MétaCan
Menu
Back to cohort
Record W4410596951 · doi:10.1186/s12909-025-07235-2

Challenging cases of hyponatremia incorrectly interpreted by ChatGPT

2025· article· en· W4410596951 on OpenAlexaboutno aff
Kenrick Berend, Ashley J. Duits

Bibliographic record

VenueBMC Medical Education · 2025
Typearticle
Languageen
FieldMedicine
TopicArtificial Intelligence in Healthcare and Education
Canadian institutionsnot available
Fundersnot available
KeywordsHyponatremiaMedical educationPsychologyMedicineInternal medicine

Abstract

fetched live from OpenAlex

BACKGROUND: In clinical medicine, the assessment of hyponatremia is frequently required but also known as a source of major diagnostic errors, substantial mismanagement, and iatrogenic morbidity. Because artificial intelligence techniques are efficient in analyzing complex problems, their use may possibly overcome current assessment limitations. There is no literature concerning Chat Generative Pre-trained Transformer (ChatGPT-3.5) use for evaluating difficult hyponatremia cases. Because of the interesting pathophysiology, hyponatremia cases are often used in medical education for students to evaluate patients with students increasingly using artificial intelligence as a diagnostic tool. To evaluate this possibility, four challenging hyponatremia cases published previously, were presented to the free ChatGPT-3.5 for diagnosis and treatment suggestions. METHODS: We used four challenging hyponatremia cases, that were evaluated by 46 physicians in Canada, the Netherlands, South-Africa, Taiwan, and USA, and published previously. These four cases were presented two times in the free ChatGPT, version 3.5 in December 2023 as well as in September 2024 with the request to recommend diagnosis and therapy. Responses by ChatGPT were compared with those of the clinicians. RESULTS: Case 1 and 3 have a single cause of hyponatremia. Case 2 and 4 have two contributing hyponatremia features. Neither ChatGPT, in 2023, nor the previously published assessment by 46 clinicians, whose assessment was described in the original publication, recognized the most crucial cause of hyponatremia with major therapeutic consequences in all four cases. In 2024 ChatGPT properly diagnosed and suggested adequate management in one case. Concurrent Addison's disease was correctly recognized in case 1 by ChatGPT in 2023 and 2024, whereas 81% of the clinicians missed this diagnosis. No proper therapeutic recommendations were given by ChatGPT in 2023 in any of the four cases, but in one case adequate advice was given by ChatGPT in 2024. The 46 clinicians recommended inadequate therapy in 65%, 57%, 2%, and 76%, respectively in case 1 to 4. CONCLUSION: Our study currently does not support the use of the free version ChatGPT 3.5 in difficult hyponatremia cases, but a small improvement was observed after ten months with the same ChatGPT 3.5 version. Patients, health professionals, medical educators and students should be aware of the shortcomings of diagnosis and therapy suggestions by ChatGPT.

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 imitation

Not 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.

metaresearch head score (Codex)0.001
metaresearch head score (Gemma)0.008
Version: metacan-v3-hybrid-931329e0061cValidation status: machine_predicted_unvalidated
Candidate categoriesMetaresearch
Consensus categoriesnone
DomainCandidate signal: Methods · Consensus signal: none
Study designCandidate signal: Observational · Consensus signal: Observational
GenreCandidate signal: Empirical · Consensus signal: Empirical
Teacher disagreement score0.999
Threshold uncertainty score0.007

Distilled classifier scores by category (both heads)

CategoryCodexGemma
Metaresearch0.0010.008
Meta-epidemiology (narrow)0.0000.000
Meta-epidemiology (broad)0.0000.000
Bibliometrics0.0010.001
Science and technology studies0.0010.001
Scholarly communication0.0010.001
Open science0.0010.001
Research integrity0.0010.000
Insufficient payload (model declined to judge)0.0020.001

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.

Opus teacher head0.060
GPT teacher head0.440
Teacher spread0.380 · how far apart the two teachers sit on this one work
Validation statusscore_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it

Classification

machine, unvalidated

Machine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.

Study designObservational
DomainMethods
GenreEmpirical

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".

Quick stats

Citations0
Published2025
Admission routes1
Has abstractyes

Explore more

Same venueBMC Medical EducationSame topicArtificial Intelligence in Healthcare and EducationFrench-language works237,207