Joining the conversation: introducing a dedicated medical education corpus

Area of Science:

• Medical Education
• Bibliometrics
• Information Science

Background:

• Medical education scholars face challenges accessing relevant literature due to broad biomedical databases and limited search functionalities.
• Existing search methods, such as PubMed's Medical Subject Headings (MeSH), often miss a significant percentage of pertinent articles, hindering scholarly discourse.

Purpose of the Study:

• To develop the first dedicated Medical Education Corpus (MEC) to facilitate literature discovery for medical education scholars.
• To create a specialized machine learning classifier for accurately identifying and extracting medical education articles.

Main Methods:

• A 3-step process was employed: defining a core-periphery model for Medical Education Journals (MEJ), developing and training a machine learning classifier (MEC Classifier) on labeled articles, and applying the classifier to extract articles from selected journals.
• The MEC Classifier was trained on 4,032 manually labeled articles and evaluated for its ability to identify medical education content.

Main Results:

• The Medical Education Corpus (MEC) was established, containing 119,137 articles from core and adjacent medical education journals as of December 2024.
• The MEC Classifier demonstrated superior sensitivity (90%) compared to MeSH (66%) in identifying medical education articles, with a comparable positive predictive value (82% vs 81%).

Conclusions:

• The MEC provides a focused and comprehensive resource, enabling medical education scholars to efficiently access literature and engage in field-wide analyses.
• The underlying methodology supports complementary tools like the MedEdMentor Paper Database, enhancing accessibility for the medical education community.