Evaluation of large language models as a diagnostic tool for medical learners and clinicians using advanced prompting techniques
Bibliographic record
Abstract
BACKGROUND: Large language models (LLMs) have demonstrated capabilities in natural language processing and critical reasoning. Studies investigating their potential use as healthcare diagnostic tools have largely relied on proprietary models like ChatGPT and have not explored the application of advanced prompt engineering techniques. This study aims to evaluate the diagnostic accuracy of three open-source LLMs and the role of prompt engineering using clinical scenarios. METHODS: We analyzed the performance of three open-source LLMs-llama-3.1-70b-versatile, llama-3.1-8b-instant, and mixtral-8x7b-32768-using advanced prompt engineering when answering Medscape Clinical Challenge questions. Responses were recorded and evaluated for correctness, accuracy, precision, specificity, and sensitivity. A sensitivity analysis was conducted presenting the three LLMs with basic prompting challenge questions and excluding cases with visual assets. Results were compared with previously published performance data on GPT-3.5. RESULTS: Llama-3.1-70b-versatile, llama-3.1-8b-instant, and mixtral-8x7b-32768 achieved correct responses in 79%, 65%, and 62% of cases, respectively, outperforming GPT-3.5 (74%). Diagnostic accuracy, precision, sensitivity, and specificity responses all outperformed those previously reported for GPT-3.5. Results generated using advanced prompting strategies were superior to those based on basic prompting. Sensitivity analysis revealed similar trends when cases with visual assets were excluded. DISCUSSION: Using advanced prompting techniques, LLMs can generate clinically accurate responses. The study highlights the limitations of proprietary models like ChatGPT, particularly in terms of accessibility and reproducibility due to version deprecation. Future research should employ prompt engineering techniques and prioritize the use of open-source models to ensure research replicability.
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 | 0.027 | 0.153 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.003 | 0.002 |
| Open science | 0.001 | 0.004 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.003 | 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".