Challenging cases of hyponatremia incorrectly interpreted by ChatGPT

• DOI: 10.1186/s12909-025-07235-2
• Publication Date: 2025-05-22T14:50:08Z
• Authors: Kenrick Berend; Ashley Duits; Reinold O. B. Gans

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