When information is the treatment? Precision medicine in healthcare

Abstract

Profoundly more data-intensive than conventional medicine, precision medicine’s distinctive informational needs present new challenges for healthcare management. Data protection and privacy law are key determinants in precision medicine’s future. This article examines legal and regulatory barriers to the incorporation of precision medicine into healthcare. Specific attention is paid to analyzing recent health privacy laws, court cases, and medical device regulations. Considering the challenges identified, recommendations and guidance are crafted for health leaders with reference to domestic and international initiatives.

Introduction

Profoundly more data-intensive than conventional medicine, precision medicine presents new challenges for health leaders. With its potential for tailored interventions that permit early diagnosis and treatment of disease, precision medicine promises to improve disease prevention and reduce the costs associated with providing care.¹ Such interventions raise the prospect of better targeting resources for those who need them, contributing in turn to the affordability and sustainability of universal healthcare in Canada.

This article examines precision medicine’s informational needs, then focuses on health privacy law and the role health leaders must play in the creation of data-sharing frameworks that will make precision medicine possible. The article then offers guidance and recommendations for health leaders with reference to domestic and international initiatives.

Precision medicine’s informational needs

Making precision medicine a reality implicates health leaders in the creation of healthcare management infrastructure that meets precision medicine’s distinctive informational needs. Before delivering interventions tailored to the individual patient, clinicians must first understand what precisely makes this patient unique vis-à-vis the rest of the population through situating the patient in a rich network of other patients’ health, sociodemographic, and environmental data. Such an approach is in stark contrast to conventional medicine, where clinicians share patient data only for the purpose of treating that same patient. The informational infrastructure of conventional medicine reflects this: “Clinical systems are built to isolate different data sets such as imaging, pathology and laboratory tests…”²

Precision medicine, by contrast, requires that health data flow from individual medical record into different research contexts and then back into a learning healthcare system.³ Precision medicine is inextricably intertwined with the research context, presenting a workflow to which many health leaders are not accustomed. More concretely, data collected during the course of clinical care are essential to the study of real-world clinical outcomes, which in turn informs future courses of treatment for other patients.² Precision medicine blurs the boundaries between the traditional categories of “clinical” and “research” data by requiring data for multiple purposes.

Big health data

An individual’s genomic data forms the bedrock of precision medicine. With next-generation sequencing techniques, it is now faster and more cost-effective to sequence an entire genome than sequencing a single gene using traditional Sanger techniques.⁴ Experience with datasets from The Cancer Genome Atlas and elsewhere has already shown the potential of genomics in clinical settings.^5
–7 For example, we have shifted from speaking of site-specific cancers to ones defined by genetic biomarkers. Parkinson’s is also now understood to be a group of diseases, each one owing to a different genetic mutation.⁸

The amount of data required to move beyond cancer and rare diseases, where genomics in healthcare is beginning to show success, should not be underestimated. Due to the diversity of human genomes in modern heterogeneous populations, information from thousands of individual genomes, across diverse populations and contexts, is needed to interpret the information contained in just one.⁹

An individual’s genome, however, is just one of multiple important data sources upon which precision medicine relies. The delivery of personalized, precise medical interventions is part of the larger trend of big data in health, defined as “encompass[ing] high volume, high diversity biological, clinical, environmental, and lifestyle information collected from single individuals to large cohorts, in relation to their health and wellness status, at one or several time points.”¹⁰

Such “standardized” contributions of health and genetic information from medical care may well contribute to the well-being of both actual and future patients. Nevertheless, this would require a major shift in healthcare management infrastructure as the need for more data from diverse sources means that hospital managers and political leaders in healthcare systems will have to drive this forward at a time when an emphasis on the privacy protections of personal, individual data is expanding. Moreover, stagnating budgets coupled with growing populations make the cost-saving prospects of targeted medical inventions even more appealing.

