Prognostic assessment of sepsis-induced acute respiratory distress syndrome in older patients using clinical and CT-based radiomic features.

Summary

In 302 older SI-ARDS patients, a CT radiomics–clinical fusion model achieved superior 28-day mortality prediction (C-index 0.850 training; 0.839 validation) with good calibration and net benefit, outperforming SOFA and single-domain models. A nomogram enabled risk stratification into distinct survival groups.

Key Findings

• A fusion radiomics–clinical model yielded C-index 0.850 (training) and 0.839 (validation) for 28-day mortality.
• Outperformed SOFA and single-domain (radiomics-only or clinical-only) models with better calibration and decision-curve net benefit.
• Nomogram-based risk stratification separated patients into distinct survival groups (log-rank p<0.001).

Clinical Implications

Supports early risk stratification at admission to guide monitoring intensity, resource allocation, and trial enrollment for older SI-ARDS patients.

Limitations

• Single-center retrospective study with a relatively small validation cohort (n=60)
• CT acquisition/segmentation variability and lack of external multicenter validation

Future Directions

External, multicenter validation and prospective impact studies; harmonization of CT protocols and automated segmentation to facilitate clinical deployment.