Systematic Reviews in the Health Sciences

There are three levels of screening and data analysis – title and abstract screening, full-text screening, and finally, data extraction or study content coding.

Title and abstract screening is the first cut in the screening process. It’s where two trained raters independently screen the titles and abstracts of all the articles they’ve found through their literature searching using the inclusion and exclusion criteria identified in their protocol. If the articles match the inclusion criteria, they’ll move on to the second phase; if they don’t, they’ll be weeded out.

The second level is full-text screening or the second cut. This is very similar to title and abstract screening, but this time the full-text articles are being reviewed, where the reviewers are reading the articles. You must document reasons for excluding any articles at this point because you’ll need to include those reasons in your PRISMA flow chart. 

Once you’ve selected the studies to be included in your review, you’ll want to actually extract the data or code the content. Because you want consistency in the data you extract from each study, you should develop a data extraction form that’s been adapted to your question; provide instructions, decision rules, and training on using the form; and pilot the form with a selection of studies to make sure the same information is being drawn out from each one. This will allow you to be objective throughout the process. 

It’s possible that some studies will not report certain data that you’re interested in. This missing or unclear information should be reported in your review.

For Title & Abstract Screening and Full Text Screening, the library recommends using Covidence for these two steps. Covidence keeps track of each reviewers’ responses and blinds reviewers to their own and each other’s responses. You can also pin your inclusion and exclusion criteria to the top of the screen so that it’s at the forefront of your mind in your decision making. Covidence also allows you to document your reasons for excluding articles within the program instead of having to make separate notations.

If you’re wondering what categories you should be including in your data extraction form, here are some sample categories from the Centre for Reviews and Dissemination. It includes things like study design, participant characteristics, and details on the intervention. But these are samples and your categories could be different. 

For other coding templates:

• Check systematic review organization websites
• Adapt a Cochrane Collaboration template
• Google
  • "data extraction form/template” and “Cochrane” or other SR organization
  • "study eligibility/selection checklist” and “Cochrane” or other SR organization
• Review published systematic reviews on your topic

Data Extraction Software: 

• Excel, Google forms
• Covidence (free to UCalgary affiliates)
• NVivo (free to UCalgary affiliates, used for qualitative data analysis, regular workshops scheduled at HSL)
• RevMan (free for those completing Cochrane reviews, or when used for purely academic purposes)
• JBI Sumari (free)
• EPPI-Reviewer (subscription-based pricing)
• DistillerSR (subscription-based pricing)