Two Coauthors Edited Different Versions of the Same Literature Review: A Repair Workflow

Two coauthors edited different versions of the same literature review. One file has a stronger introduction and a cleaner reference list; the other has newer studies and a sharper limitations section. Neither is obviously wrong, and merging them line by line will not settle the question that matters: which claims does the shared evidence actually support? The practical answer is to stop treating the task as a text merge and start treating it as an evidence reconciliation. You rebuild one claim register, one verified reference set, and one applicability statement, then let the prose follow.

This guide is written for clinicians and medical researchers preparing a focused evidence brief, a research update, or a team discussion. It uses a single running example so the decisions stay concrete, and it separates what the evidence shows from how far that evidence travels to your population.

A diverged literature review usually fails in four places at once, and each failure needs a different repair.

1. Claim drift. The two versions make slightly different assertions from the same source, often because one coauthor read the abstract and the other read the full text.
2. Reference drift. The bibliographies overlap but do not match: one version has a preprint, the other a published version, and a third entry may be a duplicate under a different identifier.
3. Scope drift. One version quietly widens the population or outcome until the conclusion sounds stronger than the cited studies allow.
4. Recency drift. One coauthor added newer material that the other never saw, so the two documents are answering different questions.

Before editing, write a one-paragraph statement of the review's question: the population, the intervention or exposure, the comparator, and the outcome. In the running example, the question is whether a structured discharge-planning program reduces 30-day readmission in older adults discharged from general medical wards. Every disputed sentence in either version can then be tested against that question. Sentences that answer a different question are not errors to fix; they are material to move to a separate section or cut.

Do not merge paragraphs. Build a table with one row per factual claim and columns for the claim as written, the supporting source, the study design, the population studied, the outcome measured, and the status (agreed, disputed, unsupported).

Read both drafts and extract every sentence that asserts something about the evidence. You will usually find that the two coauthors agree on more than the visible differences suggest, because most divergence sits in a handful of sentences where one version generalizes and the other stays specific.

For each disputed row, go back to the source text rather than to either draft. A DOI resolves through doi.org to the current destination registered for that identifier, which helps you locate the work, but the bibliographic fields and the source text still need to be checked for identity and claim support. That distinction matters in a merge: a working link proves the paper exists, not that the sentence beside it is accurate.

When you need to locate a work whose citation is incomplete, searching by author and title, sorting or limiting by date, and checking cited-by and related-article links are documented ways to find the record and alternative versions. Use those routes to confirm which version a coauthor actually read.

This is where most coauthored reviews go wrong, and where a careful repair earns its value.

Evidence level describes the design and conduct of the studies: whether the comparison is randomized, how outcomes were ascertained, and how directly the study measured the outcome you care about. Applicability describes whether the studied population, setting, and delivery of the intervention resemble the context your brief addresses. A study can be strong evidence and still be a poor fit, and a study can be a close fit and still be weak evidence. Keep the two judgments in separate columns so neither silently substitutes for the other.

In the discharge-planning example, a trial conducted in a different health system with a different discharge workflow may be high-level evidence with limited applicability to your wards. Record that as a limitation of the body of evidence, not as a reason to delete the study.

Indirectness is the gap between what a study measured and what your question asks. Common forms in a clinical review include a different population, a different comparator, a surrogate rather than a final outcome, and a different care setting. Name the specific gap in the claim register rather than writing a vague sentence about heterogeneity.

Limitations belong at two levels. Study-level limitations describe what a single paper could not establish. Body-of-evidence limitations describe what the collection of papers, taken together, cannot establish: for example, that no included study reported the outcome in the population you serve. Writing both levels prevents a common merge artifact, where one coauthor's cautious study-level caveat gets attached to a broad conclusion it was never meant to qualify.

Reference drift is the easiest problem to fix and the easiest to overlook. Work through the combined bibliography and classify each entry as confirmed, duplicate, version conflict, or unresolved.

• Confirmed: the identifier resolves and the bibliographic fields match the cited work.
• Duplicate: the same work appears twice, often once as a preprint and once as the published article.
• Version conflict: the two drafts cite different versions of the same work, and the claims may differ between them.
• Unresolved: the entry cannot be matched to a record.

Publisher-deposited metadata available through services such as Crossref can include titles, authors, venues, dates, identifiers, licenses, funding, ORCID and ROR identifiers, abstracts when deposited, and post-publication updates. Use those fields to confirm identity and to catch updates, but treat deposited abstracts as metadata rather than as a substitute for the source text when a claim is disputed.

If your team works from BibTeX, run the combined file through TrueCite, which checks BibTeX against real academic databases to flag hallucinated references. That step catches entries that look plausible but do not correspond to a real record, and it is worth doing before the merge rather than after.

A merged document inherits the blind spots of both searches. Before finalizing, restate the search question and re-run it deliberately, so the combined review reflects a defined search rather than two coauthors' browsing histories.

Use WisPaper search to find academic papers through the scholarly search interface, and keep the query terms and date limits in the project record so a third coauthor could reproduce the set. If you want a public explanation of the search entry point before sharing it with the team, the WisPaper Scholar Search page describes finding academic papers.

As you screen results, save the ones that survive into My Library, where papers can be saved or uploaded and organized. A shared, organized library is the artifact that prevents the next divergence: both coauthors work from the same set instead of from their own downloads.

Once the claim register and the reference set exist, the remaining work is convergence, and convergence needs a shared place to happen.

Set up the review in Research Projects so the merged work has a project home rather than living in two personal folders. For longer synthesis tasks, Scholar Agent supports literature review work within WisPaper's agent area.

When a specific claim is disputed, ask a question against the papers already in the library rather than re-reading everything. Library QA lets you ask questions based on papers uploaded to My Library or added from search results. Frame the question narrowly, for example asking what a given study reported about the 30-day readmission outcome in its enrolled population, and record the answer next to the claim it resolves.

The corrected deliverable is ready when it passes a short set of checks that a coauthor who was not involved could apply.

• Every factual sentence maps to a row in the claim register with a confirmed source.
• Every reference resolves, duplicates are removed, and version conflicts are resolved in favor of the version actually cited.
• Evidence level and applicability are recorded separately for each key claim.
• Indirectness is named specifically wherever the studied population, setting, comparator, or outcome differs from the review question.
• Limitations appear at both study level and body-of-evidence level.
• The search that produced the included studies is documented well enough to be repeated.

If a claim cannot pass these checks, the honest move is to narrow it or remove it, not to average the two drafts.

The finished review in the running example no longer argues that discharge planning reduces readmission in older adults in general. It states what the included studies reported, in which populations and settings, for which outcomes, and where the evidence becomes indirect for the wards in question. The introduction is drawn from the stronger draft, the newer studies from the other, and both now sit under one claim register and one verified reference set.

The transferable lesson is that coauthor divergence is usually an evidence problem wearing a formatting costume. A claim register, a separated judgment of evidence level and applicability, an identifier-checked reference set, and a documented search turn two competing drafts into one defensible review. Set those artifacts up at the start of the next collaboration and the merge becomes a review step rather than a rescue operation.