More concretely, health leaders must take a broad view of the healthcare system, ensuring that there is strategic oversight of both infrastructure and personnel. Effective coordination and administration of infrastructure such as clinical and laboratory services, information technology (IT) systems, and data interpretation is essential. This infrastructure does not exist in a vacuum. Healthcare professionals will depend on leaders to deliver adequate training, ensuring that this new infrastructure is being effectively used and best practices are disseminated.¹¹ All the while, health leaders also owe patients a duty to create trustworthy data governance frameworks that accord with patient expectations regarding the privacy of their health data.¹² The UK’s new National Genomic Medicine Service, with its emphasis on coordinating informatics and clinical infrastructure along with workforce development, exemplifies the type of strategy Canadian health leaders should be working toward.^13,14 Indeed, the implementation of precision medicine in healthcare invites leaders to consider the diverse groups they serve and see to it that their needs are met.

Health privacy

Given precision medicine’s distinctive informational needs, health privacy legislation and policy will be key factors shaping medical progress. In recognition of the specificities of Personal Health Information (PHI), almost all Canadian provinces and territories have adopted legislation that governs PHI. At their core, these health privacy laws create a privacy-respecting framework for the collection, use, and disclosure of PHI.

Since just 2015, three new PHI statutes have come into force: Northwest Territories’ Health Information Act (HIA) (October 2015), Yukon’s Health Information and Privacy Management Act (August 2016), and Prince Edward Island’s HIA (July 2017).^15
–17 While our focus is legislation that deals specifically with PHI, it is worth noting that the Canadian federal government, as well as all provinces and territories, have their own public and private sector privacy legislation that governs data that are not PHI.

Defining “identifiable”

For health privacy laws, the essential question is always whether the data are personally identifiable. If not personally identifiable, such data can flow freely, not subject to the duties in the relevant privacy legislation. Data that are subject to privacy legislation change in time, moving in tandem with re-identification techniques and the existence of other samples and data on that person.

Advanced computational techniques associated with big data and genomics in healthcare present two primary difficulties. One is that as techniques evolve, the risk of identification changes over time. Data that would have once been considered de-identified are now no longer.¹⁸ The second issue is that of genomic information’s sensitivity in light of the uniqueness of an individual’s genome. The development of novel re-identification techniques where genetic data are used has greatly narrowed the circumstances in which we can confidently say that the data cannot be traced back to an individual.¹⁹ While researchers pursuing legitimate aims, for example, the identification of redundant individuals in data sets,^20,21 have been responsible for many of such developments, the techniques remain at the disposal of individuals who may act in bad faith.

Health leaders find themselves in a position where maintaining the trust of patients can be difficult. Indeed, a 2017 survey revealed that patients are more reluctant to share medical information in light of reports of data breaches.²² While prohibitions on re-identifying research participants are often part of data access agreements, the deployment of security safeguards, coupled with the development of accountability and compliance frameworks, is essential in minimizing potential harms flowing from re-identification.²³ Research ethics guidelines also recommend that participants be informed of “reasonably foreseeable risks,” including potential privacy harms.²⁴

Compounding the problem is the fact that the definition of “identifiable” is uniform in neither law nor genetics literature. Most provinces and territories define identifiable as “reasonably foreseeable from a combination of data.”^{15
–17,25

–29} Some provinces, however, define identifiable data as those data that make the identity of a person “readily ascertainable,”^30,31 as allowing for identification,^32,33 or do not clearly define it.³⁴ The definitions that abound in genetics literature represent a “dangerously inconsistent and confusing set of terms describing the identifiability of genetic data,”¹⁹ which has only gotten worse in recent years.

This Tower of Babel poses a fundamental problem for data sharing. Personal health information custodians must determine whether the data are identifiable and treat it appropriately. Where an information custodian less stringently interprets “identifiable,” they risk sharing data capable of being re-identified outside of the legislation’s protective framework. At the other extreme, an overly cautious interpretation of “identifiable” may result in the needless removal of potentially valuable information from datasets, and worse, no “precise” information going back for the treatment of the individual patient. In this case, we risk stifling progress by leaving information, valuable for both research and care, underutilized. Indeed, it has been recognized that the variations in data regulations among Canadian jurisdictions hinder national and international studies.³⁵

This challenge is an especially important one for health leaders to confront. Achieving better outcomes does not always require incurring expenses and collecting more data. Rather, health leaders must critically take stock of the data they have and create data-sharing policies that maximize the potential utility that can be derived from such data, while complying with legal requirements.

