Perioperative Artificial Intelligence Driven Integrated Modeling of Surgeries using Anesthetic, Physical and Cognitive Statuses for Predicting Hospital Outcomes.

The association between preoperative cognitive status and surgical outcomes is a critical yet scarcely explored area. We assessed how preoperative cognitive status, as measured by clock drawing tests, contributed to predicting length of hospital stay, charges, pain during follow-up, and 1-year mortality beyond intraoperative variables, demographics, physical status, and comorbidities. We expanded our analysis to 6 surgical groups where sufficient data was available. Clock drawings were represented by 10 constructional features discovered by a semi-supervised deep learning algorithm, validated to differentiate between dementia and non-dementia patients. Machine learning models were trained using 5-fold cross-validation to classify postoperative outcomes. Shapley Additive Explanations analysis was used to find the most predictive features. Our results showed that the perioperative cognitive dataset served as the best dataset for 12 of 18 possible surgery-outcome combinations. Interpretability analysis showed that surgery duration was the most significant predictor of adverse outcomes, followed by anesthetic concentration. Disruptions in baseline correlations between intraoperative variables revealed that low average blood pressure and high standard deviation of blood pressure predicted adverse outcomes. Among the clock features, clock size was the most significant predictor of adverse outcomes. Our findings have relevance for improving healthcare modeling and perioperative risk prediction.