining regional healthcare structures with a fully interoperable national ecosystem that supports genomic information management in both clinical and research settings. Until the NAGIM strategy is implemented, Australia’s research and clinical genomic data remain siloed (web resources).

Denmark is divided into five regional governments, each of which has a great deal of autonomy in healthcare delivery while drawing on the central government for 80% of its funding and municipalities for the other 20%. The Danish National Genome Center (DNGC) was founded in 2019 to develop personalized medicine through a collaboration between the country’s regions and the entire healthcare system, as well as research institutions and patient associations. The DNGC has implemented this federated model with support from a legal mandate that requires the regions to send all patient-consented genomes to the central DNGC database, where they are made available to clinicians and researchers across the country. In addition to government funding of about $5 million a year for operating costs, the DNGC received a grant in 2018 equivalent to $140 million, to be used over 4.5 years, from Denmark’s Novo Nordisk Foundation, a non-governmental enterprise foundation that supports research in biomedicine, biotechnology, and human health (web resources).

Finland’s current system has taken decentralization one step further by moving healthcare responsibilities to specially elected well-being services councils. Finland Genomics (FinnGen), which was founded in 2017, has grown in size and reputation thanks to a number of unusual features. One, it is a public-private partnership funded in part by Business Finland and 13 pharmaceutical industry partners, which together provide a 6-year budget of $100 million, a level of non-government financial support that is unusual in this study cohort, with some exceptions, like the DNGC. Two, FinnGen collects both genomic and digital healthcare data, creating combined datasets of much greater utility to researchers than either type of data on their own. Three, FinnGen can draw on the wealth of information provided by 500,000 Finnish biobank participants, whose genetic data holdings are consented for research, with results being returned to the biobanks. FinnGen works closely with the FinData agency, which maintains health data and grants permits for the secondary use of social and healthcare data to improve data protection for individuals (web resources).

Sweden’s healthcare is managed at all three levels of government—national, regional, and municipal—with the country’s 21 county councils having wide autonomy in providing services, as well as being responsible, along with municipalities, for most healthcare funding. The Genomic Medicine Sweden (GMS) program (not to be confused with England’s Genomic Medicine Service discussed below) is a nationwide endeavor to establish the necessary infrastructure for genomics research and personalized medicine. It was founded in 2018 and is funded by the Swedish Innovation Agency, Vinnova, a specialized government agency whose mandate extends to other areas of the Swedish economy besides healthcare. The GMS also receives co-funding from the seven regions with teaching hospitals and the seven universities with medical faculties (web resources).

Switzerland’s decentralized healthcare system divides responsibilities between the federal government and 26 cantonal governments. Switzerland has benefited from the wide consent provisions in the 2014 Swiss Human Research Act, which has encouraged many Swiss citizens to make their routine data and samples available for research—almost 700,000 at the five university hospitals alone. Switzerland launched the Swiss Personalized Health Network (SPHN) in 2017, dedicated to the development of research infrastructure in collaboration with a wide range of academic, healthcare, and other partners in order to make healthcare data collected by the five university hospitals more suited for research purposes. The SPHN combines bottom-up support for research projects with top-down infrastructure development on an 8-year budget of $135 million. The Swiss State Secretariat for Education, Research and Innovation granted the SPHN Data Coordination Center another $22 million for the next funding phase (2025–2028) to extend their activities and act as a national coordination and competence center for all research endeavors involving Swiss healthcare data (web resources).

Healthcare in the United States is administered at the state level through a complex blend of public funding, private insurance, and private-sector delivery. While the United States does not have a comprehensive national mandate to develop genomic or precision medicine, the federal government funds a number of major genomic projects through the National Institutes of Health (NIH). One of its most prominent projects is the All of Us research program, whose aim is to enhance healthcare through research. Its original target was one million participants, higher than most other national initiatives. Another of the program’s unique features is that data collection extends far beyond genome sequencing and includes blood, urine, and saliva samples; physical measurements; electronic health records; survey information about participants’ health histories, family medical histories, lifestyles, and communities; and wearable data. All of Us has 10-year congressional funding of $1.5 billion, renewable annually (web resources).

