Comparative evaluation of large language models in delivering guideline-compliant recommendations for topical NSAID use in musculoskeletal pain: a multidimensional analysis
Bibliographic record
Abstract
INTRODUCTION: While large language models (LLMs) are increasingly used in clinical decision support, their adherence to evidence-based guidelines-particularly for musculoskeletal pain management-remains understudied. METHODS: Four LLMs (DeepSeek-R1, ChatGPT-4o, Gemini, Grok-3) were evaluated on their responses to topical NSAID use for musculoskeletal pain through: assessments of response quality (accuracy, over-conclusiveness, supplementary information, and incompleteness), standardized readability metrics (Flesch Reading Ease, Flesch-Kincaid Grade Level), and the PEMAT-P tool to quantify actionability. RESULTS: The four LLMs showed significant variability in accuracy (ANOVA p = 0.045), with Gemini scoring highest (8.33 ± 0.77) and DeepSeek-R1 lowest (7.72 ± 1.52) and in over-conclusiveness (ANOVA p = 0.025), with Grok-3 scoring lowest (4.56 ± 1.42) and ChatGPT-4o highest 6.72 ± 1.49). ChatGPT-4o provided the most supplementary content (6.94 ± 2.29, p = 0.106) and DeepSeek-R1 had the highest incompleteness (5.00 ± 2.52, p = 0.261). All models exceeded recommended readability thresholds (9th-10th grade level), and none met the actionability standard (≤ 33.5%). CONCLUSIONS: LLMs demonstrate potential as clinical aids. The comprehensive performance of Gemini and Grok is relatively favorable, yet their readability and actionability remain unsatisfactory. Future development should integrate clinician feedback and real-world validation to ensure safety. Human oversight and targeted AI training are critical for safe implementation. Key Points • The study reveals significant differences in accuracy among LLMs, highlighting inconsistencies in clinical decision support. • While all models generated readable text, the complexity remained high, potentially limiting accessibility for some patients. • Glucocorticoid use for patients in remission was more strongly associated with impaired physical function in patients aged 75-84 than in patients aged 55-74 years. • Over-conclusiveness and incomplete adherence to evidence-based guidelines underscore the necessity for human oversight and targeted AI training in clinical applications.
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 machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.087 | 0.346 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.002 | 0.005 |
| Bibliometrics | 0.003 | 0.002 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.005 | 0.003 |
| Open science | 0.002 | 0.004 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.004 | 0.001 |
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 source (direct Gemma or distilled Codex), 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".