Covidence: Automation

Covidence has several features that use machine learning (ML) and large language models (LLMs) for automation, including relevance sorting for screening, the Cochrane RCT classifier, removing non-RCT references before screening, and suggestions for extraction when using the Extraction 1 template. More information on automation within Covidence is available below:

• Covidence's approach to automation
• From manual to machine: How Covidence's ML is streamlining systematic reviews
• Machine learning - the gamechanger for trustworthy evidence

Automation features using ML and LLMs can be disabled by the user (see FAQ: Turning automation features on and off)

Reporting the Use of Automation Tools

Note that the PRISMA-2020 requires reporting the details of automation tools used, including in the selection and data collection processes. The PRISMA Flow Diagram also requires the reporting of the number of records excluded by automation tools.