Generating clinical cases for diagnosing urgent conditions using artificial intelligence: opportunities for clinical practice

Introduction

Timely and accurate diagnosis of critical conditions is essential in emergency medicine, as errors or delays increase the risk of complications and mortality. Artificial intelligence (AI), such as ChatGPT, can generate clinical cases for educational purposes, allowing practitioners to practice symptom recognition, rapid decision-making, prioritization of interventions, and outcome prediction—skills vital for effective emergency care.

Methods

We analyzed 50 multiple-choice clinical scenarios: 25 developed by experienced emergency physicians and 25 generated by ChatGPT using text prompts. Cases covered urgent conditions: cardiac arrest, various shocks, trauma, acute coronary syndrome, and acute respiratory failure. Five independent experts evaluated each case using ten criteria: clinical accuracy, completeness, logical consistency, realism, practical relevance, absence of errors, relevance of questions, uniqueness, and correctness of distractors. Quantitative assessment used the Item Content Validity Index (I-CVI) and Aiken’s V coefficient.

Results

Expert-prepared cases demonstrated high quality (3.8 ± 0.13; I-CVI = 0.984; S-CVI/Ave = 0.99; V = 0.936). AI-generated cases scored lower (3.0 ± 0.59; I-CVI = 0.496; S-CVI/Ave = 0.50; V = 0.671), mainly due to insufficient symptom detail, inconsistent logic, and variable distractor quality. AI cases can quickly cover a broad spectrum of scenarios but require refinement to ensure clinical accuracy and educational value.

Conclusion

High-quality clinical cases enhance practitioners’ ability to recognize critical symptoms, select appropriate diagnostics, and predict outcomes, reducing errors and accelerating decisions in emergencies. ChatGPT can aid in generating training cases for urgent conditions; however, expert review and revision are necessary. AI-generated content should complement, not replace, validated cases and established protocols, supporting improved diagnostic accuracy and practical value in real-world clinical practice.