Using Natural Language Processing to Evaluate the Quality of Supervisor Narrative Comments in Competency-Based Medical Education
Bibliographic record
Abstract
PURPOSE: Learner development and promotion rely heavily on narrative assessment comments, but narrative assessment quality is rarely evaluated in medical education. Educators have developed tools such as the Quality of Assessment for Learning (QuAL) tool to evaluate the quality of narrative assessment comments; however, scoring the comments generated in medical education assessment programs is time intensive. The authors developed a natural language processing (NLP) model for applying the QuAL score to narrative supervisor comments. METHOD: Samples of 2,500 Entrustable Professional Activities assessments were randomly extracted and deidentified from the McMaster (1,250 comments) and Saskatchewan (1,250 comments) emergency medicine (EM) residency training programs during the 2019-2020 academic year. Comments were rated using the QuAL score by 25 EM faculty members and 25 EM residents. The results were used to develop and test an NLP model to predict the overall QuAL score and QuAL subscores. RESULTS: All 50 raters completed the rating exercise. Approximately 50% of the comments had perfect agreement on the QuAL score, with the remaining resolved by the study authors. Creating a meaningful suggestion for improvement was the key differentiator between high- and moderate-quality feedback. The overall QuAL model predicted the exact human-rated score or 1 point above or below it in 87% of instances. Overall model performance was excellent, especially regarding the subtasks on suggestions for improvement and the link between resident performance and improvement suggestions, which achieved 85% and 82% balanced accuracies, respectively. CONCLUSIONS: This model could save considerable time for programs that want to rate the quality of supervisor comments, with the potential to automatically score a large volume of comments. This model could be used to provide faculty with real-time feedback or as a tool to quantify and track the quality of assessment comments at faculty, rotation, program, or institution levels.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.004 | 0.008 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.002 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".