Predicting Perioperative Depression and Mild Cognitive Impairments Among Adults Using Multimodal Machine Learning Techniques
Sylvester Orimaye, PhD, MPH, Assistant Professor, College of Global Population Health, University of Health Sciences and Pharmacy
Depression is a common health condition that significantly impairs people’s ability to function and resilience in everyday challenges. People with limited income and resources (e.g., those receiving Medicaid & Medicare) suffer disproportionately and could benefit from additional support. Overall, patients who experience depression are more likely to develop mild cognitive impairments (MCI) that progress to dementia (e.g., Alzheimer’s Disease. By 2050, dementia is expected to affect nearly 13 million people in the US (doubling from 2014), and its associated costs are expected to rise to almost $1.5 trillion. This rate of increase is not sustainable. Depression experienced perioperatively is related to different adverse outcomes, including but not limited to mental impairments, higher mortality rates, and higher financial costs. Around 40% of all surgical cases are associated with some form of depression leading to postoperative cognitive disorders. Thus, the perioperative period is a window wherein additional support could be offered to diagnose and manage depressive symptoms before such postoperative complications occur. Conversely, not all surgical patients develop perioperative depression. Therefore, to sustainably provide additional services, it is crucial to identify those at risk of perioperative depression likely to progress to MCI or dementia.
As such, this study investigates a clinically relevant Artificial Intelligence tool that identifies surgical patients for whom behavioral health interventions could provide positive post-surgical outcomes. Our predictive model will leverage multimodal datasets (medical history, audio, or speech sample, for training the machine learning algorithm. The goal is to use the predictive model as the core of a Behavioral Analysis and Intervention Tool (BEVAINT) clinical tool that identifies surgical patients for behavioral health intervention. Furthermore, we will evaluate the implementation of the BEVAINT tool with guidance from a panel of stakeholders and clinicians who serve as an advisory committee for the project.
Findings from the study can inform appropriate clinical or perioperative recommendations for persons with depression. Also, the study will inform policy changes for perioperative healthcare delivery and implementing integrated behavioral health during the pre-operative phase. More importantly, the proposed tool will mitigate burdensome monetary and well-being impacts on families and the healthcare system.