Countries with centralized healthcare

Genomics England (GEL) was created by the British government in 2013 to manage the 100,000 Genomes Project in partnership with the National Health Service (NHS). GEL passed its original target for genomes sequenced in 2018. GEL operates in a national but devolved healthcare system, as the NHS is run largely independently by each of England, Scotland, Wales, and Northern Ireland. GEL has launched several other programs, both clinical and research oriented. They support the NHS Genomic Medicine Service by providing WGS diagnostics for patients with rare diseases and cancer, with a view to enhancing genomic healthcare and streamlining patient management. Core aspects of these new programs have been mandated as part of NHS strategy, although they are not codified in legislation. They include mandates for both sequencing and data sharing, covering both research and clinical purposes, as part of the NHS strategy for genomic data collection and clinical sharing. GEL’s operational budget is $640 million provided over a period of 10 years (web resources).

France has a highly centralized, universal healthcare system and coordinates its genomic medicine activities under the auspices of the government-funded 2025 France Genomic Medicine Initiative (PFMG2025), established in 2016. Its primary goal is to incorporate genomic medicine into the clinical care pathway, with a view to revolutionizing both healthcare and biomedical science within the French system. Clinical implementation is being developed in several areas, including the validation of clinical standards for access to genomic diagnosis based on pre-indications for rare diseases and cancers and the development of patient-related resources aimed at ethical issues, education, and consent processes. Although PFMG2025 is a collaboration with France’s national hospitals and two major sequencing labs, work on a comprehensive data-sharing mandate was still “preliminary” at the time of the interviews. Nevertheless, France collaborates with several other national projects, including GEL, GMS, and the DNGC, in addition to taking part in the EU’s 1+MG initiative (web resources).

Japan operates a national healthcare system that gives responsibility to the country’s 47 regions for producing budgets responsive to regional health concerns. Japan’s central genomic institution is GEM Japan (for Genome Medical Alliance Japan), whose objective is to facilitate genomic and health data sharing among research and medical lab communities, both domestically and internationally. GEM Japan has established a nationwide alliance for the promotion of genomic medicine comprising over 50 pharmaceutical and biotech companies, along with research institutions and universities, including the National Cancer Center, the University of Tokyo, and the RIKEN Center for Integrative Medical Sciences. Japan still lacks an official data-sharing mandate but has made provisions for a sequencing mandate and regular use of data sharing for both research and clinical purposes. (We were unable to conduct interviews with representatives of GEM Japan: web resources.)

While New Zealand originally developed its healthcare system on a decentralized basis, the national government has introduced reforms that centralize all funding, policymaking, and delivery of care. Genomics Aotearoa (“Aotearoa” is Māori for New Zealand) was funded in 2017 on a 7-year grant of $24 million intended to ensure a role for New Zealand in international genomic research across all life sciences. It develops resources for improving the diagnosis and treatment of genetic diseases and for promoting the use of diagnostic exome and genome sequencing. It has also created a dedicated precision medicine research platform (Rakeiora) and the Aotearoa Genomic Data Repository to provide a secure place for the research community to store and share genomic data within a Māori values context. Genomics Aotearoa does not yet operate under a broad HDE mandate as it embarks on a second 5-year funding cycle with no budgetary provision in place for new infrastructure (web resources).

Saudi Arabia operates a healthcare system that, while centralized under the broad policy framework established by the Ministry of Health, gives the country’s 20 regional health directorates considerable independence. In 2018, the Saudi Genome Program (SGP) began as part of the Saudi Vision 2030 program, dedicated to transforming healthcare in the country. The main objectives of the SGP are to create a database to document the first genetic map of the Saudi people, develop personalized medicine, and reduce the cost of healthcare, drawing on state-of-the-art genome sequencing, bioinformatics, and validation techniques. The SGP has grown to become one of the largest genome initiatives in the Middle East, operating under a national data-sharing mandate for which provisions of Europe’s General Data Protection Regulation (GDPR) have been adopted. The first phase of strategy development, SGP 2.0, was launched in 2022 (web resources).

Singapore, like New Zealand, has a centralized healthcare system without a national data mandate, although processes and safeguards are in place to enable data sharing. Singapore has embarked on the long-term development of an agency to foster the growth of genomic and precision medicine: the National Precision Medicine Program (NPMP), which has three main purposes. The first is to establish Southeast Asia’s most deeply phenotyped cohort to support understanding of Singaporean and Asian health. The second is to promote a data-based healthcare system to identify groups at higher risk of disease, optimize healthcare delivery, and develop more accurate diagnoses and targeted treatments. The third is to capitalize on NPMP’s datasets by bolstering both medical stakeholders and adjacent industries, as well as supporting local companies and attracting foreign investment. The NPMP is structured in three distinct phases: the provision of 10,000 genomes in phase 1 (SG10K Health), which are expanded to 100,000 genomes in phase 2 (SG100K), and phase 3, which will implement a whole-of-government and whole-of-ecosystem model, planned for launch in 2025–2026 (web resources).

