Development and External Validation of a Multimodal Artificial Intelligence Mortality Prediction Model of Critically Ill Patients Using Multicenter Data
Authors' conclusion
These findings highlight the importance of incorporating multiple sources of patient information for mortality prediction and the importance of external validation.
Study context
- Population
- Critically ill intensive care unit (ICU) admissions, MIMIC-III, MIMIC-IV, eICU and HiRID data sets, >200 hospitals, 2001–2022
- Participants
- 203,434 ICU admissions
- Comparison
- Structured data model vs Multimodal model
- Outcome
- In-hospital mortality predicted after the first 24 h of ICU admission (AUROC, AUPRC, Brier score)
Methodological detailsmulticentre, multimodal deep-learning model, Mortality rate ranged from 5.2 to 7.9% across the four data sets
- Multicentre
- Multimodal deep-learning model
- Mortality rate ranged from 5.2 to 7.9% across the four data sets
Performance
As reported in the abstract
Discrimination and calibration of the model integrating structured data
| Test or model | Area under the receiver operating characteristics (AUROC) | Area under the precision-recall curves (AUPRC) | Brier score |
|---|---|---|---|
| Structured data model | 0.92 (95% CI, 0.90 to 0.93) | 0.53 (95% CI, 0.49 to 0.57) | 0.19 (95% CI, 0.18 to 0.20) |
Structured data model
Area under the receiver operating characteristics (AUROC)0.92 (95% CI, 0.90 to 0.93)
Area under the precision-recall curves (AUPRC)0.53 (95% CI, 0.49 to 0.57)
Brier score0.19 (95% CI, 0.18 to 0.20)
Further findings
External validation, 8 institutions of the eICU data set
- AUROC range
- 0.84 to 0.92
Patients with available clinical notes and imaging data: adding notes and imaging
- AUROC improved
- 0.87 (95% CI, 0.85 to 0.89) to 0.89 (95% CI, 0.87 to 0.91)
- AUPRC improved
- 0.43 (95% CI, 0.36 to 0.51) to 0.48 (95% CI, 0.40 to 0.56)
- Brier score improved
- 0.37 (95% CI, 0.36 to 0.39) to 0.17 (95% CI, 0.16 to 0.19)
- Difference in AUROC statistically significant (DeLong test)
- P = 0.0167
Original publication
Development and External Validation of a Multimodal Artificial Intelligence Mortality Prediction Model of Critically Ill Patients Using Multicenter Data. Anesthesiology. Published 16 September 2026. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →