Comprehensive Summary
This paper uses artificial intelligence (AI) to predict and monitor postoperative delirium (POD) in patients undergoing cardiac surgery. Electronic health record data from 114 patients was collected for the study. Delirium was diagnosed through a Confusion Assessment Method for the ICU, and AI predictions were compared with a frailty screening and other risk assessment measures. The AI model used structured data and natural language processing from clinical notes to estimate the risk of POD. Of all the cardiac patients in the study, 23.7% were diagnosed with POD. Postoperative screening with the AI model demonstrated a good predictive ability, receiving an area under the receiver operating characteristic curve (AUROC) of 0.79. However, preoperative AI prediction performed lower with a AUROC score of 0.55 - 0.66. The authors note that while AI did not outperform traditional preoperative methods, it may benefit the health care system as it is an automated and more efficient way to identify patients at risk for POD, and may lead to earlier intervention protocols.
Outcomes and Implications
Postoperative delirium is a common and serious complication that many patients experience after cardiac surgery. It has been shown to lead to increased morbidity, longer hospital stays, and higher healthcare costs. This study is important because it demonstrates that AI tools can utilize electronic health record data to identify patients at risk for developing POD, and can potentially enable earlier interventions to prevent or mitigate POD. Clinically, the AI model can be used in addition to preexisting screening practices to improve patient monitoring after surgery. Currently, its preoperative performance is limited and needs further development. While the postoperative screening of the AI model is promising, further validation is necessary before widespread implementation in clinical practices. Overall, this study shows promise behind AI-assisted monitoring and its ability to improve outcomes for patients undergoing cardiac surgery.