H3Africa is not an HDE operating under a single national governance system as in the previous 12 cases. It was included to provide an illustration of how genomics research has developed on multinational lines in the Global South using external funding. The central objective of H3Africa, which has about 30 participating countries, is the development of sustainable, collaborative research for genomics dedicated to improvements in population health across Africa. H3A is a pool of individual academic research projects that function together in a consortium whose coordinating center is in Cape Town. It was funded from 2011 to 2021 by the NIH, the Wellcome Trust, the African Academy of Sciences, and the Science for Africa Foundation and received over $170 million from NIH/the Wellcome Trust to fund infrastructure development. Around 50 H3A studies are hosted in over 30 countries and span a range of topics and methodologies, including population-based genomic investigations into common non-communicable conditions like heart and renal diseases, as well as communicable diseases such as tuberculosis. The common purpose is to increase biomedical understanding of African peoples, long under-represented in terms of both disease genes and the genomics of the continent’s many and varied ethnolinguistic groups. H3A has played an active role in international efforts to share data and reach broad agreement on recognized standards. It does so through facilities such as its pan-African bioinformatics network, the H3ABioNet, which aggregates data from H3A research projects under the European Genome-Phenome Archive (EGA) system, with a view to harmonizing the datasets collected (web resources).

Distinguishing features of HDE initiatives

Launch dates

As shown in Table 1, timeline is a unifying factor: 11 of the 13 projects began in the same 3-year period from 2016 to 2019: Australia (2016), Denmark (2019), Finland (2017), France (2016), Japan (2018), New Zealand (2017), Saudi Arabia (2018), Singapore (2017), Sweden (2018), Switzerland (2017), and the United States (All of Us: 2018; Canada’s AFO Precision Health Initiative was launched in 2018). This clumping across such a short time span suggests that many scientists were paying close attention to the same international trends in the biomedical field, especially the widespread uptake of WGS beginning in the mid-2010s. The two exceptions are H3Africa, launched in 2012, and England’s GEL, the persistent outlier that began in 2013 and has been described as the “pioneering flagship project” in genomic medicine (web resources).⁷

Funding

Although funding is a headline feature of large public undertakings, assessing the “value” realized from a given level of budget expenditure is a fraught exercise, as is drawing conclusions from budget comparisons. One financial pattern that does emerge here concerns the role of the private sector acting in partnership with public funders. Table 1 (private sector funding) shows that a private sector role is confined exclusively to the regional projects (all but Sweden): not a single one of the national sub-groups has any significant private sector involvement. By far the broadest range of financial structures is found in the United States system. The federal government has made major public funding commitments to support HDE initiatives like All of Us (web resources), while some ventures in genomic medicine operate commercially, such as Geisinger Health System’s MyCode, a large-scale initiative under which Geisinger patients may choose to have their genome sequenced as part of their standard of care.

IT infrastructure

The 13 projects have spread their information technology (IT) choices across two major options: local (private cloud or high-performance computing [HPC]) and external (government or commercial cloud). Seven of the 13 have opted for HPC or private cloud, four use commercial or government cloud resources provided by Google or Amazon Web Services (AWS), one (England) uses a combination of AWS cloud for computation and local HPC for data storage, and one (Australia) was still in the planning stages. These choices bear no apparent relation to healthcare system type, e.g., of the four using external cloud, two operate in national systems (Japan and Singapore) and two in regional systems (Finland and the United States). These technologies have ushered in a new era of genomic research, but, as was true of WGS, they also create problems of their own. In addition to the sheer volume of data being generated, keeping up with changing software functionality and computing paradigms requires substantial effort, expense, and expertise.

Data sharing and health system integration

In the questions we asked about data sharing, we wanted to know whether each project operated under a sanctioned data-sharing mandate. The projects reporting an official mandate numbered seven (Denmark, England, Finland, Japan, Saudi Arabia, the United States, and H3A), with the other six reporting no mandate (Australia, France, New Zealand, Singapore, Sweden, and Switzerland). One caution here is that the terminology and the practical details of sharing may leave room for misconceptions. Take the key term “mandate” and note how it applies to GEL and the NIH-supported All of Us research program. Both have a publicly funded mandate to promote genomic medicine on a wide scale, but their roles in the healthcare system are dramatically different. Whereas NIH programs like All of Us operate outside the United States healthcare system, GEL and its offshoots are an integral part of England’s NHS.

