Comparing study populations across papers is a matching task, not a reading task. You are deciding, paper by paper, whether the people (or samples) studied are similar enough that their findings can sit side by side in your synthesis, and you are recording the basis for that decision so a coauthor or reviewer can retrace it. The reliable way to do this is to fix your comparison dimensions before you read, extract population details into a structured table, and keep a method trail that shows how each paper entered the comparison.
This guide walks through that sequence with checkpoints, a worked example, a reusable table, and explicit completion criteria. It is written for researchers revising a manuscript, closing a project, preserving an evidence trail, or starting a new review.
Why does population comparison go wrong in the first place?
Most comparison failures happen before the reading begins. If you start from a pile of PDFs and no eligibility criteria, you end up comparing whatever each paper happened to report, which means different papers contribute different variables and the table fills unevenly. The Cochrane Handbook's guidance on searching and selecting studies starts from the opposite direction: plan the search from eligibility criteria, then screen in stages and link multiple reports of one study. Population comparison inherits that logic. Your inclusion rules define which populations belong in the comparison, and your extraction table records what each included study actually says about them.
A second failure is treating search results as a settled set. Searches need to be documented and, per Cochrane guidance, rerun before publication. If your population table was built from a search you cannot reconstruct, you cannot tell whether a newly found paper changes the comparison.
What should you decide before opening a single paper?
Fix four things in writing first. Each becomes a checkpoint you can return to.
- The comparison question. State the population contrast in one sentence, for example whether adult participants with a given condition differ across settings.
- Eligibility criteria for populations. Define who counts as in-scope and who does not, including age bands, settings, and diagnostic definitions where they matter to your question.
- The extraction fields. Choose the population variables you will record for every included paper, so the table has a consistent shape.
- The source and search record. Note which databases and interfaces you searched and how, so the set can be reproduced.
Checkpoint: a coauthor should be able to read these four items and predict which papers you will include. If they cannot, the criteria are still too vague.
How do you build a search you can defend later?
Build the search from the eligibility criteria rather than from whatever terms come to mind. Cochrane's guidance recommends combining free-text terms with subject headings and drawing on multiple sources. PubMed supports field tags, phrase searching, filters, Advanced Search, saved searches, and a Search Details view that shows how terms were translated and combined, which is useful when you need to record exactly what ran. Google Scholar documents author and title searching, date limits and sorting, Cited by and Related articles, citation export, and alerts, which helps when you are extending a set forward or backward from a known paper.
For the search itself, you can run queries in WisPaper search and keep the resulting papers together rather than scattered across browser tabs. If you prefer a public entry point that explains the scholarly search offering before you begin, the WisPaper Scholar Search page serves that purpose.
Checkpoint: save the query strings, the interfaces used, the date each search ran, and the number of records returned. This record is the spine of your method trail.
How do you screen papers without losing the population thread?
Screen in stages, as Cochrane recommends: deduplicate records first, then screen titles and abstracts, then full texts. At the full-text stage, extract population details immediately rather than deferring them, because that is the moment you have the methods section in front of you.
Two screening habits protect the comparison:
- Link multiple reports of one study. A single study may appear as several papers. Cochrane guidance calls for linking these so one population is not counted twice.
- Record exclusion reasons at the population level. "Wrong population" is more useful than "not relevant" when you later need to justify the boundary of your comparison.
When you need to ask questions across a set of papers you have already gathered, Library QA lets you query papers uploaded to My Library or added from search results, which can speed up the first pass over methods sections before you extract details by hand.
Checkpoint: every excluded record has a reason, and every included study maps to exactly one population row (or a clearly noted set of linked reports).
What belongs in a population comparison table?
Use one row per study population and one column per extraction field. The table below is a reusable template; adapt the fields to your question but keep them fixed once extraction starts.
| Field | What to record | Why it matters for comparison |
|---|---|---|
| Study ID and linked reports | Identifier plus any companion papers | Prevents double counting |
| Source and search date | Interface, query, date run | Makes the set reproducible |
| Population definition | Who was studied, as stated | Anchors the comparison |
| Setting and recruitment | Where and how participants were identified | Explains population differences |
| Key inclusion/exclusion rules | Study-level eligibility | Shows whether boundaries align |
| Reported characteristics | The variables your question requires | Fills the comparison consistently |
| Extraction note | Who extracted, from which section | Supports the audit trail |
Keep the table in a shared location so coauthors extract against the same schema. If you are organizing the underlying papers, My Library lets you save or upload papers and keep them organized alongside the extraction work.
Checkpoint: no cell is silently blank. A blank means "not reported," and you should mark it as such rather than leaving it empty.
Worked example: comparing populations for a manuscript revision
Suppose a reviewer asks whether your discussion overstates similarity across the studies you cite. Here is the sequence.
- Restate the comparison question in one sentence tied to the reviewer's concern.
- Write population eligibility criteria and confirm them with a coauthor.
- Rebuild the search record from your original queries, and rerun the searches as Cochrane recommends before publication. Note any new records.
- Screen the new records in stages, linking multiple reports of the same study.
- Extract population fields into the table for every cited study, marking unreported fields explicitly.
- Compare column by column, not paper by paper, so differences in setting or eligibility surface across the whole set.
- Revise the manuscript claims to match what the table supports, and cite the table in your response to the reviewer.
If the revision grows into a larger literature task, Scholar Agent is the WisPaper area for longer research tasks with literature review, inspiration, or reproduction use cases. For a project that needs its own workspace, Research Projects is where you create and manage a project.
Checkpoint: the reviewer's concern can be answered by pointing to specific table rows, not by re-reading the papers.
How do you keep the method trail auditable?
An auditable trail has four parts: the search record, the screening log with exclusion reasons, the extraction table with extraction notes, and a version note each time any of these changes. Cochrane's emphasis on documenting the search process and rerunning searches before publication maps directly onto this structure.
Before you submit or hand off, check your reference list against real databases. TrueCite screens BibTeX citations against real academic databases to flag hallucinated references, which is a concrete final step before a manuscript or handoff leaves your hands.
Checkpoint: a colleague who has never seen the project can reconstruct your included set and your population table from the trail alone.
When is the comparison actually finished?
Use these completion criteria as a checklist. The comparison is ready when all of the following hold:
- The comparison question and population eligibility criteria are written down and unchanged since extraction began (or changes are versioned).
- The search record lists interfaces, queries, and run dates, and searches have been rerun before publication.
- Every included study has one population row, with linked reports noted.
- Every excluded record has a recorded reason.
- Every table cell is filled or explicitly marked as not reported.
- Manuscript claims match what the table supports, with no claim broader than the extracted fields.
- The reference list has been screened for citation integrity.
If any item fails, the comparison is not finished, regardless of how complete the prose feels.
How do you hand the comparison to a coauthor or successor?
A handoff is a test of the trail. Package the search record, screening log, extraction table, and version notes together, and write a short cover note stating the comparison question and the current status of each checkpoint. Point the recipient to the rows that support each manuscript claim. If the project lives in a shared workspace, keep the package there so the next person works from the same files rather than a copy.
Checkpoint: the recipient can answer "why is this paper included?" and "why is this one excluded?" without contacting you.




