How Well Does ChatGPT-4o Reason? Expert Evaluation of Diagnostic and Therapeutic Performance in Hand Surgery.
Bibliographic record
Abstract
Background: The application of large language model (LLM) in surgical decision-making is rapidly expanding, yet its potential in hand and peripheral nerve surgery remains largely unexplored. This study assessed the diagnostic and therapeutic performance of a large language model (ChatGPT-4o) in scenarios characterized by multiple valid management strategies and absent expert consensus. Methods: Three representative cases-thumb carpometacarpal (CMC I) arthritis, scaphoid nonunion, and carpal tunnel syndrome (CTS)-were developed to reflect frequent conditions in hand surgery with competing but accepted treatment options. Each case was submitted to ChatGPT-4o using a standardized prompt. LLM-generated responses were evaluated by 52 participants (34 board-certified hand surgeons and 18 residents) across diagnostic accuracy, clinical relevance, and completeness. Readability indices, including Flesch-Kincaid Grade Level, were analyzed to assess appropriateness for a medical audience. Results: ChatGPT-4o demonstrated coherent but limited diagnostic accuracy (mean 2.9 ± 1.2 SD), moderate clinical relevance (3.5 ± 1.0 SD), and slightly higher completeness (3.4 ± 1.1 SD). Performance was strongest in the standardized scenario (carpal tunnel syndrome, CTS) and weakest in individualized reasoning (CMC I arthritis). No significant differences were observed between experts and residents (p > 0.05). In higher-level reasoning, ChatGPT-4o performed best in CTS and weakest in CMC I arthritis. Readability confirmed professional-level language (mean Flesch-Kincaid Grade Level: 16.4). Conclusions: ChatGPT-4o shows promise as a supportive tool for diagnostic reasoning and surgical education, particularly where standardized frameworks exist. Its limitations in ambiguous scenarios highlight the ongoing need for expert oversight. Future large language model development should emphasize specialty-specific training and context-aware reasoning to enhance their role in surgical decision support.
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 | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| 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".