The paper presents *Prescience*, a machine-learning-based system designed to predict the risk of hypoxemia during surgery and provide interpretable explanations for these predictions. The system was trained on minute-by-minute data from over fifty thousand surgeries, integrating comprehensive datasets from a hospital's Anaesthesia Information Management System (AIMS). *Prescience* improves the performance of anaesthesiologists in predicting hypoxemia risks and identifying contributing factors. The system's predictions are consistent with prior knowledge and literature, and it can help anaesthesiologists anticipate and proactively address hypoxemia events, potentially reducing patient harm. The study demonstrates that *Prescience* can enhance anaesthesiologists' ability to predict hypoxemia by providing both accurate predictions and detailed explanations of the factors influencing the risk. The system's predictions are based on a wide range of patient and procedure characteristics, including static and real-time data, and it can identify both modifiable and non-modifiable risk factors. The findings suggest that *Prescience* can significantly improve the clinical understanding and management of hypoxemia during anaesthesia care.The paper presents *Prescience*, a machine-learning-based system designed to predict the risk of hypoxemia during surgery and provide interpretable explanations for these predictions. The system was trained on minute-by-minute data from over fifty thousand surgeries, integrating comprehensive datasets from a hospital's Anaesthesia Information Management System (AIMS). *Prescience* improves the performance of anaesthesiologists in predicting hypoxemia risks and identifying contributing factors. The system's predictions are consistent with prior knowledge and literature, and it can help anaesthesiologists anticipate and proactively address hypoxemia events, potentially reducing patient harm. The study demonstrates that *Prescience* can enhance anaesthesiologists' ability to predict hypoxemia by providing both accurate predictions and detailed explanations of the factors influencing the risk. The system's predictions are based on a wide range of patient and procedure characteristics, including static and real-time data, and it can identify both modifiable and non-modifiable risk factors. The findings suggest that *Prescience* can significantly improve the clinical understanding and management of hypoxemia during anaesthesia care.