The Relationship between Biological Aging and Cognitive Function: A Machine Learning Model

Purpose:The relationship between cognitive impairment and biological age (BA) is unclear.This study aimed to investigate the association between BA and cognitive impairment.Materials and Methods: A total of 2202 participants from the 2011-2012 and 2013-2014 National Health and Nutrition Examination Survey were included.Klemera-Doubal method age (KDM-Age), phenotypic age (PhenoAge), and their acceleration values were calculated based on laboratory parameters.Six machine learning models were constructed to compare performance metrics such as area under the receiver operating characteristic curve (AUC-ROC) and accuracy, and Shapley additive explanations (SHAP) was used to interpret feature contributions Results: After fully adjusting for confounders, each 1-year increase in PhenoAge was associated with a 6%, 8%, and 5% higher risk of cognitive impairment on the Consortium to Establish a Registry for Alzheimer's Disease (CERAD), Digit Symbol Substitution Test (DSST), and Animal Fluency Test (AFT) tests, respectively (all p<0.001); the highest quartile group exhibited a 5.98-17.48-foldincrease in risk compared to the lowest.KDM-Age showed only a marginal association with DSST odds ratio (OR)=1.02,p=0.049.Among all models, the random forest (RF) performed best for CERAD prediction (AUC-ROC=0.997,sensitivity=99.5%),significantly outperforming the other algorithms (p<0.001).SHAP results demonstrated that sociodemographic factors had mean contribution values (0.014-0.135) significantly higher than those of BA indicators (<0.011), underscoring the dominant role of social determinants.Conclusion: PhenoAge is a biological aging indicator with suggestive value for identifying the risk of cognitive impairment in older adults.Combining the RF model with key sociodemographic information may help facilitate early risk screening.