Discussion

As mentioned in the introduction, the most notable finding from our survey is that HDEs are highly idiosyncratic and exhibit far more differences than similarities from one project to another. What makes this finding even more remarkable is the degree to which these differences persist in the face of two countervailing tendencies. First, the projects all share the same general goals, particularly the integration of genomics into routine clinical care as a core feature of precision health. Second, in the course of their work, HDE teams continue to expend significant effort and resources to overcome these differences in order to maximize opportunities for data sharing and collaboration, the overarching rationale for undertaking these projects in the first place.

In analyzing the survey results, we looked for patterns that might explain the range of differences we identified. Take, first of all, the launch dates noted above. Does an early launch confer what might be interpreted as an advantage in project development? Here, we construe “advantage” as achieving a data-sharing mandate, although there are other ways in which this could be interpreted. Two examples suggest there is no apparent advantage. The earliest project in the centralized group was the France Genomic Medicine Initiative, launched in 2016, the same year as AG, in the decentralized group. However, 8 years later, the French project still lacked a data-sharing mandate, and AG had not completed the development of its proposed NAGIM. On the other hand, in 2013, GEL became the first of all to launch (except H3A, in 2012) and included a national data sharing mandate, then went on to beat its own timetable for sequencing 100,000 genomes.

Funding for HDEs takes many different forms, particularly those operating on a mixed public-private model, which includes all the regional projects except Sweden’s. In some cases, private sector contributions come from commercial partners, such as for Finland’s FinnGen initiative, which is funded in part by its Business Finland and pharmaceutical partners, whereas Denmark’s DNGC received its non-government funding from the Novo Nordisk Foundation, the world’s wealthiest enterprise foundation. One reaction common to all our interviewees was worrying about whether their funding was going to hold up over the long term. Financial disruption can extend even to well-established and prominent projects like All of Us. In 2016, it was allocated $1.5 billion over 10 years, but it is subject to the congressional appropriations process, and in April 2024, the program budget was reduced by 34% (web resources). One question arising from the distribution of private sector funding is why it is all confined to regional systems. One possibility is that projects having distinct components that are separated geographically and enjoy some autonomy may be in a better position to find opportunities for growth that can start small and more easily accommodate specialized private sector interests.

There is a more general question to address about HDEs and the two types of healthcare systems, namely whether a decentralized system is inherently less welcoming than a centralized one toward establishing a national mandate for data sharing. That is the argument advanced by a group of Swedish researchers in connection with the growth of genomic projects in Sweden’s decentralized healthcare system.⁷ Their challenge was to nurture a national initiative in the absence of a government-sanctioned, top-down mandate, especially as Swedish healthcare and research resources are distributed across multiple government entities. Fioretos ascribes Sweden’s success to a “bottom-up approach” in which a number of small scientific groups joined forces to build what eventually became a nationally distributed infrastructure under the banner of GMS, funded by the innovation agency Vinnova.

The idea that HDEs have a more difficult time growing in regional systems may be intuitively appealing but does not accord with how HDEs are actually faring. In the decentralized group, Denmark and Finland each operate their HDE under a national data-sharing mandate, while the United States has implemented its own form of data-sharing mandate through its NIH programs. On the other hand, Australia and Sweden are, by their own account, still at the proof-of-concept stage. The centralized countries demonstrate the other side of this proposition: four of them, France, Japan, New Zealand, and Singapore, were still functioning without an official data-sharing mandate as of their respective interview dates. In the Swedish example of bottom-up growth, the development of its national HDE was slowed not just by the multiple regional entities involved in healthcare but also by the existing legal framework set up expressly to restrict data sharing. Moreover, the case of Sweden’s grassroots initiatives and their evolution into a national project is not unusual. Indeed, Canada may be on a path like Sweden’s, whereby the grassroots growth of Canada’s foundational genomic initiatives will eventually benefit from a top-down, national health data strategy, as recommended by Canada’s expert advisory group, the authors of the 2022 report on a pan-Canadian health data strategy (web resources).

Top-down and bottom-up growth look different again for multinational projects like the H3Africa initiative and the EU’s collective pursuit of genomic medicine. The developing nations of Africa overcame a lack of resources and scientific talent by pooling their efforts and creating a transnational, collaborative top layer that allows them to run infrastructure, host research studies, and develop homegrown talent. Through its 1+MG and B1+MG initiatives, first launched in 2018 and running until 2027, the EU has also created a “top” structure to link nearly 30 countries in joint endeavors, but it is built on the already well-resourced scientific establishments in the participating countries. These collective EU initiatives have enabled the launch of the Genome of Europe (web resources), as well as the creation of the MLM, a tool for countries to self-assess the maturity of genomic medicine practices in their healthcare systems.³ In other words, the bottom-up concept covers a number of different starting points and objectives, and given the universal aim of maximizing data sharing, it seems likely that bottom-up strategies in partnership with top-down approaches will dominate in the coming years.

