Evidence synthesis · Lesson 2 of 3

The systematic review workflow

Criteria before searching, screening documented, exclusions counted. The process is what a reader is being asked to trust.

10 min read · Free — no account needed to read

Nine steps, in order, each documented. The documentation is not bureaucracy — it is the entire difference between a review and a reading list.

1. Define the question precisely, usually in a structured form naming the population, the intervention or exposure, the comparison and the outcome.

2. Write inclusion and exclusion criteria before searching. Which designs, which populations, which period, which languages, which outcomes. Writing them afterwards means they can be shaped by what you found.

3. Search, in several databases, with documented strings, plus grey literature and citation chaining. Record the date and the counts — a search is only repeatable if it is written down.

4. Deduplicate. The same study appears under several records, and in several forms.

5. Screen titles and abstracts against the criteria, then full texts. Ideally two people independently, with disagreements resolved by discussion. Record why each full text was excluded.

6. Extract data into a structured form — the same fields for every study, decided in advance.

7. Assess risk of bias in each study, using a tool appropriate to the design. This is how you weight, rather than treating every study as equal evidence.

8. Synthesise — narratively, or statistically if the studies are comparable enough to pool.

9. Report with a flow diagram showing how many records were found, screened, excluded and included, and why. That diagram is what lets a reader see the whole funnel.

Pre-registering the protocol before you start is standard in several fields, and it is what stops the criteria drifting towards the answer.

See it happen

Halfway through screening, you notice most included studies are from one country and want to exclude it for generalisability.

The read that arrives on its own

That is a reasonable methodological concern — add the exclusion and rerun the screening.

The same situation, with this idea

You notice the timing. Changing criteria after seeing what they do to your set is the same move as excluding outliers after seeing the result — it makes the review unable to come out the other way. The options that stay honest: keep the criterion and report the geographic concentration as a finding, or add a pre-specified subgroup analysis by country. Either is defensible; a silent mid-screening change is not.

After this, you can

notice when a criterion is being changed because of what it does to your set.

The takeaway

Criteria before searching, screening documented, exclusions counted. The process is what a reader is being asked to trust.

Easy to take the wrong way

That the flow diagram is a formality. It is the honest picture of how 4,000 records became 12, and where the 3,988 went.

Try this

Write your inclusion and exclusion criteria now, before searching, and date them.