Data publishing best practices

---

Data supporting the results of a systematic review should be made available to the users of the evidence (and this is a requirement for Cochrane reviews). Publishing the data in the form most likely to be useful to users can amplify the impact of your review.

This page outlines best practices for data publishing using RevMan. Expect this to be expanded and refined over time. For suggestions on how to improve this page, or if you have any queries, please contact us at support@cochrane.org. 

Reporting intervention review data 

Capturing the underlying included study data using study centric data enables maximum flexibility for users to re-analyze the data according to their needs, as it is more representative of how the studies were reported.

When study centric data has been used, RevMan will automatically include both the study results and the analysis data tables in the data package. Whenever possible, avoid manual transformations of the data, and rely on the transformations that RevMan can apply automatically.

When some analyses are not possible using study centric data (see When should I use study centric data management?), use manual input for those analyses only, and automatic analyses for the remaining ones. 

Reporting network meta-analysis data 

Network meta-analysis datasets are incredibly valuable, and the key study data informing the analysis should be included in the review.

Even though RevMan currently does not have the capability to perform a network meta-analysis, it is possible to accurately capture the underlying data using study centric data. This is the recommended method for reporting the data. Forest plots can be included for key comparisons if they convey important information to the reader. 

Ideally, if a data extraction tool like Covidence has been used, the outputs of the data extraction process can be imported into RevMan as for any other review.

However, in some cases it may be desirable to convert data tables that have been used for the analysis software for import into RevMan. Such data tables can be in either “long format” (one row per arm) or in “wide format” (one row per study). The long format is straightforward to adapt to fit into the RevMan import templates, while the wide format will first need to be converted to long format.

If you have performed analyses at different levels of granularity of the interventions (lumping/splitting), import the most granular definition used into RevMan and use “intervention groupings” to define the less granular levels. Any properties of the studies that have been used as effect modifiers or to exclude some studies from certain analyses, should be imported as covariates. 

Reporting for other review types 

For other review types, use the data management features provided, and consider how users may be able to re-analyze the data. If the data presented in RevMan are insufficient, consider publishing additional data files and citing these in your data availability statement. 

Reporting data not managed in RevMan 

Currently, RevMan does not allow data files in arbitrary formats to be attached to reviews. Instead, use an external data sharing service such as Zenodo, Dryad, or Figshare.

Make sure to use a service that assigns a permanent identifier (e.g. DOI) to allow you to cite the dataset and the user to retrieve it. Further guidance on including external data and files with Cochrane reviews is available in our author guidance.

If you’re publishing data as external files, pay particular attention to the file formats used and avoid overly complex formats or formats that are tied to specific tools. For example, CSV (simple, open) should generally be preferred over Excel (complex, proprietary) for publishing data.

Consider aligning the formats you use with the formats used in the RevMan data package. When formatting your data as CSV files, this external guidance may be helpful.