PRECEDE II Presurgical Cognitive Evaluation Via Digital Clockface Drawing Focusing on Disparities - Project Summary Preoperative cognitive impairment is a significant predictor of postoperative complications. However, despite recommendations for preoperative cognitive screening in older adults, current tools, like the clock drawing test, face challenges in accurately assessing patients with low education and limited reading. The overall objective of this application is to examine how education impacts clock drawing production, how it modifies its predictive value on the postoperative outcome, and how to evaluate its screening accuracy compared to alternative tools for detecting preoperative cognitive impairment in less-educated patients. Building on the productivity and infrastructure of the original R01, this renewal will develop and validate innovative approaches tailored to sociodemographically diverse patient populations. The rationale is that improving cognitive screening tools for older adults with varying educational backgrounds will lead to better detection of preoperative cognitive impairment and enhanced prediction of postoperative complications. The overall objective will be achieved by pursuing three specific aims. (1) Preoperative Features: Develop an explainable AI-augmented digital Clock Drawing (dCDT) tool that considers educational disparities in cognitive features among older adults undergoing surgery, hypothesizing that certain clock drawing features will be absent in individuals with lower education levels. (2) Postoperative Outcome: Create a multimodal federated learning model to enhance predictive accuracy for postoperative outcomes, integrating preoperative cognitive and intraoperative data with sociodemographic factors. (3) Prospective Validation: Evaluate the accuracy of dCDT against two alternative digital metrics (Paragraph Recall and Symbol Cancellation) for identifying cognitive impairment in low- education older adults, with a hypothesis that dCDT metrics may underperform compared to these novel metrics. The approach is innovative because it represents the first attempt at (1) the focus on cognitive assessment accuracy in the lower education older adult patient population, (2) the comprehensive integration of explainable AI in the dCDT tool, and (3) the use of federated learning for cognitive predictive modeling. The proposed research is significant as it addresses critical gaps in perioperative cognitive assessment, enabling fair and accurate evaluation of older adults with lower education. The expected outcomes include improved perioperative cognitive screening tools, better prediction of postoperative complications, and enhanced care for older adults at risk for cognitive impairment.
