Open Access News

On May 8, Elizabeth Pisani released the first draft of the Bamako data sharing code of conduct.

The code arose from last year's Global Ministerial Forum on Research for Health (Bamako, Mali, November 17-19, 2008), where participants formulated the Bamako Call to Action on Research for Health, which included a call for "open and equitable access to research data, tools, and information...."  For more background, see Pisani's slide presentation at the November 2008 meeting on the need for a data sharing code of conduct, and a report on the the discussion following Pisani's presentation. 

From the May draft code:

...What is driving the exponential growth in knowledge in areas such as genetics, astrophysics, information technology? Data sharing....

Epidemiology and public health have been left behind in this data sharing revolution, mired in a culture that restricts access to data and information. This is in part because of a perceived need to protect the privacy of individuals involved in research. But public health is a public good; in public health research there’s an ethical imperative to use information gathered from individuals to benefit the greatest possible number of people. Public health deserves to advance at the same speed as genetics, where data sharing has led to an explosion of progress. The World Health Organisation and several funders of public health research, led by the Wellcome Trust, are thus supporting the development of a code of conduct to encourage greater sharing of public health data. The code seeks to provide guidance for funders of data collection and for institutions that collect and analyse data, including those who perform secondary analysis on data collected by other people. The principles espoused by the code are universal....

The draft code presented here is the product of initial discussions between epidemiologists and data managers from all continents. They gathered with a number of representatives from governments, international organisations and major funders of public health research in London on October 6th, 2008 to agree on the core principles in the code. The discussions of this Working Group were informed by a background paper which reviewed the major challenges to more open exchange of public health data, challenges that can be categorised broadly as incentive-related, capacity-related, ethical and technical. The draft code is structured around these four areas. The background paper has been updated to reflect the outcome of the meeting, and is appended here....

A code of conduct on data sharing is an important first step in striking the balance between the advancement of science and the rights and needs of individuals and communities....

To the extent possible, the code promotes the sharing of micro-level data -- that is, individual level records. There may occasionally be reason to restrict access to individual level data. There is rarely any reason at all to restrict access to aggregated data....

We support the maximum public access to data of public health importance compatible with the following principles:

• The protection of privacy of individuals from whom data are gathered
• Fair reward for the work of data collectors and primary investigators
• Maximum public health benefit delivered in a reasonable time frame....

Limited-time exclusive access for primary researchers

Data are available to the research team involved in data collection and their institutional partners for a fixed period (between six and 18 months) before they are shared. This allows the research team a head start on data analysis and publication....

Following a period of exclusive access for primary researchers where necessary, the most common levels for access to data of public health importance will be:

Fully open access

Data (anonymised where necessary) are made available in machine-readable formats on publicly-accessible websites. This is most desirable and should be encouraged where feasible and compatible with privacy....

Controlled public access

Data are made available to authorised users after a screening process. This is likely to be the most common form of access for data of public health importance....

Collaborative access among scientists

Data are made available to other scientists in a collaborative network. Collaborative access may be necessary for complex datasets that include sensitive information where anonymisation is difficult (e.g. longitudinal data sets including HIV status)....

Exclusive access for primary researchers

Data are only available to the research team involved in data collection and their institutional partners. This is currently the norm in public health data collection, but it is precisely this norm that the current code seeks to change. There are few cases in which this degree of exclusivity is necessary in the long term....

Increasing the incentives to share data

Under the Code of Conduct on Data Sharing we agree to:

Put past data sharing performance on a par with publication as a criterion for evaluating the performance and job suitability of scientists, as well as evaluating grant proposals.

Reward concrete plans for data sharing when evaluating funding proposals for research and routine health systems functions such as surveillance.

Develop citation standards and indices for shared data sets; commit to using them when publishing secondary analysis.

Require registration of public-health related research and data collection in open access data-bases to facilitate data discovery and create demand for shared data.

Encourage submission of micro-data to public repositories as a condition for journal publication of research results.

Promote a “creative commons” approach, in which derived datasets and secondary analysis files based on shared data are in turn made publicly available.

Support an ombudsman system to oversee the fair use and proper acknowledgement by secondary users of shared data....

Using technology to increase data sharing

Under the Code of Conduct on Data Sharing we agree to:

Commit to a single metadata standard for datasets of public health interest....

Ensure that metadata are open access and machine-readable, even for data that are shared under the controlled or collaborative access standards.

Support the development of “open source” software for management, documentation and analysis of public health data....

Taking the code forward...

In trying to meet the needs of [a] huge and varied constituency, the current draft code is vague: phrases such as “promote x” and “encourage y” predominate. As the code develops, we hope that it will become more concrete: “Funding institutions commit to investing in x”, “Secondary analysts agree to provide y”....

Labels: Hot