ICU sedation, delirium & recovery
Evaluating the efficacy of large language models in predicting intensive care unit admission needs
Authors' conclusion
ChatGPT demonstrated superior accuracy and reliability in predicting ICU needs compared to Gemini.
Study context
- Population
- ICU consultation cases in a tertiary hospital, January 2020–June 2024
- Participants
- 8,043 ICU consultation cases
- Exposure
- ChatGPT vs Gemini
- Outcome
- Prediction of ICU admission needs (accuracy, Kappa, Pearson Chi-square, logistic regression vs anesthesiologists' ICU decisions)
Methodological detailssingle-centre (tertiary hospital)
- Single-centre (tertiary hospital)
Main finding
As reported in the abstract
Agreement with anesthesiologists' ICU decisions
| Group | Value | Kappa | Pearson Chi-Square | P |
|---|---|---|---|---|
| ChatGPT | accuracy 93.8% | 0.802 | 5293.310 | p < 0.001 |
| Gemini | Kappa 0.396 |
Value
ChatGPTaccuracy 93.8%
GeminiKappa 0.396
Kappa
ChatGPT0.802
Gemini— not stated in the abstract
Pearson Chi-Square
ChatGPT5293.310
Gemini— not stated in the abstract
P
ChatGPTp < 0.001
Gemini— not stated in the abstract
Observational data: associations, not effects.
Original publication
Evaluating the efficacy of large language models in predicting intensive care unit admission needs. J Crit Care. Published 23 January 2026. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →