Open access · ISSN pendingPublished by MedRise
The DifferentialJournal of Clinical Reasoning

Peer review

Peer review policy

Two independent human experts review every peer-reviewed article. AI helps them check the details; it never decides.

In short

  • Original Research, Reviews, Innovation Reports, Reasoning Cases, Perspectives and Point–Counterpoint pieces are each reviewed by at least two independent external peer reviewers.
  • Letters and Editorials get editorial review by at least two editors and are labeled "Not peer-reviewed".
  • With the authors' permission, the journal runs two independent AI pre-reviews that flag points for the reviewers and editor to check. One reviewer works with these notes; the other reviews without them.
  • Every article states the review it received and whether AI pre-review was used.

Step by step

  1. Technical check (within 3 working days)The editorial office checks scope, completeness, anonymization, required statements, reporting checklist, and text overlap with published work.
  2. AI pre-review (within 2 days, if the authors allow it)Two separate AI reviews of the manuscript, each run independently, produce a list of points to check. See AI pre-review.
  3. Editor's assessment (within 10 days)The handling editor reads the manuscript and decides whether to send it for review. Manuscripts outside scope or below the journal's standard are declined at this stage with reasons.
  4. Independent peer review (within 21 days)At least two experts review the manuscript, each without seeing the other's report. Reviewer A receives the AI pre-review with the manuscript and must mark each AI point as agreed or disagreed, with a reason, alongside their own assessment. Reviewer B reviews without the AI notes; after filing, they see the AI pre-review and may add a short addendum.
  5. DecisionThe handling editor weighs the human reviews and recommends accept, minor revision, major revision or reject; the Editor-in-Chief confirms. Authors receive all reviewer reports and the AI points the editor found valid.
  6. RevisionAuthors respond point by point. Major revisions usually return to the same reviewers.
  7. PublicationAccepted articles are copyedited, proofed and published with their dates and review statement.

AI pre-review

AI is good at checking details that tired humans miss: a missing checklist item, a reference that does not exist or does not say what the text claims, an unexplained exclusion, a statistic that does not add up. It is not good at judging whether a question matters or whether a conclusion follows from messy evidence. So it does the first and never the second.

  • Consent first. Authors choose at submission whether to allow AI pre-review. Saying no never affects the decision; the manuscript simply goes to the reviewers without AI notes.
  • Run by the journal, never by reviewers. The pre-review runs on the journal's own AI service under commercial terms that bar the provider from training on the manuscript and require it to delete the manuscript from its systems within 30 days. Reviewers and editors may not put manuscripts into any other AI tool.
  • Two independent passes. Two separate AI reviews, each run without seeing the other, check: the reporting guideline items, methods and statistics, whether each reference exists and supports the sentence that cites it, internal consistency of numbers, and possible ethics or consent gaps.
  • Suggestions, not judgments. The AI lists points to check. It gives no score and no accept or reject recommendation.
  • One review always AI-free. One reviewer works with the AI notes; the other reviews without them and sees them only after filing. Every manuscript gets at least one assessment untouched by AI, and the reviewer who used the notes must accept or reject each point with a reason and write their own summary and recommendation.
  • Human oversight. The editor decides which AI points are valid before anything reaches the authors. Each year the editors compare a sample of AI notes with the human reviews, including how the AI-assisted and AI-free reviews differed, and publish what they find, including the errors.
  • Disclosure. Each article says whether AI pre-review was used.

Choosing reviewers

Editors choose reviewers with expertise in the topic or the method. Reviewers must not have a conflict of interest with the work or its authors: no collaboration or shared institution in the last 3 years, no close personal relationship, no financial interest, and no competing work under review. Authors may suggest reviewers and ask that specific people be excluded; editors make the final choice and never rely only on authors' suggestions. Residents and fellows may review with a faculty mentor, who co-signs the report.

Anonymity

Original Research, Reviews and Innovation Reports get double-anonymized review: reviewers do not know who the authors are, and authors do not know who the reviewers are. Perspectives, Reasoning Cases and Point–Counterpoint pieces get single-anonymized review, because the author's position or setting is often part of the piece. Reviewers may sign their reports if they wish.

Decisions and revisions

Decisions are based on the importance of the question, the soundness of the methods, whether the conclusions follow from the evidence, and clarity. When reviewers disagree, the editor may ask a third reviewer. Revised manuscripts are due within 60 days (minor) or 90 days (major); after that they are treated as new submissions.

How long it takes

Our targets: first decision within 6 weeks of submission, and publication within 4 weeks of acceptance. Each article shows its received, accepted and published dates.

Editors' own papers

Manuscripts from editors, board members or MedRise staff are handled by an editor with no conflict, go through the same review, and the author is excluded from every step of the decision. No more than one article per issue may come from these groups, and each such article says so.

Appeals

Authors may appeal a rejection once, within 30 days, by explaining in writing where the decision went wrong. An editor not involved in the first decision reviews the appeal, with a new reviewer if needed. Their decision is final. See complaints and appeals.

Guidance for reviewers

  • Treat the manuscript as confidential. Do not share it, use it, or put any part of it into an AI tool.
  • Declare any conflict of interest as soon as you see the manuscript, and decline if you have one.
  • Write a report the authors can act on: a short summary, major issues, minor issues, and a recommendation (shown to the editor only).
  • Judge the work, not the authors. Be specific and courteous.
  • Tell the editor if you suspect plagiarism, duplicate publication, fabricated data or an ethics problem. Do not contact the authors.
  • Reviewers receive a certificate of review and are thanked by name each year unless they ask not to be.

Join the reviewer pool