SARS-CoV-2 data sharing and privacy in Canada
Guest host Emma Griffiths talks with Finn McGuire and William Hsiao about harmonising SARS-CoV-2 surveillance metadata and sharing genomic data across Canada. They explain CanCoGen’s metadata tiers, practical tools for standardisation, and the provincial privacy rules and governance challenges that can delay public release. The discussion asks how technical infrastructure, trust and clearer policy can make a decentralised public health system more effective.
Listen to Episode 54: SARS-CoV-2 In Canada and addressing data sharing and privacy on SoundCloud
In this episode
Three tiers of SARS-CoV-2 metadata
Information from multiple contributors feeds Canada’s national genomic surveillance database. William Hsiao explains that CanCoGen divided metadata into three tiers: minimum, low-risk, de-identified information intended for public release; a middle tier for national surveillance and sharing between provinces and the national laboratory; and identifiable local metadata that partners were encouraged to retain themselves. Keeping local records in a standardised format would make later de-identification and research use easier, even when there was no time to prepare them during the pandemic.
Agreeing on the minimum tier took several months. Age, gender and sampling date became points of disagreement between jurisdictions over what counted as identifiable information. The team consulted legal experts and prepared memos explaining its position. Hsiao stresses that sampling dates are particularly valuable for tracking viruses and understanding transmission.
Standards, sampling context and Data Harmonizer
Hsiao describes a SARS-CoV-2 metadata standard developed through the Public Health Alliance for Genomic Epidemiology and adopted for CanCoGen. Emma Griffiths leads the harmonisation work. Consultation involved academic researchers, public health workers and other stakeholders, and the standard was made publicly available for reuse and contributions.
The fields extend beyond minimum metadata. Reasons for sequencing, reasons for sample collection, sampling scope and inclusion or exclusion criteria help analysts understand the denominator behind epidemiological calculations. Without that context, interpreting a collection of genomes is harder.
Hsiao’s group developed Data Harmonizer to capture standardised metadata, validate its format and export records for public repositories or national laboratory reporting. Griffiths also leads a curation team that helps stakeholders clean and convert raw data. Some records are sufficiently incompatible that manual harmonisation remains necessary.
Why public data release was slow
Griffiths raises criticism of Canada’s pace of sharing data with public repositories. McGuire adds that completeness matters as well as availability: he cites plots by Art Poon at Western showing incomplete date metadata in Canadian genome submissions, with substantial differences between provinces. He connects this variation to provincial legal systems, privacy rules and relationships with the federal government.
Hsiao distinguishes several sources of delay. Provinces have different laws, but also different privacy officers and lawyers interpreting them. Some are willing to release sequences while withholding full date information. He characterises Canada’s approach as comparatively risk-averse, while recognising legitimate quality concerns. Sequence and metadata quality control, including de-hosting the sequence data, also took time. CanCoGen’s coordinated work was intended to address these barriers individually rather than treat slow release as one simple problem.
Federal powers, trust and public opinion
McGuire points to the Public Health Agency of Canada Act and the Statistics Act as sources of federal powers to compel data sharing. He argues that exercising those powers is not straightforward when the system depends on working relationships between provinces, territories and the federal government. Relationships with sovereign First Nations add another dimension, including understandable concern about federal intervention. He also reports arguments that the provincial–federal gap could leave Canada falling short of international health information-sharing obligations.
Hsiao identifies public opinion as a missing part of the discussion. Better evidence about people’s willingness to share minimum de-identified information could help authorities make decisions with greater confidence. He sees fear of breaching privacy or public trust as an important reason for caution.
What is working, and what remains unresolved
McGuire sees progress when working groups and informal calls connect laboratory staff across jurisdictional boundaries. Even where governments disagree about data sharing, practitioners are tackling similar problems and want to exchange experience. He also welcomes discussion of lasting infrastructure and broader use of genomic epidemiology, including for foodborne disease and within hospitals.
