1、Sharing Sensitive Health Data in a Federated Data Consortium Model An Eight-Step Guide I N S I G H T R E P O R T J U LY 2 0 2 0 Contents Foreword Introduction Step 1 Establish and sustain trust Step 2 Jointly determine the problem for a federated approach Step 3 Align on incentives and organizationa

2、l capacity Step 4 Identify resourcing team leadership and funding Step 5 Identify institutional differences or gaps in policy Step 6 Create a consortium governance model Step 7 Structure the data Step 8 Deploy the API technology Conclusion Appendix Acknowledgements Endnotes 3 4 6 9 11 13 15 17 20 21

3、 22 23 23 24 2020 World Economic Forum. All rights reserved. No part of this publication may be reproduced or transmitted in any form or by any means, including photocopying and recording, or by any information storage and retrieval system. Sharing Sensitive Health Data in a Federated Data Consortiu

4、m Model2 Foreword At the World Economic Forum, we think of data as the oxygen that fuels the fire of the Fourth Industrial Revolution. It is readily available and necessary but, if used improperly, it can generate dangerous and unwelcome results. Concerns over how to protect valuable data, especiall

5、y sensitive, personal data, are at the core of many countries and institutions data policies. We see a complex and dynamic data policy landscape evolving around health data in particular; it is becoming more and more complicated to share data to the extent desired to advance research, innovation and

6、 patient outcomes. The need to rapidly provide access to health data, while protecting patient privacy and data security, has never been more urgent in the fight against the COVID-19 pandemic. This paper is part of the Forums work to create actionable resources for policy-makers, healthcare professi

7、onals and leaders of the Fourth Industrial Revolution to navigate complex and sensitive health data policies globally. The Forum is testing a federated approach where data sets are accessed remotely without movement of data from its secure location of origin as a practical way to access the disparat

8、e genomic and health data sets needed to accelerate the diagnosis of rare disease in patients in four countries. Federated data systems are not new per se, but they are starting to be used more frequently as a solution to accessing multiplying, disparate data repositories in a multinational and mult

9、i-jurisdictional world. Being able to quickly and securely access disparate data sets accelerates the ability to gather insights and inform care decisions for precision medicine approach, which uses data to drive more personalized and tailored diagnosis and treatment of disease in patients. Offering

10、 practical advice on how to build a federated data consortium is only possible with the partners in the Breaking Barriers to Health Data project, key to the Forums precision medicine portfolio of projects. Four genomics institutions in Canada, Australia, the United Kingdom and the United States have

11、 worked tirelessly to have the difficult conversations and build the governance model that inform this eight-step guide. We applaud their leadership. This guide also forms a critical input to the Forums Data for Common Purpose Initiative focused on new models of data governance in the Fourth Industr

12、ial Revolution. A recently released Roadmap for CrossBorder Data Flows: Future-Proofing Readiness and Cooperation in the New Data Economy expressly recommends that governments should recognize Federated Data Learning as a valid means of cross-border data (insight) sharing and should not be blocked b

13、y legislation. Proactive efforts will be needed to motivate government officials, business leaders and civil society members to establish real-world pilots and to enable continuous and active experimentation with federated data systems, particularly in situations where they are most valuable. Access

14、ing sensitive health data at scale will advance research, innovation and patient outcomes Genya Dana Head of Healthcare Transformation, Shaping the Future of Health and Healthcare, World Economic Forum Arnaud Bernaert Head, Shaping the Future of Health and Healthcare, World Economic Forum Sharing Se

15、nsitive Health Data in a Federated Data Consortium Model An Eight-Step Guide July 2020 Sharing Sensitive Health Data in a Federated Data Consortium Model3 Introduction Accessing global health data through federated consortiums will reveal disease causes and cures Genomic data represents our shared D

16、NA and can be broken down into a machine-readable format in a process called genetic sequencing. During genetic sequencing, DNA is broken down into its four chemical bases (adenine, guanine, cytosine and thymine) for analysis. Each human DNA consists of about 3 billion bases.2 Every human being has

