Abstract
INTRODUCTION: Cognitive decline has become a growing public concern, yet large-scale exposure data identifying the contributing factors remain limited. METHODS: We conducted an exposure-wide association study involving 1142 participants and 207 exposures, using machine learning to assess the relative contribution and joint effects of key factors. Cluster analysis and intervention simulation trials helped identify high-risk subpopulations and the potential benefits of targeted interventions. RESULTS: In adjusted mixed models, the socioeconomic status domain emerged as the strongest predictor of longitudinal global cognitive score (β = 2.91, p < 0.0001, q < 0.0001), while the dietary domain also played an important role in memory function. The cluster analysis found that the "unfavorable lifestyle" dominated phenotype was associated with the poorest cognitive outcomes. Simulation trials indicated that cognitive scores could improve by shifting individuals from unfavorable to favorable phenotypes. DISCUSSION: Cognitive health requires multidomain interventions, particularly in the socioeconomic and dietary fields, and necessitates collaboration between government and individuals. HIGHLIGHTS: The exposure-wide association study design, which assesses a broad range of exposures, is used to identify novel variables and understand their contributions to cognitive function. The findings from the multidomain analysis indicate that socioeconomic status is the most significant contributor to global cognitive function, while diet plays the largest role in memory function. Increasing the proportion of favorable phenotypes through multidomain interventions can significantly enhance public cognitive health.