CONSORT-NLP

CONSORT-NLP automates extraction and generation of CONSORT reporting checklists from PDF randomized clinical trial articles to standardize and evaluate adherence to CONSORT guidelines.

Key Features:

• Natural Language Processing (NLP): Leverages NLP to identify and generate CONSORT checklist items from trial text.
• CONSORT item coverage: Covers 34 of 37 CONSORT reporting items, with 30 items fully implemented.
• Accuracy: Achieves greater than 90% accuracy for the fully implemented items on validation sets.
• Efficiency: Generates checklists in an average of 23 seconds per article versus 11.9–57.6 minutes for manual review by three reviewers.
• Input format: Processes articles provided in Portable Document Format (PDF).

Scientific Applications:

• Trial reporting assessment: Automates evaluation of randomized clinical trial reports for adherence to CONSORT guidelines.
• Manuscript review support: Supports manuscript reviewers and journal editors by providing rapid CONSORT checklist generation during peer review.
• Author and researcher support: Assists clinicians, researchers, and scientists in preparing manuscripts to improve CONSORT compliance and reporting quality.

Methodology:

Development and evaluation used 158 published journal articles from high-impact journals in general/internal medicine, oncology, and cardiac/cardiovascular systems, split into training (111), testing (25), and validation (22) sets; the tool processes PDF articles and was evaluated on validation sets.