Consent

Nestled within this complexity is consent, the fulcrum to privacy legislation in Canada. The collection, use, and disclosure of personal information and PHI all require either individual consent or authorization by law. Meaningful, specific, and informed consent is the standard in privacy legislation and in clinical settings, though broad consent, subject to oversight, is recognized for certain research settings, for example, longitudinal studies. Precision medicine, however, with its intimate link to genetic testing challenges the established paradigm in the clinical setting of an individual consenting to a procedure, as often the provision of the resulting genetic at-risk information is “the” treatment.

Biobanks are also becoming part of medical care and are key repositories of genomic and phenotypic information, providing the building blocks for understanding the multivariate relationships a given disease has with an individual.³⁶ Nonetheless, consent in the context of population biobank research has raised several issues because individuals consent for future, unspecified research purposes at the time biospecimens and data are taken, which will be approved by ethics boards.^37,38 Such broad consent has been recognized by the Government of Canada’s 2014 Tri-Council Policy Statement: Ethical Conduct for Research Involving Humans but not explicitly by legislation.²⁴ A solution for the next decade of healthcare could be to explicitly mention broad data sharing in medical care and not just research as part of hospital admission notifications while including a possible individual opt out.

Health privacy in the courts

The courts are also shaping the future of precision medicine. In a recent decision, the Quebec Court of Appeal struck down key provisions of the federal Genetic Non-Discrimination Act (2018) that protected individuals from needing to undergo or disclose genetic test results for insurance and employment contracts.^39,40 By reopening the door to legitimate public concerns about genetic discrimination, the decision risks discouraging participation in genetic testing. The very purpose of the sections that the court struck down was to allow access to the benefits of genetic testing without fear that the information could be used against genetic test participants. At the same time, the law had limited scope, internal flaws, and indirectly could be seen as exacerbating genetic exceptionalism. The saga is not over—the Canadian Coalition for Genetic Fairness has initiated an appeal at the Supreme Court of Canada, with a hearing scheduled for October 2019.

In another recent case, a unanimous Supreme Court protected the privacy of patient health records in British Columbia, finding that anonymized healthcare databases did not need to be released as evidence for the benefit of a big tobacco company’s defense.⁴¹ While anonymized databases do not contain “personal” data, the decision forms part of the complex puzzle of maintaining the public’s trust that the privacy of health records will be preserved.

New players in the clinical setting: Cloud storage and AI

Big health data has spurred developments in data processing in the domain of cloud computing and Artificial Intelligence (AI) applications. Notably, the increased reliance of cloud computing to deliver economical, scalable solutions to the needs of big data storage and processing means that entities engaged in healthcare management and delivery are transmitting patient data to Cloud Service Providers (CSPs). This development further complicates the patchwork of laws and regulations in three ways: the division of obligations between data custodians and CSPs is unclear, the assessment of risks involved in transnational data transfers to CSPs is left to ill-equipped data custodians, and the data handling and privacy practices of CSPs are opaque.⁴²

Implementing machine learning and other so-called AI technologies is one of the key ways we can translate big health data into tangible clinical results.⁴³ Artificial intelligence technologies have already begun outperforming expert clinicians in the diagnosis of melanoma, metastatic breast cancer, and certain eye diseases.^44

–48 Yet the promise and potential of AI in healthcare is not without issue. Current regulations surrounding genomic software applications and other algorithmic technologies make determining whether such applications are clinical support tools, medical devices, or in vitro diagnostic devices is unclear and has great implications for the continued development and safe adoption of such technologies.⁴⁹

Getting strategic about solutions

Are Canadian health leaders sufficiently trained in digital health and its systemic implications? We have seen that the barriers to the creation and management of a healthcare system that delivers on the promise of precision medicine and big health data are diverse. In light of such challenges, Innovation, Science and Economic Development Canada in a 2018 report recommended a national digital health strategy built upon three pillars: interoperable systems, harmonized data and privacy frameworks, and a single electronic health record for every patient.⁵⁰ Achieving these goals will not happen without innovative leadership.

