Predictive Modeling Of Alzheimer’s Disease Risk And The Simulated Long -Term Impact Of Covid-19 Severity.

Abstract

Alzheimer's disease (AD) is a profoundly challenging neurodegenerative disorder^[2], and an expanding body of clinical literature increasingly links the post-acute sequelae of COVID-19 to long-term neurological and cognitive deficits. This computational inquiry investigates the potential impact of long-COVID conditions on the latent risk of developing AD using an advanced machine learning framework. A comprehensive clinical Alzheimer's disease dataset was leveraged for model training and feature extraction^[3], while empirical insights from a COVID-19 symptom and cognitive impact dataset were utilized to construct multi-tiered simulated patient profiles. Three distinct classification algorithms—Random Forest^[2], Logistic Regression^[3], and XGBoost—were trained to predict AD likelihood, with the XGBoost Classifier emerging as the superior predictive model^[7]. Feature importance analysis highlighted functional assessments, Activities of Daily Living (ADL), and specific cognitive state examinations as the primary computational drivers of predicted decline^[2]. By deploying the trained models against hypothetical post-COVID profiles ranging from mild to severe impairment, a distinct escalation in the predicted Alzheimer's risk probability was observed as post-COVID health markers deteriorated. These findings suggest a compelling structural vulnerability for accelerated cognitive decline in individuals experiencing severe long-term post-COVID conditions^[9]. However, because this machine learning simulation serves as an indirect proxy^[2], these results underscore an urgent mandate for prospective longitudinal clinical studies to establish definitive causal pathways between viral-induced sequelae and neurodegenerative outcomes^[8]..