17、such DNA represented by billions of bases, but it is only possible to understand more about our shared DNA and, more importantly, how our DNA impacts or even predicts our health by mode of comparison using large volumes of DNA. This is because more than 99% of bases are the same in all people, makin

18、g any differentiation more difficult to discern in smaller data sets. In contrast to a base, a gene is the unit by which an individuals one-of-a-kind combinations of DNA bases are inherited. Genes can vary in size from a few hundred DNA bases to more than 2 million bases per gene.3 Both in the sheer

19、 scale of genomic data and in the complex health data policy regulatory landscape, aggregating such data to improve patient outcomes is complicated. The human genome (your genome is the sum of the DNA in your body or the sum of your genetic data) represents roughly 100 gigabytes (GB) of data, which

20、is equivalent to the size of about 100,000 digital photos. In 2011, our sequencing capacity hit 13 quadrillion bases, which was the equivalent of two miles of stacks of DVDs in data storage (which were used for storage in this era before data storage moved to the cloud). By 2018, however, the human

21、genome (roughly 3 billion bases) fit on a single DVD disk rather than on the hundreds of discs spanning two miles in 2011.4 Storing the human genome is progressively getting easier, smaller in size and cheaper. Comparing genomic data to Silicon Valleys Moores Law, which states that computers double

22、in speed but half in size every 18 months, genomic data is outpacing Moores Law by a factor of four in storage size.5 Why genomic data?BOX 1 In the current era of the Fourth Industrial Revolution, data is our most valuable resource.1 The five leading companies of our time Alphabet, Amazon, Alibaba,

23、Facebook and Microsoft rely on data to fuel their successful enterprises. Data is also a resource in the healthcare ecosystem that can improve the standards, quality and outcomes of healthcare and healthcare delivery for patients worldwide. But just how are health ecosystems using data? As volumes o

24、f healthcare data increase, genomic data and other types of sensitive health data provide a treasure trove of information on how to diagnose, treat and generally manage the most complex and destructive diseases but only if we can look at data across the global population. Genomic data is a particula

25、rly valuable type of health data because it represents the hereditary material in humans (and almost all organisms) called deoxyribonucleic acid (DNA), which stores the “master code” dictating how our bodies operate. More than 99% of genetic code is the same in all people, making it difficult to pic

26、k out “glitches” or specific small differences in the genetic code useful for research, diagnosis and treatment of disease without ways to comb through large amounts of data. Aggregating large genomic data sets in ways that researchers and clinicians can use to improve patient outcomes is complicate

27、d, in part due to the flood of genomic data from national and institutional genetic sequencing efforts. The human genome (your genome is the sum of the DNA in your body or the sum of your genetic data) represents roughly 100,000 digital photos. It now takes approximately a day to sequence most of th

28、e genome of one person, and several hundred dollars, compared to 13 years and $1 billion in 2003. Countries and institutions are sequencing hundreds of thousands of people. In 2018, the UK announced the completion of 100,000 sequences from National Health Service patients. Accessing all of this data

29、, however, remains a challenge due to a complex landscape of data protection laws and health data privacy regulations. Sharing Sensitive Health Data in a Federated Data Consortium Model4 The World Economic Forums Global Precision Medicine Council, in its May 2020 Precision Medicine Vision Statement,

30、 cited the gap in data- sharing and interoperability as key to preventing the wider adoption of a more personalized approach to healthcare.6 Precision medicine depends on the availability of health data in the aggregate. For genomic data in particular, the costs of storage and analysis are usually m

31、ore expensive than the lab costs of sequencing. The cost to store, process and analyse the data can be justified in the global patient interest if the data can be used beyond its initial diagnostic capacity for a single patient.7 Accessing and using sensitive health data and genomic information to i

32、ts full potential requires care and creativity, with strong governance protocols to guide this process. To tackle the challenge of governance of cross- border access to health data, the World Economic Forum led the Breaking Barriers to Health Data project, from July 2018 to July 2020. The project te