Speaking the same language—Defining “identifiable”

Health leaders have a central role to play in developing data-sharing frameworks for their organizations. The choices surrounding when data can be shared greatly relies on what is considered identifiable data. Accordingly, the definition of identifiable should be harmonized across Canada. Defining identifiability as “reasonably foreseeable from a combination of data” not only allows for a definition that responds to context but also aligns with that used in Europe.⁵¹ Likewise, a flexible, context-specific approach should be adopted to the question of de-identification. In a recent report, the Council of Canadian Academies recommended viewing de-identification as a continuum, with access controls responding to the appropriate form of de-identification (eg, anonymization, pseudo-anonymization via coding, etc).⁵² Such an approach would not only change the all-or-nothing bite of current privacy legislation but also would be responsive to changing re-identification methods and the purposes for which data are being shared.

The way that we speak about privacy and consent should also develop with advances in medicine and science. In treating privacy either in the abstract, or by solely focusing on hypothetical privacy harms, we fail to take seriously the potential benefits that flow from data sharing. A flexible, responsive approach should be adopted that balances the public’s interest in medical advances spurred by big health data with realistic privacy risks for individuals. Possible ways forward include a three-part privacy test, which takes into account the sensitivity, the potential harm resulting from re-identification, and the expectations of individuals regarding the sharing of their health data.⁵³ Adopting such an approach would assist in determining the degree of protection such data merits, and, in turn, under which conditions such data can be shared.

Avoiding genomic exceptionalism

We must also avoid falling into the trap of genomic exceptionalism within the privacy debate. Genomic information’s relevance to scientific progress is too great to treat it as a “no-go” zone. Code-protected genomic information should be treated in a way that accords with society’s expectations and responds to realistic potential privacy harms that can result from its improper use.

In recognition of this interdependence of precision medicine and open data, a critical rethinking of current data practices is needed.⁵⁴ Opt-out organ donation programs provide inspiration. Individual genomic information would, by default, be shared with clinicians and researchers and be accompanied by notification of this approach with a right to opt out, thereby respecting an individual’s preferences and sensitivities. Similarly, some recommend that automatic sharing of a small number of individual genetic variants (alleles) to be included in the medical record along with limited clinical info become the new norm in medical care.⁹

The right to science in Canada

Inspiration can be drawn from the 1948 Universal Declaration of Human Rights and the 1976 International Covenant on Economic, Social and Cultural Rights. Both maintain the human right of everyone to enjoy the benefits of scientific progress and its applications (“the right to science”). Health leaders can help in developing the right in tandem with the informational needs of precision medicine so that everyone may benefit. For example, the Global Alliance for Genomics and Health’s 2014 Framework interprets the right to science as including the duty of scientists to “access and share genomic and health-related data across the translation continuum, from basic research through practical applications.”⁵⁵ In a similar vein, the World Health Organization has now begun advocating for global data sharing outside the contexts of public health emergencies, such as in tuberculosis clinical trials.⁵⁶

Potential lessons from Australia

Australia’s health leaders, themselves working in a federal country like Canada, have recently shown the kind of strategic, collaborative thinking integrating genomics into healthcare requires. While Australia has a federally administered national health service, the states and territories fund clinical and laboratory genetic services. Founded in 2014, the Australian Genomics Health Alliance is a partnership of 80 hospitals, research institutes, universities, and laboratories, whose aim is to “provide evidence for the equitable, effective and sustainable delivery of genomic medicine in healthcare.”⁵⁷ Its current projects include a pilot for a shared, interoperable national database of clinical genomics information.⁵⁸

With a view to appropriately harnessing the potential of genomics in the healthcare system, the Council of Australian Governments (COAG) has developed a National Health Genomics Policy Framework, with responsible data practices as one of its five strategic priorities.⁵⁹ In 2017, COAG also approved a National Digital Health Strategy, which includes interoperability of data as one of its strategic priorities.⁶⁰ In July 2016, the Australian Digital Health Agency was launched, which is tasked with the creation of an electronic health record for all Australians.

Conclusion

We will not achieve precision medicine without engaging health leaders in the construction and management of a learning health system. Such a system relies on big data to enable stratification for targeted allocation both of resources and of medical care. It supports a data trajectory from research to clinic and clinic back to research. Consents to research will necessarily vary by the specific type of research and foresee and facilitate secondary use and linkage along the trajectory. For example, broad consent is needed for longitudinal studies and for research across possibly related diseases and environments. Privacy and security protections should be proportionate to the real risks of re-identifiability. While allowing for opt-out, residual medical care samples and core elements from the medical record could, as mentioned, be automatically shared within the health learning system for improved care for all based on real-time research.

