A pilot evaluation of the diagnostic accuracy of ChatGPT-3.5 for multiple sclerosis from case reports
Bibliographic record
Abstract
The limitation of artificial intelligence (AI) large language models to diagnose diseases from the perspective of patient safety remains underexplored and potential challenges, such as diagnostic errors and legal challenges, need to be addressed. To demonstrate the limitations of AI, we used ChatGPT-3.5 developed by OpenAI, as a tool for medical diagnosis using text-based case reports of multiple sclerosis (MS), which was selected as a prototypic disease. We analyzed 98 peer-reviewed case reports selected based on free-full text availability and published within the past decade (2014-2024), excluding any mention of an MS diagnosis to avoid bias. ChatGPT-3.5 was used to interpret clinical presentations and laboratory data from these reports. The model correctly diagnosed MS in 77 cases, achieving an accuracy rate of 78.6%. However, the remaining 21 cases were misdiagnosed, highlighting the model's limitations. Factors contributing to the errors include variability in data presentation and the inherent complexity of MS diagnosis, which requires imaging modalities in addition to clinical presentations and laboratory data. While these findings suggest that AI can support disease diagnosis and healthcare providers in decision-making, inadequate training with large datasets may lead to significant inaccuracies. Integrating AI into clinical practice necessitates rigorous validation and robust regulatory frameworks to ensure responsible use.
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 | 0.017 | 0.087 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.004 | 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 | 0.001 |
| Insufficient payload (model declined to judge) | 0.003 | 0.002 |
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".