Evaluating the Efficacy of Large Language Models for Dizzy History Taking and Peripheral Vestibular Disorder Diagnosis
Bibliographic record
Abstract
ImportanceVertigo accounts for one of the most frequent presenting symptoms in primary care. However, complexities in differential diagnoses and reliance on clinical history contribute to frequent specialist referrals and diagnostic delays. Large language models (LLMs), like LLaMA-3.1-8B, offer new potential for assisting in clinical decision-making.ObjectiveTo assess the utility of a small-scale, open-source LLM in diagnosing peripheral vestibular disorders (PVDs), and evaluate the impact of synthetic data augmentation on diagnostic accuracy.DesignRetrospective chart review.Setting/ParticipantsA retrospective analysis included adult patients presenting with dizziness to a neuro-otologist at St. Joseph's Healthcare Hamilton between 2018 and 2023. The dataset comprised 100 clinical cases, supplemented with 40 synthetic cases generated using GPT-4. The LLaMA-3.1-8B model was evaluated on the clinical, synthetic, and combined datasets. Diagnostic reasoning approaches, including chain-of-thought reasoning and multi-shot prompting, were employed to optimize model performance.Main Outcome MeasuresMetrics for evaluation included top 1 and top 3 diagnostic accuracy, Cohen's kappa for inter-rater agreement, and accuracy in predicting symptom laterality.ResultsThe LLaMA-3.1-8B model achieved a top 1 diagnostic accuracy of 60.7% and a top 3 accuracy of 71.4% in the combined dataset. The most frequent diagnosis was Meniere's disease (55.7%), followed by vestibular migraines (9.3%) and labyrinthitis (9.3%). Diagnostic accuracy was highest for benign paroxysmal positional vertigo (90%), followed by Meniere's disease (80.8%). Less common conditions, such as superior canal dehiscence syndrome and vestibular paroxysmia, exhibited lower diagnostic accuracies. Cohen's kappa indicated substantial agreement for symptom side prediction (κ = 0.96) and moderate agreement for diagnosis (κ = 0.41) in the combined dataset.Conclusions and RelevanceThe LLaMA-3.1-8B model demonstrated promising accuracy in diagnosing PVDs. The model's performance highlights its potential to serve as a high-yield screening tool for primary care physicians and general otolaryngologists.
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 | 0.014 | 0.036 |
| Meta-epidemiology (narrow) | 0.003 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.003 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.002 | 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".