Precision medicine’s informational needs exacerbate the cacophony of privacy frameworks and uncertain regulations and further challenge fundamental concepts such as consent and the now-inextricable role of genomic data in medical care. We consider that the treatment of the informational needs of precision medicine may contribute to making Canada’s universal healthcare system sustainable and affordable for the long term via targeted approaches. Who will lead in crafting the solutions required for sustained dialogue among the panoply of stakeholders precision medicine implicates—patients, clinicians, policy-makers, IT professionals, and others? Precision medicine is simultaneously all about the individual patient and the rest of us actual and future patients, and so implies a radical shift in healthcare systems and their data management. Do we dawdle or deal with it?

Footnotes

ORCID iD

Michael Beauvais

References

Alyass

Turcotte

Meyre

. From big data analysis to personalized medicine for all: challenges and opportunities. BMC Med Genomics. 2015;8(1):33. doi:10.1186/s12920-015-0108-y.

Hulsen

Jamuar

Moody

, et al. From big data to precision medicine. Front Med. 2019;6:34. doi:10.3389/fmed.2019.00034.

Faden

Kass

Goodman

Pronovost

Tunis

Beauchamp

. An ethics framework for a learning healthcare system: a departure from traditional research ethics and clinical ethics. Hastings Cent Rep. 2013;43(s1):S16–S27. doi:10.1002/hast.134.

Pennell

Mutebi

Zhou

Z-Y

, et al. Economic impact of next generation sequencing vs sequential single-gene testing modalities to detect genomic alterations in metastatic non-small cell lung cancer using a decision analytic model. J Clin Oncol. 2018;36(suppl 15):9031–9031. doi:10.1200/JCO.2018.36.15_suppl.9031.

Sanchez-Vega

Mina

Armenia

, et al. Oncogenic signaling pathways in The Cancer Genome Atlas. Cell. 2018;173(2):321–337.e10. doi:10.1016/j.cell.2018.03.035.

Cava

Bertoli

Colaprico

Olsen

Bontempi

Castiglioni

. Integration of multiple networks and pathways identifies cancer driver genes in pan-cancer analysis. BMC Genomics. 2018;19(1):25. doi:10.1186/s12864-017-4423-x.

The Cancer Genome Atlas Research Network, Chang

Creighton

Mills

, et al. The Cancer Genome Atlas Pan-Cancer analysis project. Nat Genet. 2013;45(10):1113.

Weiner

. There is no Parkinson disease. Arch Neurol. 2008;65(6):705–708. doi:10.1001/archneur.65.6.705.

Wright

Ware

Lucassen

, et al. Genomic variant sharing: a position statement [version 1; peer review: 1 approved, 1 approved with reservations]. Wellcome Open Res. 2019;4:22. doi:10.12688/wellcomeopenres.15090.1.

10.

Auffray

Balling

Barroso

, et al. Making sense of big data in health research: towards an EU action plan. Genome Med. 2016;8(1):71. doi:10.1186/s13073-016-0323-y.

11.

Health Education England. The Topol Review: Preparing the Healthcare Workforce to Deliver the Digital Future. National Health Service; 2019:1–53. Available at: https://topol.hee.nhs.uk/. Accessed June 3, 2019.

12.

Kalkman

Mostert

Gerlinger

van Delden

JJM

van Thiel

GJMW

. Responsible data sharing in International Health Research: a systematic review of principles and norms. BMC Med Ethics. 2019;20(1):21. doi:10.1186/s12910-019-0359-9.

13.

Health Minister: NHS must lead the world in genomic healthcare. GOV.UK. 2019. Available at: https://www.gov.uk/government/news/health-minister-nhs-must-lead-the-world-in-genomic-healthcare. Accessed June 3, 2019.

14.

NHS England. NHS Genomic Medicine Service. 2019. Available at: https://www.england.nhs.uk/genomics/nhs-genomic-med-service/. Accessed June 3, 2019.

