Robert Moskovitch

Acute Kidney Injury (AKI) progression from KDIGO Stage 1 to Stage ≥ 2 in the Intensive Care Unit (ICU) is associated with increased mortality, need for renal replacement therapy, and prolonged stay. Most existing predictive models are limited in handling heterogeneous multivariate temporal data, and real-time interpretability, hindering real-time clinical decision-making. We present a framework for continuous prediction of AKI progression that employs temporal abstraction, Time Interval-Related Patterns (TIRPs) mining, and the Fully Continuous Prediction Model (FCPM) to deliver continuously interpretable risk estimates. Using temporal abstraction, raw multivariate clinical temporal data are transformed into Symbolic Time Intervals (STIs) series from which frequent TIRPs ending with the AKI progression event are mined. FCPM learns a model that consists of the mined patterns, which then estimates in real-time their completion probability as new clinical data arrives, leveraging the time-duration distributions when the pattern is unfolding. We rigorously evaluate the FCPM with three data-driven state abstraction methods (EWD, EFD, SAX) and a knowledge-based abstraction defined by domain experts. A rigorous evaluation on a cohort of 1,343 ICU patients shows that FCPM combined with EWD achieves an AUC-ROC of 0.83 and a mean lead time of 6.35 hours, significantly outperforming LSTM, ResNet, TFT, and XGBoost baselines in both discrimination and lead time.