FAIR data pipeline: provenance-driven data management for traceable scientific workflows.

Modern epidemiological analyses to understand and combat the spread of disease depend critically on access to, and use of, data. Rapidly evolving data, such as data streams changing during a disease outbreak, are particularly challenging. Data management is further complicated by data being imprecisely identified when used. Public trust in policy decisions resulting from such analyses is easily damaged and is often low, with cynicism arising where claims of 'following the science' are made without accompanying evidence. Tracing the provenance of such decisions back through open software to primary data would clarify this evidence, enhancing the transparency of the decision-making process. Here, we demonstrate a Findable, Accessible, Interoperable and Reusable (FAIR) data pipeline. Although developed during the COVID-19 pandemic, it allows easy annotation of any data as they are consumed by analyses, or conversely traces the provenance of scientific outputs back through the analytical or modelling source code to primary data. Such a tool provides a mechanism for the public, and fellow scientists, to better assess scientific evidence by inspecting its provenance, while allowing scientists to support policymakers in openly justifying their decisions. We believe that such tools should be promoted for use across all areas of policy-facing research. This article is part of the theme issue 'Technical challenges of modelling real-life epidemics and examples of overcoming these'.

Following the science? Comparison of methodological and reporting quality of covid-19 and other research from the first wave of the pandemic.

BACKGROUND: Following the initial identification of the 2019 coronavirus disease (covid-19), the subsequent months saw substantial increases in published biomedical research. Concerns have been raised in both scientific and lay press around the quality of some of this research. We assessed clinical research from major clinical journals, comparing methodological and reporting quality of covid-19 papers published in the first wave (here defined as December 2019 to May 2020 inclusive) of the viral pandemic with non-covid papers published at the same time. METHODS: We reviewed research publications (print and online) from The BMJ, Journal of the American Medical Association (JAMA), The Lancet, and New England Journal of Medicine, from first publication of a covid-19 research paper (February 2020) to May 2020 inclusive. Paired reviewers were randomly allocated to extract data on methodological quality (risk of bias) and reporting quality (adherence to reporting guidance) from each paper using validated assessment tools. A random 10% of papers were assessed by a third, independent rater. Overall methodological quality for each paper was rated high, low or unclear. Reporting quality was described as percentage of total items reported. RESULTS: From 168 research papers, 165 were eligible, including 54 (33%) papers with a covid-19 focus. For methodological quality, 18 (33%) covid-19 papers and 83 (73%) non-covid papers were rated as low risk of bias, OR 6.32 (95%CI 2.85 to 14.00). The difference in quality was maintained after adjusting for publication date, results, funding, study design, journal and raters (OR 6.09 (95%CI 2.09 to 17.72)). For reporting quality, adherence to reporting guidelines was poorer for covid-19 papers, mean percentage of total items reported 72% (95%CI:66 to 77) for covid-19 papers and 84% (95%CI:81 to 87) for non-covid. CONCLUSIONS: Across various measures, we have demonstrated that covid-19 research from the first wave of the pandemic was potentially of lower quality than contemporaneous non-covid research. While some differences may be an inevitable consequence of conducting research during a viral pandemic, poor reporting should not be accepted.