15.

Health Information Act, SNWT 2014, c 2.

16.

Health Information Privacy And Management Act, SY 2013, c 16.

17.

Health Information Act, RSPEI 1988, c H-1.41.

18.

Kosseim

Brady

. Policy by procrastination: secondary use of electronic health records for health research purposes. McGill J Law Health. 2008;2(5).

19.

Phillips

Knoppers

. The discombobulation of de-identification. Nat Biotechnol. 2016;34(11):1102–1103.

20.

Homer

Szelinger

Redman

, et al. Resolving individuals contributing trace amounts of DNA to highly complex mixtures using high-density SNP genotyping microarrays. PLoS Genet. 2008;4(8):e1000167–e1000167. doi:10.1371/journal.pgen.1000167.

21.

Cai

Hao

Winslett

, et al. Deterministic identification of specific individuals from GWAS results. Bioinformatics. 2015;31(11):1701–1707. doi:10.1093/bioinformatics/btv018.

22.

Black Book Research. Healthcare’s Digital Divide Widens, Black Book Consumer Survey. 2017. Available at: https://www.prnewswire.com/news-releases/healthcares-digital-divide-widens-black-book-consumer-survey-300384816.html. Accessed May 24, 2019.

23.

Joly

Dove

Knoppers

Bobrow

Chalmers

. Data sharing in the Post-Genomic world: the experience of the International Cancer Genome Consortium (ICGC) Data Access Compliance Office (DACO). PLoS Comput Biol. 2012;8(7):e1002549. doi:10.1371/journal.pcbi.1002549.

24.

Canadian Institutes of Health Research, Natural Sciences and Engineering Research Council of Canada, and Social Sciences and Humanities Research Council of Canada. Tri-Council Policy Statement: Ethical Conduct for Research Involving Humans. Ottawa, Ontario, Canada: Canadian Institutes of Health Research, Natural Sciences and Engineering Research Council of Canada, and Social Sciences and Humanities Research Council of Canada; 2014.

25.

The Health Information Protection Act, SS 1999, c H-0.021.

26.

Personal Health Information Protection Act, 2004, SO 2004, c 3, Schedule A.

27.

Personal Health Information Privacy and Access Act, SNB 2009, c P-7.05.

28.

Personal Health Information Act, SNS 2010, c 41.

29.

Personal Health Information Act, SNL 2008, c P-7.01.

30.

Health Information Act, RSA 2000, c H-5.

31.

The Personal Health Information Act, CCSM c P33.5.

32.

Act Respecting Access to Documents Held by Public Bodies and the Protection of Personal Information, CQLR c A-2.1.

33.

Act Respecting the Protection of Personal Information in the Private Sector, CQLR c P-39.1.

34.

E-Health (Personal Health Information Access and Protection of Privacy) Act, SBC 2008, c 38.

35.

Cate

. Provincial Canadian Geographic Restrictions on Personal Data in the Public Sector. Washington, DC: The Centre for Information Policy Leadership, Hunton & Williams LLP; 2008. Available at: https://www.huntonak.com/files/Publication/2a6f5831-07b6-4300-af8d-ae30386993c1/Presentation/PublicationAttachment/0480e5b9-9309-4049-9f25-4742cc9f6dce/cate_patriotact_white_paper.pdf. Accessed April 2, 2019.

36.

Saha

Hurlbut

. Treat donors as partners in biobank research. Nature. 2011;478(7369):312.

37.

Allen

Joly

Moreno

. Data sharing, biobanks and informed consent: a research paradox. 2013. McGill J Law Health. 7(1):85–120.

38.

Zawati

Joly

Knoppers

. L’autonomie basée sur l’Individualisme libéral: les limites dans le contexte des biobanques populationnelles. In: Mamzer

Martinent

Stanton-Jean

, eds. Réflexion et Recherches En Éthique. Paris, France: Dalloz; 2018:279–293.

39.

Dans l’affaire du: Renvoi relatif à la Loi sur la non-discrimination génétique édictée par les articles 1 à 7 de la Loi visant à interdire et à prévenir la discrimination génétique., 2193 (Quebec Court of Appeal 2018).

40.

