What makes a good systematic review?

In general, a good systematic review will include the following basic components:

  • A focused question – such as: ‘Are parenting interventions effective for reducing adolescent antisocial behaviour?’ (as opposed to an unfocused question, such as ‘Are parenting interventions effective?’) or ‘Do teen pregnancy prevention programmes reduce pregnancies among girls in urban secondary schools?’ (an unfocused question would be, for example, ‘How can we prevent teen pregnancy?’).
  • Inclusion/exclusion criteria – which studies the authors included or excluded, and why. For example, a review may include only randomised studies, because that is the most reliable method. Authors need to provide a logical reason for including or excluding studies.
  • Search strategy – the reviewer must clearly define how he or she went about finding relevant studies; this should include searches for both published and unpublished reports of trials, and may involve online database searching as well as contacting subject-matter experts to find unpublished reports.
  • Study selection – to determine which identified studies fit the criteria for the review, and collect relevant data for those studies.
  • Assess quality of studies – helps to identify risks of bias in studies.
  • Synthesise study results – if the included studies are similar, the author can determine the overall effectiveness of an intervention using meta-analysis (see below); if the studies are not very similar (e.g. if they measure different things, or if some trials are randomised and some are not), then the author may provide a narrative analysis, simply describing the results.
  • Analysis and limitations – the author should discuss what the results mean for the area of study being addressed, including how the results might change what is already known about a subject. There should also be a description of the limitations of the study.

Some systematic reviews also include meta-analyses, which provide a good measure of the overall effects of the intervention that is being tested. Read more about Meta-analysis.