The diversity we found among the initiatives we examined, affecting almost every aspect of their strategy and operations, extends to the terminology we use to describe HDEs, as noted earlier in connection with the term “mandate.” Even when the context is more concrete, like IT infrastructure, Kovanda points to anomalies in how this resource is described, suggesting that this reflects “cultural conceptual differences in what is considered as infrastructure.”⁵ A similar observation suggests that there is social evidence that healthcare users and providers, when asked about details of personalized genomic medicine and personal health data spaces, draw on differing interpretations of fundamental concepts like data ownership and control.⁸ A study undertaken in Canada in 2022 looked at how four countries—Australia, France, the United Kingdom, and the United States—used data sharing as a public health tool to combat the COVID-19 pandemic.⁹ Despite their similar legal and policy structures, the study concluded that the four countries performed very differently in how they shared their data. It suggested several factors that may explain these differences: stricter approval formalities, the involvement of multiple consortia, the absence of centralized data platforms, and cultural factors, such as the degree of familiarity with open science. Taken together, these findings lend support to the idea that a diversity of assumptions and working methods, while potentially beneficial in certain ways, tend to make data sharing and other collaborative scientific activities more difficult, a conclusion confirmed in the interviews.

The survey gave interviewees an opportunity to offer candid views about their personal experiences working in HDE teams in terms of both warnings and advice on solutions. Some expressed their surprise at how lengthy and difficult—in one case, “chaotic”—they found the planning and early implementation stages and wished they had had more time and patience for important choices. One said their team had greatly underestimated the complexities of genomics and needed three years to understand “the landscape” and get agreements drafted—legal negotiations being cited as a major source of delay. Some found difficulties in implementing data sharing with other labs because of differences in infrastructure, staff skills, and established workflows, despite the many efforts being directed at overcoming barriers of this very kind. Interviewees made clear they experienced disagreements over the functionality and cost of new technologies, as well as over disruptions like moving to cloud computing, and that few objective criteria were available to resolve such disagreements. As one put it, “Even if everybody wants to work together and they’re all using the same tools, implementation still takes a lot of work.”

Much of the advice offered was about how to manage the less tangible but no less important working conditions that determine motivation, teamwork, resilience, and job satisfaction. Some effective measures can be planned and managed: well-designed awareness and outreach campaigns, education and training programs to retain highly skilled staff, and partnerships that enhance long-term interoperability. Other human resource goals are less amenable to control and include collaborative working environments, an inclusive approach to decision-making, creating opportunities for meaningful engagement with collaborators, and promoting tolerance for conflicting viewpoints in early project stages. This aspect of project management should extend to winning strong buy-in across the stakeholder community, which can be helped by taking care to listen to stakeholders from the outset and including them in information sessions and decision-making as often as possible.

Survey research of the kind we used has limitations as well as benefits. First, our 13 participants constitute only a small sample, and no selection of countries would be representative of all those operating an HDE project. Our lineup was also limited in number by self-selection: some of the projects we approached declined our invitation or did not respond in time. Second, the project personnel made available to us were mostly senior scientists, who could reasonably be expected to provide an accurate and authoritative account of their team’s work. Other personnel may have responded differently or may have been more familiar with certain aspects of their country’s developments. Nevertheless, certain questions, such as those concerning challenges and lessons, may have elicited different responses if asked of different interviewees. Third, the survey questionnaire was open ended and could have produced different responses if the wording or order of questions was changed. If the countries, interviewees, or questionnaire had been different or more inclusive, many of the survey responses would undoubtedly have differed in their details. Nevertheless, short of a sweeping change in coverage (e.g., surveying HDE projects in developing countries only), it is unlikely the results would have changed in a material way.

