Figure 2
From: Development of a deep learning model that predicts Bi-level positive airway pressure failure
Model performance. (Left) Hourly AUROC values from LSTM-RNN, LR models, and S/F ratio for predicting BIPAP failure. The rolling cohort shows each model’s n-hour AUROC for BIPAP episodes with at least n-hours of BIPAP. (Right) Number needed to alert plotted as a function of missed alarm rates for predictions at 6 h after BIPAP initiation, capturing intermediate and late BIPAP failures.
