Data sharing in Epidemiology

This past Monday, I was invited to speak on technical issues surrounding data sharing at the Wellcome Trust, in coordination with the WHO's Health Metrics network. The meeting - "Code of Conduct for the Collection, Analysis and Sharing of Health Related Research Data in Developing Countries" - brought together public health researchers from the developed and developing world, social scientists, demographers, representatives from the Trust, members of the Health 8 (H8), and other experts (me being one?). As the title suggests, the Trust is looking to craft a draft code of conduct for data sharing in developing countries, to hopefully be presented at their next meeting in Bamako, Mali this November.

Data sharing is an issue near and dear to my heart, and a key focal point of our work at Science Commons. With that said, the majority of the conversations we engage in with the community on this topic don't traditionally to date include social scientists, or at least, not as much as those in the life sciences. It offered an interesting perspective to me, with a distinct cultural barrier and set of blocking factors.

Out of the entire day, there were a few main points of agreement that emerged:

1. the need for capacity building, first and foremost - especially if looking at implementing this code of conduct in developing countries. This includes trainings, money invested in the infrastructure, etc.

2. the agreement that in terms of ethics - it is unethical to not share data if doing so can help a community or assist public health research. (a very powerful statement for an audience that many self-ascribed as "dinosaurs" when it came to data sharing policy and practice).

3. the need for standardization - on a social level: to share funds, resources, human power, skills and training; as well as technical standardization in terms of metadata, common languages and formats that allow for interoperability.

4. speaking of metadata - there was considerable agreement that such standards used for the social sciences, specifically DDI, should be built into the code for all forms of data. I'm not overly familiar with DDI, but worth exploring ...

5. we need to lower transaction barriers at whatever cost is necessary, as is allowed. there are a number of issues - privacy and confidentiality among those - that call for complete public disclosure to be slightly modified, but the intent to share is there.

6. we are a long way off from fully accepting this paradigm shift and there is a significant cultural barrier to sharing data / putting data in the public domain. that reality can't be glossed over, and needs to be recognized.

Another part of Monday's conversation revolved around "access" - as was expected. :) With such a variety of stakeholders, the discussion kept coming back to a hierarchy of access, along a spectrum from an embargoed access, to closed access, to fully "open access" data (which I tried to change to describing as "putting data in the public domain", but c'est la vie).

Back to capacity building, one participant (under Chatham House rules, so this is intentionally not attributed to anyone) posited that data management needs to move away from a stigma of being a clerical position to being its own scientific enterprise, to attract more skilled professionals to the field. This reminded me of conversation around repository managers moving from just an administrative, grunt work position to actually receiving support and acknowledgment.

All in all, the proposed code of conduct draft needs some work, and this is an iterative process, still gathering feedback from policy makers, think tanks, scientists and public health officials.

It does raise some interesting food for thought, though in regards to data sharing. What is the best approach, if targeted towards developing countries? Should there be a distinction? A case can be made (and I often tend to make) that this is more of a global issue to address, than developing vs. developed. With that said, should this be honed down and target-specific to allow for quicker implementation or be crafted as a higher level document that could be a living, breathing recommendation?

We shall see. Gives me a lot to noodle on though, that's for sure.