Open almost any systematic review indexed in PMC and you will find the same figure near the introduction: a chain of boxes narrowing from thousands of records down to a handful of included studies. That figure is the PRISMA flow diagram, and it is not decoration. It is the single fastest way to judge whether a review’s conclusion rests on a defensible search or on a convenience sample the authors happened to like. For a desk that reads global-health literature for a living, learning to read this diagram properly is as basic a skill as checking a DOI.
What PRISMA Actually Is
PRISMA stands for Preferred Reporting Items for Systematic Reviews and Meta-Analyses. It is a reporting guideline, not a research method: it does not tell an author how to run a review, only how to document one so a reader can audit the work. The guideline is maintained by the PRISMA Group, an international collaboration of methodologists and journal editors, and the current version, PRISMA 2020, was published in 2021, replacing the original 2009 statement. Journals that require PRISMA compliance for systematic reviews expect the flow diagram as a mandatory figure, which is why it appears so consistently across publisher platforms and PMC full-text records.
The Four Stages, and What Each One Verifies
The flow diagram is built around four sequential stages. Each one answers a different question about how trustworthy the final study set is.
- Identification. This box reports the total number of records located across all databases and other sources, before anything is removed. It is the only place a reader can check whether the search was actually broad — a review that searched a single database is a different, weaker exercise than one that searched several plus trial registries and reference lists, even if both end with the same number of included studies. Duplicate records are typically removed at this stage and the count is reported.
- Screening. Titles and abstracts of the deduplicated records are checked against the review’s eligibility criteria. Most records are excluded here, and the diagram should report how many, though not always the individual reasons — that detail usually waits for the next stage.
- Eligibility. The full text of every record that survived screening is retrieved and assessed in detail. This is where exclusions get itemized by reason — wrong population, wrong outcome, wrong study design, full text unavailable — which is the part of the diagram most worth reading closely, because it shows the judgment calls the authors made.
- Included. The final count of studies that met every criterion and were carried into the review’s qualitative synthesis, and into the meta-analysis if one was performed. This number is what gets quoted in abstracts and headlines, but it is meaningless without the three boxes that preceded it.
Why the Gaps Between Boxes Matter More Than the Boxes
The habit worth building is reading the arithmetic between stages, not just the final total. A review that starts with several thousand identified records and ends with four included studies is not necessarily weak — that is normal for a narrow, well-defined question. What is worth flagging is a steep, unexplained drop at eligibility with vague exclusion reasons, or an identification count that looks implausibly small for a topic with a large published literature. Both are signals to open the full methods section rather than trust the diagram at face value. The diagram is a summary of decisions, and summaries can still be misleading even when every number in them is accurate.
Where This Fits an Evidence Desk’s Reading Habits
This desk exists to point readers past a headline finding and toward the sourcing behind it — PMC records, WHO guidance documents, publisher DOIs. A PRISMA flow diagram is one of the fastest checks available before citing a systematic review: if a paper claims to synthesize "the evidence" on a global-health question but has no flow diagram, or a threadbare one with no eligibility detail, that absence is itself informative. Reviews published without PRISMA reporting are not automatically wrong, but they ask a reader to trust the author’s search process on faith rather than showing it.
Conclusion
A PRISMA flow diagram compresses an entire literature search into four boxes, and each box is answering a specific, checkable question: how broad was the search, how much was screened out, why were full texts excluded, and what actually survived. Reading past the final included-studies number to the path that produced it is a small habit that catches a surprising number of overstated conclusions before they get repeated as fact.