Proof of Principle

The preliminary study performed on >15M randomized observations (SMART DST2; SMART STU3; MIT, Harvard, Boston Children's Hospital) showed completeness, computability, and accuracy of the raw data sources (Fast Healthcare Interoperability Resources - FHIR).

RESULTS: Cross-validated (70/30) AI-Driven model demonstrated good discrimination for unplanned ICU Admission - AUC 0.90; 95% CI, 0.76-0.94) and performed as well, or better than the current models (C statistic ranges: 0.56– 0.74).

The model calibration across different risk categories showed a close association of predicted and observed outcomes (PSI11; PSI04).