In conclusion, our survey findings and other research have shown that, while there are clearly lessons to be learned from active national HDE projects, there is no evidence of a comprehensive blueprint or model to explain what makes any given HDE effective, valuable, or successful. The main goal pursued by researchers in this space is to expand data sharing and international collaboration on a broad scale while bridging the differences that make the pursuit of this goal so challenging. HDE designers can only benefit from gaining a deeper understanding of how other teams are handling shared problems. To do so requires that we have good empirical methods for measuring progress, beginning with an agreed-upon definition of the factors that make progress possible. Under its 1+MG and B1MG initiatives, the EU has developed its MLM, which is designed to measure the level of maturity of national genomic medicine practices using a matrix based on a 5-point rating scale that is applied to 49 indicators grouped into eight domains, with results presented in both qualitative and quantitative form. It is intended for individual countries to carry out self-assessments, some of which have been published to encourage meaningful comparisons between countries.^10,11 National HDE teams outside the EU project, including Canada, may benefit from adopting the MLM for their own purposes. At the same time, we see a continuing role for the survey research behind this paper, making firsthand contact with HDE teams to convey their experiences on the job, along with more formal feedback about strategy and operations. Further research of both kinds is needed to advance our knowledge of HDEs and genomic medicine, in addition to acting as a channel for advocacy by promoting more and better communication among those working in the field or hoping to do so in the future.

Data and code availability

The published article includes all datasets generated or analyzed during this study.

Acknowledgments

This work was carried out as part of the AFO Precision Health Partnership. It was funded by the government of Canada through Genome Canada and the Regional Genome Centres (Ontario Genomics, Genome Alberta, Genome Quebec, Genome Atlantic, Genome Prairie, and Genome British Columbia). K.M.B. is supported by a CIHR Foundation Grant (FDN-154279) and a Tier 1 Canada Research Chair in Rare Disease Precision Health. We extend our sincere thanks to all the participating HDE teams for making time to help us understand and assess their projects. (See the supplemental information, participant acknowledgments, for details about the interviewees and their affiliations.)

Author contributions

Conceptualization, J.P.L.-E., E.M.P., S.S., and K.M.B.; data curation, S.S., J.P.L.-E., and E.M.P.; formal analysis, J.P.L.-E., S.S., and E.M.P.; funding acquisition and project administration, K.M.B., E.M.P. C.R.M., M.Z., V.F., J.L.M., D.B., J.P.L.-E., and F.B.; writing – original draft, J.P.L.-E., S.S., and E.M.P.; writing – review & editing, J.P.L.-E., E.M.P., S.S., T.B., M.-J.B., A.R., L.C., J.G., D.C., N.B., S.S.W., S.R., B.B., K.C., S.O., V.W., P.S., A.L., F.N., I.A., N.J.M., A.G., P.G., L.D.S., C.R.M., M.Z., V.F., J.L.M., D.B., F.B., and K.M.B.

Declaration of interests

The authors have no conflicts of interest to declare.

Web resources

• 2025 France Genomic Medicine Initiative (PFMG2025), https://pfmg2025.aviesan.fr/en

• All for One (AFO), https://genomecanada.ca/challenge-areas/all-for-one

• All of Us, NIH, USA, https://allofus.nih.gov, https://allofus.nih.gov/about/faq

• Australian Genomics, https://www.australiangenomics.org.au

• Danish National Genome Center (DNGC), https://www.eng.ngc.dk

• Finland Genomics (FinnGen), https://www.finngen.fi/en

• GEM Japan, https://www.amed.go.jp/en/aboutus/collaboration/ga4gh_gem_japan.html

• Genome Canada, https://genomecanada.ca

• Genome of Europe, https://b1mg-project.eu/1mg/genome-europe

• Genomic Medicine Sweden (GMS), https://genomicmedicine.se/en

• Genomics Aotearoa, https://www.genomics-aotearoa.org.nz

• Genomics England (GEL), https://www.genomicsengland.co.uk

• Government of Canada, https://www.canada.ca/en/public-health/corporate/mandate/about-agency/external-advisory-bodies/list/pan-canadian-health-data-strategy-reports-summaries/expert-advisory-group-report-03-toward-world-class-health-data-system.html

• H3Africa, Human Heredity and Health in Africa, https://h3africa.org

• NIH, https://www.nih.gov/news-events/news-releases/nih-awards-27m-establish-new-network-genomics-enabled-learning-health-systems

• National Precision Medicine Program (NPMP), https://medicine.nus.edu.sg/trp/precision-medicine

• Novo Nordisk Foundation, https://novonordiskfonden.dk/en/

• Saudi Genome Program (SGP), https://www.vision2030.gov.sa/en/explore/projects/the-saudi-genome-program

• Swedish Innovation Agency, Vinnova, https://www.vinnova.se/en

• Swiss Personal Health Network (SPHN), https://sphn.ch