Ensuring the success of data sharing in Canada

Few biomedical researchers have had exposure to training to enable data sharing, nor is this part of typical graduate or clinical training. We have outlined the educational and training considerations (Taylor et al. 2020) for implementing data-sharing practices. While universities may elect to develop in-house educational resources, this runs a risk of creating heterogeneity of content and (or) missing key content. We recommend that a central committee be established to work collaboratively on a syllabus of core content that can be further tailored to meet individual institutional and discipline-specific needs. This approach will help reduce mixed messaging of content to researchers across the country and reduce duplication of effort and costs.

It will be critical to engage researchers as educational training is conceived of and developed. There are likely a few core data-sharing practices that will benefit from being taught across fields within medicine. For example, the FAIR principles (Findable, Accessible, Interoperable, Reusable) are likely key. It makes sense to roll out any training of the FAIR principles and other key data-sharing concepts using standardized content. This will help ensure that core messages seamlessly cross disciplines. However, within and between disciplines there will be a need for educational tailoring. For example, sharing sequencing data is different from sharing data from randomized trials. Even within disciplines there should a push to use core content across the field where possible. To do this effectively will require identifying ways to create dialogue with the research community to build grassroots informed training that identifies and addresses context-specific needs of different research communities.

Once a syllabus is developed, we recommend that a central group lead the development of the content and as needed contracting out particular modules for development. We have developed OS material that includes a content pretest and post-test with learners who reach a certain threshold (i.e., 80%) of knowledge being able to receive a certificate of competence. These practices may be relevant to consider. Whatever resources are developed they need to be easily scalable. These resources will also need to be available in English, French, and Indigenous languages. Similarly, making them available on an open educational resource platform will be critical. We also think it important to build a database of best-practice examples of how researchers have shared their data.

While requiring data sharing itself is a few years off in Canada, there are important considerations to think about now to ensure its successful implementation when it is required. The resources required will take time to develop and implement. To facilitate data-sharing costs, we recommend that funders allow grant applicants to include a budget line item for data sharing. The costs might differ considerably between research areas and between clinical and preclinical research. Given that many universities will be starting from a low baseline of data sharing, it is likely that for the first few years the costs of data sharing may need to be shared jointly by funders and universities. We can imagine a scenario where for every successful grantee their home research institution will contribute to a central OS fund. The ratio of funding between an institution and funder could be 1 to 2; for every dollar provided by a funder 2 dollars would be given by the research organization. This would help facilitate data sharing of unfunded research at the institution. This fund could be built up based on all funding from successful grantees thus enhancing the sustainability for data sharing resources. Such a program could be in place for a limited period of time (e.g., 2 years) to develop the data-sharing core resource program.

There is no university or funder in Canada that can easily provide information on their proportion of data sharing. While mandates and policies are important, they are likely to have a greater chance of success if audit and feedback is provided. Developing and implementing an automated dashboard of OS practices will allow universities and funders to gauge and monitor their successes with implementing data sharing and other OS practices. Prototypes of such dashboards already exist (Charité Metrics Dashboard 2021).

Institutions may elect to have different immediate thresholds of achieving data sharing. For example, Institution A may set a low benchmark (e.g., 40%) after which they will institute a steep upward trajectory to reach 80% within two years. Alternatively, institution B may already have a series of educational and training modules shared across campus and have an expectation of a higher benchmark threshold (e.g., 65%). We believe it is critical for audit and feedback to be integrated into academic institutions and funders benchmarking data sharing and other OS practices. There are examples of how audits can enhance an entire system (Goldacre et al. 2018).

Meta-science has recently been highlighted as one of five science strategies for a better COVID-19 recovery by the Canadian federal government and the United Nations (United Nations 2021). Meta-scientists can facilitate a comparison of data sharing (e.g., FAIR principles) across and within disciplines, academic institutions, researchers, or funders to evaluate policy success. The recommendation to use standard content for education, the role of audit and feedback, and the integration of OS dashboards can be greatly enhanced through use of meta-science. There are few opportunities in the current Canadian funding structure to support researchers conducting meta-science. To facilitate this type of research we call upon Canadian federal and provisional funders to initiate specific calls for meta-science as has happened in Europe (Wellcome Trust 2018; Dutch Research Council 2020).

In the current Canadian landscape of publish or perish (Rice et al. 2021), researchers may feel pressured to extract as many papers out of their collected data as possible. There is now a strong movement away from this mindset. The Declaration on Research Assessment (DORA) strongly advocates against using these metrics when assessing researchers for career advancement (DORA 2021). About 2000 organizations, globally, have signed DORA, including 27 in Canada (e.g., University of Calgary; Lunenfeld-Tanenbaum Research Institute, Toronto; Tri-Agencies). Importantly, patients participating in clinical trials appear keen on having their data shared (Mello et al. 2018). This is an important group of stakeholders without whom clinical research is not possible. Key to the success will be incentivizing and rewarding data sharing as part of a researcher’s career advancement. Universities will need to integrate data sharing as part of promotion and tenure criteria. Cambridge University first developed the concept of data champions. These are typically researchers who volunteer to provide advice and assistance to the research community about ways to share their data. Champions themselves receive support and are managed by Cambridge’s Office for Scholarly Communication. The Technical University of Delft, the Netherlands, have taken this one step further. Data champions at their institution can now include their role providing data-sharing help as part of their promotion and tenure portfolio. We think building networks of data champions should be seriously considered within all faculties of medicines in Canada.

KDC is an academic representative on the Policy Committee for Research Data Canada and sits of the Advisory Board of DORA (Declaration on Research Assessment). DM leads the Hong Kong Principles initiative.