Thank you for reviewing for JMIDH. Please follow the COPE Ethical Guidelines for Peer Reviewers and the guidance below.

Before accepting

Accept only if the manuscript is within your expertise and you can respond within 21 days. Declare competing interests, including consultancy, employment or equity in companies with competing or related products. Tell the editor if you believe you know the authors, and do not involve colleagues without the editor’s permission.

What to assess in all papers

•     Health relevance: Does the work address a real clinical, public health or health system problem? Would its findings matter to patients, clinicians or health services?

•     Design and data: Are data sources, dates, setting and inclusion criteria described? Are data quality and missing data handled appropriately?

•     Ethics and privacy: Is there ethics approval or a documented exemption? Is patient data protected? Could any figure, screenshot or dataset identify a person?

•     Reporting: Does the manuscript follow the relevant guideline?

•     Reproducibility: Are code, models and data available as stated, or are restrictions justified? Is there enough detail to reproduce the work?

•     Conclusions: Are claims about clinical usefulness, safety or readiness for deployment supported by the evidence?

Additional points for AI and prediction model studies

•     Is the outcome clearly defined and clinically meaningful? Is the prediction time point realistic for use in practice?

•     Is there any risk of data leakage — for example the same patient in training and test sets, or predictors that would only be known after the outcome?

•     Is validation adequate? Internal validation alone is weak evidence; external or temporal validation is preferred.

•     Are discrimination, calibration and clinical utility (for example decision curve analysis) all reported with confidence intervals?

•     Is performance compared with a sensible baseline, such as existing scores or clinician judgement?

•     Is performance reported across relevant subgroups (for example sex, age, site and socioeconomic groups) to check for bias?

•     For imaging AI: are image sources, devices, reference standard and annotation process described?

•     For large language model studies: are model versions, prompts, settings and evaluation methods reported, and were outputs checked for errors and fabricated content?

Additional points for software, data and implementation papers

For software papers, check whether the software is installable, documented, openly licensed and archived, and whether its usefulness is demonstrated. For data descriptors, check the collection methods, de-identification, quality control and access conditions. For implementation reports, check whether adoption, use, outcomes, costs, and failures as well as successes are reported.

Writing the report

1.    A short summary of the manuscript and its main claims.

2.    Major comments affecting validity or conclusions, numbered.

3.    Minor comments on presentation and clarity.

4.    Confidential comments to the editor (optional), including any ethical or privacy concerns.

5.    A recommendation: accept, minor revision, major revision or reject.

Ethical obligations

Keep the manuscript, code and data confidential and delete them after the review. Do not upload any part of the submission or your report into generative AI tools. Do not run supplied code on systems where it could expose data to third parties. Do not contact the authors, and do not request citations of your own work unless essential. Report suspected misconduct to the editor.

Recognition and joining

Reviewers receive a certificate of review on request. Researchers who would like to review for JMIDH are welcome to write to editor@corvusmedpress.com with their CV, ORCID iD and areas of expertise.