Covidence Extraction 2.0 Offers More Flexibility

In the last few months, the Covidence systematic review software (Veritas Health Innovation, Melbourne, Australia) has updated its data extraction module, offering users additional flexibility in the case of many of its features in Extraction 2.0. At least for the time being, users still have the ability to revert from the now 2.0 default back to Extraction 1.0 in the Settings if they so choose.

What’s involved in Systematic Review data extraction, in general?

Before discussing the Covidence Extraction 2.0 changes, it’s good to review some best practice standards for data extraction. The standards related to data extraction are mainly  intended to provide quality control to the extraction process by creating an environment where errors of omission and errors of inaccuracy can be minimized. From a PCORI presentation:

Relevant IOM Standards for Data Extraction
Standard 3.5: Manage data collection

  • At a minimum, use two or more researchers, working independently to extract quantitative and other critical data from each study
  • For other types of data, one person could extract data, while a second person independently checks for accuracy and completeness
  • Establish a fair process for resolving discrepancies (do not give final decision making to the senior reviewer)
  • Link publications from the same study to avoid including data more than once
  • Use standard data extraction forms developed for the specific review
  • Pilot-test the data extraction forms and process

Added flexibility and customization in Extraction 2.0 

Covidence Extraction 1.0 was best suited for systematic reviews where quantitative data would be extracted. As such, the default setting for the number of extractors was automatically locked in to two reviewers and could not be changed (in the way that the Title/Abstract Screen and Full Text Screen stages could be set to just one reviewer). Extraction 2.0 has made it possible to proceed with just one person carrying out the extraction. If two people are set to extract, consensus checking is still required. The comparison step, however, can be carried out as soon as both reviewers are done, at the individual study level.

Another noteworthy difference with Extraction 2.0 is that users now have the ability to begin designing the extraction form as soon as the Covidence review is created, even before any references have been added to the project. In other words, the extraction templates can be set up separate from the studies, something which benefits the process of pilot-testing the form/template. Also, whereas previously the addition of quantitative intervention outcome data automatically was configured into a table format and strictly followed a PICO framework, Extraction 2.0 recognizes that systematic reviews can vary and is less table intensive and more customizable in terms of the broader range of information that can be captured.

There are a couple of areas worth mentioning that have been intentionally left inflexible (most likely in the interest of better data quality/integrity). There is currently no way to restrict who on the team can be an extractor or do consensus – it is done by whoever gets there first. And once the data extraction has begun in Extraction 2.0, assigned roles cannot be re-assigned (so basically, only two people from the team can be involved in the data extraction stage). Also, once data extraction has begun, publications from the same study can be merged, however, this step cannot be undone.

Clearly, the Covidence Extraction 2.0 updates are still very much in line with the Standards for Data Extraction set out by the IOM (listed above). To learn more about the systematic review process and how to use Covidence, be sure to check out the MSK Library’s upcoming workshop schedule or Ask Us!