33、sted how a distributed federated data system could be set up and run sustainably across countries with clear governance optimizing for operational efficiency, patient privacy and data security. Federated data systems are a promising way to enable access to health data, including genomic data, that m

34、ust remain inside a country or institution because of their sensitivity. Although examples of federating health and genomic data sets are growing, how to practically create the federated data system with a group of institutions was not as clear.8 Allowing access to data sets is not particularly diff

35、icult technically, but there are larger challenges in how to form the necessary relationships between institutions that enable trust and transparency, and sustained, predictable operations in a consortium model. In close partnership with Australia (the Australian Genomics Health Alliance), Canada (G

36、enomics4RD), the United Kingdom (Genomics England) and the United States (Intermountain Healthcare), the Forum created and led a multistakeholder community that supported these institutions through the journey of determining how to maximize the benefits and minimize risks of federating genomic data

37、to diagnose rare diseases.9 In order to federate data, a consortium of institutions must be formed. As outlined in Figure 1, this eight-step guide distils the learnings from the Breaking Barriers to Health Data projects work to set up a federated data consortium for the purposes of diagnosing rare d

38、isease using genomic data from a global, distributed data set. Other institutions are also encouraged to adapt this federated data consortium model for additional use cases. Before creating such a data consortium leveraging sensitive health data, it is crucial to carefully plan for such a consortium

39、 and meticulously consider how to effectively craft and implement clear governance structures. Global federated data consortiums provide a tremendous opportunity to improve patient outcomes and healthcare delivery pathways but also require robust security, continually improving policy to provide saf

40、eguards against bad actors, data breaches or other types of preventable risk. Federated Data Consortium 1 Step 1: Establish Trust 2 Step 2: Defi ne Problem8 Step 8: Deploy the Technology 3 Step 3: Align Incentives7 Step 7: Structure the Data 4 Step 4: Identify Resources 5 Step 5: Identify Institutio

41、nal Gaps 6 Step 6: Create Governance Model Eight steps to follow to build a federated data consortiumFIGURE 1 Federated data systems are a promising way to enable access to health data, including genomic data, that must remain inside a country or institution because of their sensitivity. Sharing Sen

42、sitive Health Data in a Federated Data Consortium Model5 Establish and sustain trust 1 Generating trust is more important than ever and requires the right partners, thorough relationship building and support from leadership teams Step The first step, and the singular component that appears to make o

43、r break a federated data consortium, is establishing trust with identified prospective partners entering a data consortium. Establishing trust between partners is also the most time-consuming component in establishing a successful data consortium. The creators of a new data framework called Trust :

44、Data Consortium which include the Massachusetts Institute of Technology, United Nations, White House Cybersecurity Initiative and the Forum argue that todays social structures do not readily accommodate the new reality of integrated systems that can leverage autonomous, dynamic, digital feedback mec

45、hanisms. Our social structures struggle to adapt to digital methods, which can illuminate trust between data-sharing systems by transparently tracking when and how data is accessed or exchanged.10 In other words, despite many technical solutions designed to encourage trustworthy behaviour between da

46、ta-sharing partners once a consortium is up and running, establishing trust at the beginning of the relationship is nevertheless contingent on our everyday social structures and perceived social relationships. Before beginning to form social relationships with partners, however, it is important to s

47、elect the correct partners for a data consortium. Identifying the best partners requires understanding of another institutions origin, strategic goals and its research objectives for prospective data consortium partners and whether or not these align with similar metrics from your institution. A tho

48、rough vetting process at the beginning of the relationship cannot be facilitated with a quick website check or even a phone call but requires a series of in-person meetings. At the start of the Breaking Barriers to Health Data project, the Forum found that several iterations of discussion and reiter

49、ation of purpose were necessary with each prospective institution before it was possible to move on to discuss details of a partnership. Traveling in person to the location of a prospective partner institution eases the process of uncovering the day-to-day operations and team norms that will be contributed to the data consortium by a prospective partner. At this recommended in-person meeting (or series of meetings), it is important to discuss: (1) what type of data each institution is currently collecting; (2) how each institution runs its day- to