The PRISMA Flow Chart is a process showing how you searched for and then filtered down your search results. This process is used when conducting a systematic review, when you want your search process to be transparent. The general process looks something like this:
- Search for articles, and then remove any duplicates.
- Look at the articles titles and abstracts, exclude articles based on your screening criteria (you decide what this criteria is).
- Retrieve the full text of your remaining records. If you are unable to get the full-text of any articles, you would document this.
- Read your remaining full-text articles, and then run them through your exclusionary criteria again.
- The remaining articles are the ones you will be using for the study.
So let's break down the flow chart, line by line.
For the Identification section, you will run your search using your search strategy (including any keywords, subject headings and synonyms)
BOX 1
- Databases "n" = the number of reports you found in databases.
- Registers include clinical data registries -- these have information about studies, but might not include a traditional full-text article (like you would find in a database). An example of a registry would be ClinicalTrials.gov.
BOX 2
- For the Records removed before screening section, you'll want to remove any articles from your list that are automatic "no's." You will remove any duplicates from your results, and removing articles from consideration by using filters (i.e., articles that are over 5 years old, are in a foreign language, specific study types, etc.); you won't need to dig into the article abstract yet (that comes later). If you are using filters, this can be documented in the "records marked as ineligible by automation tools" area.
For the Screening section, you will be digging further into the records to decide whether it is worth keeping.
BOX 3
- Records screened refers to the new number of results you are working with; this is the number of records minus the records excluded in the last step.
BOX 4
- In the Records excluded step, you review the abstracts of each citation and think about what factors would make an article ineligible for your research question.
- Some examples of exclusionary criteria could include: studies that upon closer investigation are covering the wrong population, are outdated, not peer reviewed, are written in a foreign language, are not the correct study type or have a poor design, etc.
- These exclusionary factors are up to you to decide, and should be based on your specific research question.
BOX 5
- Records sought for retrieval is your new total (records screened minus the records you excluded). This is your list of eligible studies. At this point, you will need to retrieve and read the full text of each article in your list.
BOX 6
- Reports not retrieved is the total number of articles in which you are unable to get the full text for your review.
BOX 7
- Reports assessed for eligibility is the new list you have (Reports sought for retrieval minus Reports not retrieved).
BOX 8
- After reading the full text of the articles, you will list your reasoning for any final exclusions in the Reports excluded section. Again, the exclusionary criteria is up to you to decide, and is based on the question you are asking.
The last section (Included) refers to the total number of articles you will be using.
BOX 9
- This last section is the total number of results you've selected after all of your exclusions. This section also allows you to make a note if the number of studies (trials) you selected differs from the number of total reports you selected. A "report" is defined as "a journal article, preprint, conference abstract, study register entry, clinical study report, dissertation, unpublished manuscript, government report, or any other document providing relevant information."(2) If you searched registries, you might have found trials that don't include a report, so you would use this box to make a note of this.
For example, if you found 15 studies, and two of them were studies found in a registry that didn't include a report, then your box would include this information:
- Studies included in review (n=15)
- Reports included in review (n=13)
