Once you have screened the full-text articles to decide which will be included in the final review, the next step involves thoroughly reading the full text of each included article again and extracting relevant data using a standardised data extraction form. As with other steps of the review the decision-making and process should be well documented to enable full reporting.
The objective of the data extraction phase is to collect relevant information from studies objectively and accurately in a consistent format to make use of the data in future stages of the review. The extracted data will be presented in the review as a summary table or summary of findings table and described in the narrative.
It is recommended that at least two reviewers independently extract data from each study, with a clear process in place to address any discrepancies that may arise and that extraction should be piloted to ensure all reviewers are recording similar data and the template used is appropriate. Once all reviewers have finished their extraction, they meet to compare their extracted data and create a final single data set.
The data extraction of included studies should adhere to established guidelines. Information such as title, author, year, journal, research question, specific aims, conceptual framework, hypothesis, research methods or study type, and conclusions may need to be extracted from each included study and well as effects sizes, population charactertistics, etc., based on the purpose of the review. It is important to carefully review the methodology to categorise studies by type in the results section of the review. Additionally, for any intended meta-analysis, raw and refined data should be extracted from each study result.
Chapter Nine of the Cochrane Handbook explains how to summarise study characteristics and prepare for synthesis.
Covidence's downloadable manual A Practical Guide: Data Extraction for Intervention Systematic Reviews provides definitions, advice, real-world examples, links to the Cochrane Handbook and downloadable templates. The manual is tailored for use in intervention or effectiveness systematic reviews.
If you are using Covidence to design and populate your data extraction form you have a choice between using the Data Extraction 1 and Data Extraction 2 tools. A link below in the "Templates and reporting" section of this chapter aids decision making in which one will be better suited for your project. The extraction workflow in Covidence ensures each reviewer works independently to extract data and then brings both data sets together for a consensus stage, creating one agreed data set. Extraction 2 enables a single reviewer extraction process. Data can then exported in a variety of formats for further analysis.
Extraction 1 and Extraction 2 both include a quality assessment template (E1 has the Cochrane Risk of Bias v1 template as default and the E2 template is fully customisable) making it easier to do the critical appraisal of each study at the same time as extraction.
The template or form you use for recording extracted data will depend on the study methodology and data type - a form designed for data in RCTs will not be suitable for a non-interventional review. While certain information about studies will always be included such as title, authors etc it is expected any template used may be customised according to the needs of all the data being collected.
The PRISMA checklist requires that reviewers:
“Specify the methods used to collect data from reports, including how many reviewers collected data from each report, whether they worked independently, any processes for obtaining or confirming data from study investigators, and, if applicable, details of automation tools used in the process.”
“List and define all outcomes for which data were sought. Specify whether all results that were compatible with each outcome domain in each study were sought (for example, for all measures, time points, analyses), and, if not, the methods used to decide which results to collect.”
This article is an explanation and elaboration of PRISMA 2020 and gives details of how they expect you to conduct and report this (items 9 and 10.)