November 9, 2026

How to Record Why Studies Were Excluded from a Review

When you finish screening, the number that matters is not how many records you removed but whether you can explain each removal. A defensible exclusion record lets a reader retrace your decisions, lets a supervisor or re

Written byWisPaper TeamAI Research Workflow Team
Editorial illustration for How to Record Why Studies Were Excluded from a Review

When you finish screening, the number that matters is not how many records you removed but whether you can explain each removal. A defensible exclusion record lets a reader retrace your decisions, lets a supervisor or reviewer check them, and lets you pick the project back up months later without re-reading everything. This guide walks through the decisions and artifacts that make that record, using a winter-break cleanup of a thesis chapter as the running example.

What a usable exclusion record actually contains

An exclusion record is not a single list. It is a small set of linked artifacts: the eligibility criteria you screened against, the search and deduplication trail, a screening log with a reason for every excluded record, and a way to connect multiple reports of the same study. Cochrane's guidance on searching and selecting studies frames these as one documented process — plan searches from eligibility criteria, use multiple sources, deduplicate records, screen in stages, and link multiple reports of one study. Your exclusion log is the visible output of that process.

The practical test is simple: could a colleague who has never seen your project take one excluded record and understand, from your files alone, which criterion it failed and at which stage you decided that?

Diagnose your starting state before you build anything

Most readers arrive here in one of a few situations. Naming yours tells you which artifact to build first.

  • You screened informally and kept no reasons. You have a folder of PDFs and a vague memory of why some were dropped. Start by writing the eligibility criteria down, then re-screen against them.
  • You have reasons but they are inconsistent. One record says "wrong population," another says "not relevant." You need a controlled reason vocabulary before the log is worth anything.
  • You have a clean log but no link between stages. You cannot tell whether a record was removed at title screening or full-text screening. Add a stage column.
  • You are restarting after a break. The files exist but the logic is gone. Reconstruct the criteria and reason codes first, then audit a sample of decisions.

In the winter-break example, the reader has a thesis chapter with 40 full-text PDFs, a rough list of 12 that "didn't fit," and a supervisor comment asking why three specific papers were dropped. The starting state is the second and third bullets combined: reasons exist but are inconsistent and unlinked to stages.

Write the eligibility criteria as testable statements

Exclusion reasons only work if they point back to criteria. Before logging a single exclusion, turn each criterion into a statement you can apply to a title, an abstract, or a full text.

A workable pattern is to state the population, the intervention or exposure, the outcome, the study design, and any date or language limits as separate lines. Cochrane's guidance ties search planning directly to eligibility criteria, which is why this step comes before screening rather than after. If a criterion cannot be phrased so that two people would usually agree on whether a paper meets it, it is not yet a criterion — it is a preference.

Checkpoint: read your criteria list and ask whether each line could generate a specific exclusion reason. If a line cannot, rewrite it until it can.

Build a controlled reason vocabulary

Free-text reasons drift. "Wrong population," "different sample," and "not our group" may all mean the same thing, and a reader cannot tell. Fix this by defining a short, closed list of reason codes, each mapped to one criterion.

A reusable structure looks like this:

CodeReasonLinked criterionStage where applied
E1Population outside scopePopulationTitle/abstract or full text
E2Intervention or exposure not studiedInterventionTitle/abstract or full text
E3Outcome not reportedOutcomeFull text
E4Study design not eligibleDesignTitle/abstract or full text
E5Outside date or language limitLimitsTitle/abstract
E6Duplicate or linked report of an included studyLinkingDeduplication or full text

Keep the list short enough to memorize. If you find yourself inventing a new code mid-screening, stop and decide whether it belongs to an existing criterion or whether your criteria need revising — and if they do, note the revision and its date.

Separate screening stages and record the stage for each exclusion

Cochrane describes screening in stages, and your log should mirror that. A record excluded at title screening for an obviously wrong population is a different decision from one excluded after reading the full text and finding the outcome missing. Recording the stage does two things: it makes your counts auditable, and it stops you from over-claiming certainty about a paper you only skimmed.

A practical stage sequence:

  1. Deduplication. Remove duplicate records and note the method. Cochrane recommends deduplicating records as part of the selection process.
  2. Title and abstract screening. Apply only the criteria that can be judged from a title or abstract. Log the reason code and the stage.
  3. Full-text screening. Apply the remaining criteria. Log the reason code, the stage, and a one-line note pointing to the specific passage or section that failed.
  4. Linking. Where several reports describe one study, link them rather than counting them as separate exclusions. Cochrane's guidance on linking multiple reports of one study supports treating them as one unit.

Checkpoint: for every excluded record, the log should show a code, a stage, and — for full-text exclusions — a locator. If any row is missing one of these, the record is not finished.