Hsiao calls the collaboration a trust-building exercise. Meetings and shared expertise help a large network reach agreement, but technical progress has outpaced understanding of the governance needed to use it. He identifies legal, ethical and regulatory questions, guidance documents and public attitudes as priorities for CanCoGen’s second year. Better pipelines alone cannot resolve those gaps.
Advice for decentralised surveillance systems
For countries building SARS-CoV-2 sequencing capacity, McGuire recommends centralising analysis, data collection and curation where possible, with shared repositories or analytical platforms. He mentions IRIDA and SP3, the latter in relation to tuberculosis, as examples of approaches to shared analysis. Clearly communicated quality-control metrics should also be checked centrally.
The guests argue that good genomic surveillance must connect to policy. McGuire praises the UK’s genomic epidemiology effort while criticising how its Conservative government used the information. He also discusses shortcomings in protecting vulnerable people in Canada, including the tension between reduced shelter capacity for distancing and exposure to winter cold. He argues that domain experts need greater involvement in implementation and political decisions.
Hsiao’s closing advice is not simply to centralise everything: web-based technology can support a connected decentralised system. The essential requirement is a trust framework that enables data, information and expertise to move between organisations rather than remain in silos.
Highlights
- 00:00:57 — Harmonising metadata from multiple contributors to national genomic surveillance
- 00:01:25 — CanCoGen’s three metadata tiers and negotiations over minimum public information
- 00:05:10 — Developing the metadata standard, adding sampling context and using Data Harmonizer
- 00:08:12 — Criticism of Canada’s pace of sharing data with public repositories
- 00:09:00 — Incomplete date metadata and differences between provinces
- 00:10:30 — Provincial privacy interpretations and quality-control work behind release delays
- 00:13:17 — Federal data-sharing powers and the relationships that complicate their use
- 00:15:08 — Understanding public opinion on sharing minimum de-identified information
- 00:16:51 — Working groups, laboratory collaboration and longer-term genomic infrastructure
- 00:17:50 — Building trust and closing the gap between technical tools and governance
- 00:20:25 — Advice on shared analysis, centralised curation and consistent quality-control checks
- 00:24:08 — Using web-based technology and trust frameworks to connect decentralised systems
In their own words
But the key there is to have a trust framework that enables data to be shared in that decentralized system rather than setting up silos in decentralized system.
— William Hsiao, 00:24:08
Who is talking
- Emma Griffiths (host)
- Finn McGuire (guest)
- William Hsiao (guest)
- Lee Katz (host)
- Andrew Page (host)
- Nabil-Fareed Alikhan (host)
Also mentioned: Art Poon.
Tools and resources mentioned
Data Harmonizer, IRIDA, SP3.
Questions this episode answers
What were CanCoGen’s three metadata tiers?
They were minimum de-identified metadata for public release, a middle tier for national surveillance and inter-jurisdictional sharing, and identifiable local metadata retained by partners. Local information was still encouraged to follow a standard format so it could support later de-identification and research.
What does Data Harmonizer do for SARS-CoV-2 metadata?
Hsiao describes it as a tool for capturing metadata in a standardised format, validating records and exporting them in formats required by public repositories or national reporting. A curation team also helps with data that require manual cleaning and conversion.
Why was Canada slow to share SARS-CoV-2 genomic data?
The guests identify several factors: different provincial privacy laws and interpretations, disagreements over releasing full sampling dates, fragmented governance and a cautious approach to public disclosure. Sequence and metadata quality control, including de-hosting, also contributed to delays.
What advice do the guests give for decentralised genomic surveillance?
McGuire recommends shared analysis and data infrastructure, with clearly communicated quality-control metrics checked centrally. Hsiao emphasises connecting decentralised organisations through a trust framework that supports sharing data, information and expertise.
This page was written from a machine transcript of the episode and checked against it. Listen to the episode for the whole conversation.
Show notes
The Canadians have taken over the podcast - AGAIN ! Join guest host Dr. Emma Griffiths, as she talks with Dr. Finn McGuire and Dr. William Hsiao about the SARS-CoV-2 genomics epidemiology efforts in Canada. Cancogen website : https://www.genomecanada.ca/en/cancogen