Genetic Non-Discrimination Act, SC 2017, c 3.

41.

British Columbia v Philip Morris International, Inc.(Supreme Court of Canada 2018).

42.

Thorogood

Simkevitz

Phillips

Dove

Joly

. Protecting the privacy of Canadians’ health information in the cloud. Can J Law Technol. 2016;14(1).

43.

Yang

Xue

, et al. Translating cancer genomics into precision medicine with artificial intelligence: applications, challenges and future perspectives. Hum Genet. 2019;138(2):109–124.

44.

De Fauw

Ledsam

Romera-Paredes

, et al. Clinically applicable deep learning for diagnosis and referral in retinal disease. Nat Med. 2018;24(9):1342.

45.

Bejnordi

Veta

Van Diest

, et al. Diagnostic assessment of deep learning algorithms for detection of lymph node metastases in women with breast cancer. JAMA. 2017;318(22):2199–2210.

46.

Haenssle

Fink

Schneiderbauer

, et al. Man against machine: diagnostic performance of a deep learning convolutional neural network for dermoscopic melanoma recognition in comparison to 58 dermatologists. Ann Oncol. 2018;29(8):1836–1842.

47.

Kristensen

Lingjærde

Russnes

Vollan

HKM

Frigessi

Børresen-Dale

A-L

. Principles and methods of integrative genomic analyses in cancer. Nat Rev Cancer. 2014;14(5):299.

48.

Kalinin

Higgins

Reamaroon

, et al. Deep learning in pharmacogenomics: from gene regulation to patient stratification. Pharmacogenomics. 2018;19(7):629–650.

49.

Thorogood

Touré

Ordish

Hall

Knoppers

. Genetic database software as medical devices. Hum Mutat. 2018;39(11):1702–1712.

50.

Report from Canada’s Economic Strategy Tables: Health and Biosciences. Ottawa, Ontario, Canada: Innovation, Science and Economic Development Canada; 2018. Available at: https://www.ic.gc.ca/eic/site/098.nsf/eng/00025.html. Accessed February 28, 2019.

51.

Regulation 2016/679 of the European Parliament and of the Council of 27 April 2016, General Data Protection Regulation.

52.

The Expert Panel on Timely Access to Health and Social Data for Health Research and Health System Innovation, Council of Canadian Academies. Accessing Health and Health-Related Data in Canada. Ottawa, Ontario, Canada: Council of Canadian Academies; 2015:1–260. Available at: https://www.scienceadvice.ca/reports/accessing-health-and-health-related-data-in-canada/. Accessed February 28, 2019.

53.

Dyke

Dove

Knoppers

. Sharing health-related data: a privacy test? NPJ Genomic Med. 2016;1(1):16024.

54.

Knoppers

Thorogood

. Ethics and big data in health. Curr Opin Syst Biol. 2017;4:53–57.

55.

Knoppers

. Framework for responsible sharing of genomic and health-related data. HUGO J. 2014;8(1):3.

56.

World Health Organization. WHO strongly supports global sharing of data from TB clinical trials to maximize learning and accelerate discoveries. World Health Organization. 2018. Available at: http://www.who.int/tb/features_archive/global-sharing-of-data-from-TB-clinical-trials/en/. Accessed April 2, 2019.

57.

Australian Genomics. 2018. Available at: https://www.australiangenomics.org.au/about-us/australian-genomics/. Accessed April 2, 2019.

58.

Australian Genomics. Genomic data storage. 2018. Available at: https://www.australiangenomics.org.au/our-research/a-national-approach-to-data-federation-and-analysis/. Accessed April 2, 2019.

59.

Health AGD of National Health Genomics Policy Framework 2018-2021. 2017. Available at: http://www.health.gov.au/internet/main/publishing.nsf/Content/national-health-genomics-policy-framework-2018-2021. Accessed April 2, 2019.

60.

Australia’s National Digital Health Strategy. Safe, Seamless and Secure: Evolving Health and Care to Meet the Needs of Modern Australia. Sydney, Australia: Digital Health Agency:1–63. 2017. Available at: https://conversation.digitalhealth.gov.au/sites/default/files/adha-strategy-doc-2ndaug_0_1.pdf. Accessed February 28, 2019.