Dynamic forecasting of severe acute graft-versus-host disease after transplantation

Abstract Forecasting of severe acute graft-versus-host disease (aGVHD) after transplantation is a challenging ‘large p , small n ’ problem that suffers from nonuniform data sampling. We propose dynamic probabilistic algorithm, daGOAT, accommodates sampling heterogeneity, integrates multidimensional clinical and continuously updates the daily risk score for aGVHD onset within two-week moving window. In studied cohorts, cross-validated area under receiver operator characteristic curve (AUROC) daGOAT rose steadily peaked at ?0.78 in both adult pediatric outperforming two-biomarker MAGIC score, three-biomarker Ann Arbor peri-transplantation features-based models XGBoost. Simulation experiments indicated algorithm well suited short time-series scenarios where underlying process event generation smooth, there are frequent irregular missing. daGOAT’s broader utility was demonstrated by performance testing on remotely different task, is, prediction imminent human postural change based smartphone inertial sensor data.