Frame a question you can actually answer
A systematic review starts with a narrow, answerable question. A structure such as population, intervention or exposure, comparison and outcome (PICO, or a variant that suits qualitative or policy topics) keeps the scope tight. If you cannot state the question in one sentence, the search will drift.
Write the protocol first
Decide inclusion and exclusion criteria, databases, date limits, languages and the screening method before you look at results. Writing this down in advance, and registering it where possible, protects the review from choices that are made to fit the findings.
- Eligibility: study design, population, setting, outcomes.
- Sources: which databases and which other routes, such as reference lists.
- Methods: how many reviewers, how disagreements are resolved.
Search and record every step
Build the search string with the key concepts, synonyms and controlled vocabulary, adapt it for each database, and save the exact string and the date of each search. Export all records into a reference manager and remove duplicates, keeping a count at each step.
Screen in two passes
First screen titles and abstracts against your criteria, then read the full text of what remains. Record a reason for every full-text exclusion. Where more than one reviewer is involved, screen independently and agree rules for resolving conflicts.
Keep your numbers consistent. The counts in the text, the flow diagram and the tables must match exactly. Reviewers check this first.
Extract data and judge quality
Use a structured extraction form, test it on a few studies, then apply it to all. Assess risk of bias with a tool that fits the study designs you included, and explain how that assessment shaped your conclusions rather than just listing scores.
Report so that others can follow
Follow a reporting guideline such as PRISMA. Include the flow diagram, a table of included studies and a clear statement of limitations. Synthesise findings by theme or outcome rather than study by study, and keep the conclusions within what the evidence supports. State the review’s dates, so readers know how current it is.