Artificial Intelligence (AI) Usage Policy

Authors may use generative AI or AI-assisted technologies to support activities such as language improvement, literature organization, idea development, data exploration, coding support, visualization, or manuscript preparation. AI tools must not replace the authors' critical thinking, scientific judgment, clinical reasoning, policy interpretation, analysis, or original scholarly contribution.

Authors are responsible for:

• verifying factual accuracy and checking references generated by AI;
• reviewing and substantially editing AI-assisted content;
• checking for bias, hallucination, fabricated citations, incorrect medical claims, or misleading policy interpretations;
• protecting patient information, personal health information, confidential clinical data, copyrighted material, and unpublished research data;
• ensuring that AI outputs do not compromise patient safety, privacy, or research ethics; and
• ensuring that the final manuscript represents the authors' own scholarly work.

Meaningful use of generative AI in manuscript preparation must be disclosed in a separate AI Declaration. The declaration should identify the tool used, its purpose, and the extent of human review and oversight.

Basic spelling, grammar, or punctuation checks do not normally require disclosure. If AI forms part of the research method, health-data analysis, predictive modeling, clinical informatics workflow, decision-support system, or policy analysis, its use must be described in sufficient detail in the Methods section.

AI tools, chatbots, language models, machine-learning systems, and other automated technologies must not be listed as authors or co-authors. Authorship requires human responsibility for the integrity of the work, approval of the final manuscript, accountability for its content, and the ability to respond to questions regarding the research.

Generative AI must not be used to create, manipulate, obscure, remove, or introduce features in medical images, diagnostic images, clinical photographs, laboratory images, research figures, or other visual evidence in a way that misrepresents the underlying data.

AI-generated illustrations that are purely explanatory or conceptual must be clearly identified and must not be presented as original empirical or clinical evidence.

An exception may apply when AI-assisted imaging, machine vision, image classification, segmentation, reconstruction, or visualization is part of the research design or methodology. In such cases, authors must describe the tool, model/version, training or validation process where relevant, procedure, and its role in generating or interpreting research data in the Methods section.

Submitted manuscripts are confidential documents. Reviewers must not upload manuscripts, manuscript excerpts, health datasets, patient information, clinical images, supporting files, or review reports into public generative AI systems.

Peer review is a human scholarly responsibility. AI tools must not be used to replace independent scientific assessment, evaluate clinical validity without human review, determine policy significance, assess manuscript quality autonomously, or determine review recommendations. Reviewers remain personally responsible for the content and integrity of their reports.

Editors must not upload submitted manuscripts, patient or participant information, confidential health data, editorial correspondence, reviewer reports, or editorial decision letters into public generative AI systems.

AI must not replace human editorial judgment or be used to determine acceptance, revision, or rejection. Editors remain fully responsible for scientific evaluation, ethical assessment, communication, and final publication decisions.

Health informatics research frequently involves sensitive clinical, administrative, genomic, imaging, electronic health record, insurance, public-health, or patient-generated data. Such information must not be entered into public AI systems unless lawful authorization, ethical approval, appropriate security safeguards, and applicable consent or data-use permissions are in place.

• Personally identifiable health information must be protected.
• De-identification or anonymization must be applied where appropriate.
• Institutional, legal, contractual, and ethical restrictions on health-data processing must be respected.
• Authors should disclose relevant privacy-preserving, access-control, or secure-processing procedures.
• AI-assisted analysis must not compromise confidentiality or expose protected health information.
• Data-governance limitations should be stated clearly where they affect reproducibility or sharing.

When artificial intelligence, machine learning, deep learning, natural language processing, predictive analytics, or automated decision-support systems are central to the research, authors must provide sufficient methodological detail to allow scientific evaluation.

• Identify the model, algorithm, software, or platform used.
• Describe training, validation, testing, and external validation procedures where applicable.
• Explain dataset composition, inclusion and exclusion criteria, and preprocessing procedures.
• Report relevant performance measures and uncertainty.
• Describe steps taken to assess overfitting, data leakage, calibration, fairness, and generalizability.
• Report clinically or policy-relevant limitations.
• Explain human oversight and how model outputs were interpreted.
• Where applicable, discuss potential harms arising from false positives, false negatives, bias, or inappropriate automated recommendations.

AI systems used in health research may reproduce or amplify biases present in training data, healthcare systems, clinical documentation, population sampling, or institutional practices.

Authors should evaluate relevant differences in model performance across populations where scientifically appropriate and should avoid unsupported claims of universal applicability. Potential effects on health equity, access to care, vulnerable populations, and discriminatory outcomes should be discussed where relevant.

AI-generated outputs must not be treated as substitutes for professional clinical judgment, public-health expertise, health-policy analysis, regulatory requirements, or appropriate human decision-making.

Authors must avoid overclaiming the clinical effectiveness, policy impact, safety, or real-world applicability of AI systems when such claims are not supported by appropriate evidence. Any recommendation derived from AI-assisted research remains subject to human interpretation and professional accountability.