
Acute kidney injury: High-performance KDIGO 2-3 prediction using basic vitals, without reliance on serum creatinine.
Digital twins: Neural network model that predicts stroke patient trajectories to inform clinical decision-making or provide virtual control arms for clinical trials.
GI bleed: Prediction of GI bleed requiring intervention during inpatient stay using demographic and vitals data from 2 hrs after admission.
Transfer learning: Site-specific performance improvement using transfer learning techniques with small amount of site data, outperforming MEWS in mortality prediction.
COViage: Identifies adult COVID-19 patients at risk of respiratory decompensation or hemodynamic instability during hospital stay.
Racial bias: Mortality prediction algorithm trained on specially pre-processed data achieved greater accuracy and lower equal opportunity difference (EOD) over MEWS, APACHE, and SAPS-II.
Acute respiratory distress syndrome: Improved disease prediction using semi-supervised learning or recurrent neural networks.
Stroke: Enhanced population selection for interventional studies by predicting with patient risk of stroke in next 12 months.