Keep the search and deduplication trail alongside the log

An exclusion log without a search trail is hard to defend, because a reader cannot tell what was screened in the first place. Cochrane recommends using multiple sources and combining free-text terms with subject headings, and documenting the process. Your trail should record which sources you searched, the terms and any field tags or filters you applied, the date of the search, and how duplicates were handled.

Database interfaces help here. PubMed's search details show how terms were translated and combined, which is useful when you need to reproduce a search later. Google Scholar documents searching by author and title, date limits, and citation export, which can help you capture records consistently. Neither replaces your own written trail — the trail is the artifact you keep.

This is also where a literature tool earns its place in the workflow. If you are rebuilding a search after a break, WisPaper search gives you a scholarly search interface to find papers, and the WisPaper Scholar Search page is a public entry point for finding academic papers. Use it to regenerate the candidate pool, then log the terms and dates exactly as you would for any database.

Worked example: cleaning up a thesis chapter over winter break

Here is how the pieces fit together for the reader with 40 PDFs and a supervisor's question about three dropped papers.

Step 1 — Reconstruct criteria. Write the population, intervention, outcome, design, and limits as separate lines. Date the list.

Step 2 — Define codes. Build the table above, mapping each code to one criterion. Keep it to six codes or fewer.

Step 3 — Rebuild the search trail. Record the sources, terms, filters, and search dates. If the original search is lost, rerun it and note that the rerun date is the recorded date.

Step 4 — Re-screen in stages. Start with deduplication, then title/abstract, then full text. For each of the 12 previously dropped papers, assign a code and a stage. For the three the supervisor asked about, add a locator — the section or sentence that failed the criterion.

Step 5 — Link multiple reports. Check whether any excluded records are actually secondary reports of an included study. If so, recode them as linked rather than excluded, and note the link.

Step 6 — Save the artifacts together. Keep the criteria, the code table, the search trail, and the screening log in one project space so the next reader finds them together. WisPaper Research Projects is where you can create a project and keep the work grouped; My Library is where you save or upload papers and organize them. Storing the PDFs and the log in the same place is what makes the record usable months later.

Step 7 — Answer the supervisor. For each of the three papers, quote the code, the stage, and the locator. That is the whole reply.

Use a checklist to audit the log before you call it done

Run this checklist over the finished log. Each line is a yes/no test.

  • Eligibility criteria are written as separate, testable statements and dated.
  • Every reason code maps to exactly one criterion.
  • Every excluded record has a code and a stage.
  • Every full-text exclusion has a locator.
  • Duplicates and linked reports are recorded as such, not as exclusions.
  • The search trail records sources, terms, filters, and dates.
  • Criteria revisions are noted with dates.
  • A second reader could reproduce the exclusion decision for any single record.

If any box is unchecked, the log is not ready. The checklist is also the fastest way to spot drift: a cluster of records with the same code and no locator usually means a stage was skipped.

Where literature tools fit — and where they do not

Tools speed up the parts of this workflow that are mechanical: finding candidates, organizing PDFs, and checking references. They do not decide eligibility for you, and the exclusion log remains your artifact.

  • Finding and regenerating the candidate pool. Use WisPaper search or the WisPaper Scholar Search entry point, then log terms and dates as you would for any database.
  • Organizing the screened set. My Library holds saved and uploaded papers so the included and excluded sets stay separable.
  • Asking questions across your saved papers. When you need to check whether a full-text exclusion was correct, Library QA lets you ask questions based on papers uploaded to My Library or added from search results.
  • Checking the reference list. Before you submit, TrueCite screens BibTeX against real academic databases to flag hallucinated references — useful when your exclusion log cites the papers you kept.
  • Longer synthesis tasks. If the review is growing into a larger synthesis, Scholar Agent supports literature review, inspiration, and reproduction use cases.

Each of these supports a step you would otherwise do by hand. None of them writes the reason code for you, and that is the point: the code is the record of your judgment.

How to know the record is ready

The record is ready when three conditions hold. First, every excluded record has a code, a stage, and — where the decision was made on full text — a locator. Second, the search trail is complete enough that the screened pool can be reconstructed. Third, a reader who has never seen the project can follow one exclusion from criterion to code to stage to locator without asking you a question.

A quick final test: pick the exclusion you are least sure about and try to defend it in two sentences using only your files. If you cannot, that row needs work before the chapter goes to your supervisor.

Your next action

Open your project space and create the code table first — six rows or fewer, each mapped to one criterion. Then re-screen your excluded set against it, starting with the papers your supervisor or reviewer is most likely to question. Save the criteria, the code table, the search trail, and the log together in Research Projects so the next person to open the chapter finds the reasoning, not just the PDFs.

FAQs

Treat duplicates separately from exclusions. Cochrane recommends deduplicating records as part of the selection process, and a duplicate is not a study that failed a criterion — it is the same study counted twice. Log duplicates in their own column or stage so your exclusion counts reflect genuine eligibility decisions.