Modeling the importance of life exposure factors on memory performance in diverse older adults: A machine learning approach

This work uses a regression tree machine learning model (XGBoost) with highly interpretable Shapley Additive exPlanation values to analyze impacts of 12 life exposure factors plus age, sex and ethnoracial identity on episodic memory outcome. This approach has valuable properties, including the ability to implicitly account for variable interactions, non-linear relations with outcome, and missing values. Age, sex, education, income, volunteering, exercise, hearing and vision, and sleep (quality and duration) have important impacts on memory outcome in a combined model and in stratified models regardless of ethnoracial identity. We also demonstrate individualized models for subgroups of participants, showing how life exposure factors vary in importance between divergent populations and suggesting an approach to personalized interventions. This approach can be valuable for both policy decisions and individualized interventions to support healthy cognitive aging.