Prognostication and integration of bedside lung ultrasound and computed tomography imaging findings with clinical features to Predict COVID-19 In-hospital mortality and ICU admission.

Summary

In 1,230 COVID-19 patients, consolidation on LUS/CT and absent A-lines were associated with mortality, while several imaging patterns predicted ICU admission. Although individual markers performed poorly (AUC <0.65), an integrated CT/LUS-clinical nomogram achieved high mortality prediction accuracy (AUC 87.3%).

Key Findings

• Consolidation on BLUS/CT and absence of A-lines were associated with in-hospital mortality.
• Ground-glass opacities, atelectatic bands, mosaic attenuation, crazy paving, and confluent B-lines were associated with ICU admission.
• Individual markers had poor predictive performance (AUC <0.65), whereas the integrated nomogram achieved AUC 87.3% for mortality.

Clinical Implications

Do not rely on imaging features alone for COVID-19 risk stratification; consider integrated nomograms combining CT/LUS and clinical variables to triage patients on ER arrival.

Limitations

• Retrospective single-center design limits generalizability and may introduce selection bias.
• COVID-19-specific model; applicability to non-COVID viral pneumonia or ARDS requires validation.

Future Directions

Prospective multicenter validation and dynamic, serial imaging-clinical integration; extend to broader viral pneumonia/ARDS populations.