How much can biomarkers explain sociodemographic inequalities in cognitive dysfunction and cognitive impairment? Results from a machine learning model in the Health and Retirement Study

Background Biomarkers may be pathways by which social adversity affects cognitive aging and Alzheimer's disease and related dementias (ADRD) risk. Objective How much variance in cognitive dysfunction and cognitive impairment onset do blood-based and physiological biomarkers provide above and beyond easily attainable sociodemographic variables, and how much can biomarkers explain differences in cognitive functioning and ADRD by sociodemographic variables? Methods We utilize machine learning to generate measures of predicted cognitive dysfunction and cognitive impairment incidence based on 91 biomarkers, identify the relative importance of each biomarker, and examine how much these biomarkers mediate sociodemographic differences. Results Markers related to cellular aging, neurodegeneration, diet and nutrition, immune functioning, and lung function were identified as important. Biomarkers mediated 47.2–77.3% of the variance associated with age, 22.7–35.2% of racial/ethnic differences in cognitive dysfunction, and 12.5–17.6% of educational differences. Conclusions Biomarkers provide the potential to understand pathways linking sociodemographic characteristics to cognitive functioning and health. Future research should consider additional biomarkers and evaluate the specific systems that put people at risk for cognitive impairment.