Predicting Intraoperative Hypoxemia with Hybrid Inference Sequence Autoencoder Networks

We present an end-to-end model using streaming physiological time series to accurately predict near-term risk for hypoxemia, a rare, but life-threatening condition known to cause serious patient harm during surgery. Inspired by the fact that a hypoxemia event is defined based on the sequence of future-observed low SpO2 (i.e., blood oxygen saturation) instances, our proposed model makes hybrid inference on both future low SpO2 instances and hypoxemia outcomes, enabled by a joint sequence autoencoder that simultaneously optimizes a discriminative decoder for label prediction, and two auxiliary decoders trained for data reconstruction and forecast, which seamlessly learns contextual latent representations that capture the transition between present state to future state. All decoders share a memory-based encoder that helps capture the global dynamics of patient measurement. For a large surgical cohort of 72,081 surgeries at a major academic medical center, our model outperforms all baselines including the model used by the state-of-the-art hypoxemia prediction system. Being able to make minute-resolution real-time prediction with clinically acceptable alarm rate to near-term hypoxemic events, particularly the more critical persistent hypoxemia, our proposed model is promising in improving clinical decision making and easing burden on perioperative care.