Editor Guidelines

An editor is accountable for a decision that is defensible on the record: the manuscript, the reviews, and the reasoning that connects them. Reviewers advise; the editor decides.

The full framework an editor works inside is the Editorial Policies. This page is the operational side of it.

All submissions undergo initial editorial screening. Manuscripts that meet the journal's scope and minimum requirements proceed to independent peer review.

Screening establishes five things. Where one fails, say which, so that the author can act on the response.

1. Scope: the work sits within the journal's stated research lanes, and the informatics or decision contribution is substantive rather than incidental.
2. Article type: the manuscript is what the submitted category says it is, and is within that category's limits.
3. Declarations: ethics approval, consent, trial registration, data and code availability, funding, competing interests and AI use are present and specific.
4. Patient data: de-identification is described, and no individual is identifiable anywhere in the manuscript, its figures, its screenshots or its supplementary files.
5. Methodological adequacy at face value: the design can answer the question asked, and the reporting is complete enough for a reviewer to assess.

A manuscript outside scope is declined promptly and clearly. Returning it for revision it cannot pass helps nobody.

Single-blind by default; double-blind review is available on request. Manuscripts that proceed to external peer review are normally evaluated by at least two independent subject-matter experts.

Composition

Work in this field usually needs more than one kind of expertise. A machine-learning study evaluated only by clinicians, or only by computer scientists, produces reviews that miss opposite halves of the same manuscript. Aim for a panel that covers the method, the clinical or operational context, and — where the analysis carries the claim — the statistics.

Independence

1. No shared institution, recent collaboration, co-authorship or funding relationship with any author.
2. No commercial interest in the system, vendor or product under evaluation.
3. Author-suggested reviewers may be used, but never as the whole panel, and their independence is checked like any other.
4. Reviewers declare conflicts on accepting; an undeclared conflict discovered later invalidates the review.

Handling the review

When double-blind review is requested, editors must ensure that author-identifying information is not shared with reviewers. Check the manuscript, the file properties and the supplementary material before it is sent.

Reviews are confidential. Discourteous or personal comments are returned to the reviewer or withheld from the author, and the substance is conveyed in the editor's own words.

Editorial decisions are based on scope, scientific quality, methodological rigor, ethical compliance, reporting quality, and relevance to the journal.

Weighing conflicting reviews

Reviews are weighed, not counted. Where two reviewers disagree, identify whether they disagree about a fact in the manuscript, about the interpretation of a result, or about whether the question matters — the three call for different responses, and only the first is settled by rereading the manuscript. A third opinion is sought where the disagreement is substantive and the editor cannot resolve it from the evidence.

What a decision letter carries

1. The decision, stated plainly at the start.
2. The reasoning, in the editor's own words, not merely the reviews attached.
3. Where revision is invited: which points must be addressed, and which are suggestions.
4. Where the manuscript is declined: the reason, at a level of detail the author can learn from.

Machine-learning and AI manuscripts

Satisfy yourself that the reviews have addressed these, and ask directly where they have not.

1. Data leakage: was the split made at the right unit, and could information from the outcome have reached the features?
2. Comparator: is the result measured against current practice or a credible baseline, rather than against nothing?
3. Validation: does the claim extend beyond the development setting, and if so, is there external validation to support it?
4. Calibration, where the output is used as a probability that informs a decision.
5. Subgroup performance, and what any difference means for the intended use.
6. Reproducibility: could another team rebuild this from what is reported?

Decision-support claims

Healthcare decision-science and formal research

Assess the decision problem, alternatives, patient preferences or utilities, assumptions, uncertainty and decision quality as applicable. Formal and conceptual contributions require sound reasoning and suitable evaluation, not necessarily a software artifact or clinical cohort. Select reviewers for the actual contribution, including decision-science expertise where needed. Apply ethics, consent and trial-registration requirements only when the study makes them applicable.

Methodological work without deployment

A sound methodological contribution is not weak because it has not yet been deployed. Where the work fits the journal's scope and the claims match the evidence, judge it on what it claims to be. The scope boundary is about healthcare relevance, not about deployment.

Ethics and patient data

Where approval, consent, de-identification or trial registration is missing, unclear or inconsistent with the described method, resolve it before review rather than after. Where a concern about integrity arises at any stage, refer it to the Editor-in-Chief rather than handling it in correspondence with the author.

Confidentiality

A submitted manuscript is confidential. It is not shared, discussed outside the review, or used in the editor's own work.

Recusal

An editor with a competing interest in a manuscript hands it to another editor and takes no further part, including access to its reviews and decision.

Editor-authored manuscripts

Handled by an editor who is not an author, entirely outside the author-editor's access.

AI use

Editors must obtain journal permission before using AI to assist with editorial assessment. Any permitted use must be disclosed, and confidential manuscript content must not be uploaded where confidentiality cannot be assured.

Appeals

Appeals are considered by an editor who was not responsible for the original decision. Do not handle an appeal against your own decision.

Suspected misconduct

Referred to the Editor-in-Chief and handled under COPE guidance, with the author and, where appropriate, the institution given the chance to respond.