Date of Award
Spring 2026
Document Type
Open Access Dissertation
Department
Epidemiology and Biostatistics
First Advisor
Monique J Brown
Abstract
There is a global demographic shift in aging, with exponential growth projected between 2020 and 2050. Older adults, aged 65 years and older, are at increased risk of Alzheimer’s disease and Related Dementias (ADRD). People living with ADRD frequently have multiple chronic conditions that require treatment with multiple medications. However, prior research has demonstrated adverse outcomes associated with polypharmacy among people living with ADRD. This study, therefore, examined the association between polypharmacy and mortality; assessed the mediating role of polypharmacy in the association between comorbidity and falls; and compared the predictive performance of machine learning (ML) approaches for mortality in the South Carolina Alzheimer’s Disease Registry (SCADR).
Data from the SCADR were limited to patient medical records. Polypharmacy was constructed as a time-varying exposure, and an inverse probability weighting Cox hazards model was used to estimate the hazard ratio (HR) and 95% confidence interval (CI) for all-cause mortality and ADRD-specific outcomes. Logistic regression models were implemented in the PROC CAUSALMED procedure in SAS to estimate the effects of polypharmacy as a mediator. Additionally, LASSO, Random Forest and Support Vector Machine (SVM) methods were used to predict mortality in the SCADR.
A total of 16550 participants were included in this study. The mean age of participants was 74. The prevalence of polypharmacy was 85%. Participants with polypharmacy had a significantly higher hazard risk of death (HR:1.467, 95% CI: 1.414 – 1.523). The mediation analyses showed that the total effect of comorbidity on falls was statistically significant, with about 15% increase in the odds of falls (p< 0.0001), and that about 41% this association was mediated through polypharmacy. A statistically significant interaction between comorbidity and polypharmacy (p< 0.0001). Twenty-five predictors were evaluated across the ML models. The LASSO demonstrated strong discrimination for mortality prediction (ROC-AUC=0.81), and the SVM achieved the highest accuracy (74%).
Polypharmacy was associated with a significantly higher risk of mortality among participants in the SCADR and was both a mediator and an effect modifier in the association. These findings underscore the complex role of medication management in older adults living with ADRD. Broader interdisciplinary strategies are required to address polypharmacy with emphasis on medication appropriateness. ML approaches may further support risk stratification. Future studies should examine prescription patterns, appropriateness and specific drug-drug interactions in the SCADR.
Rights
© 2026, Daniel Amobtiwon Amoatika
Recommended Citation
Amoatika, D. A.(2026). Polypharmacy, Causal Pathways, and Mortality in Alzheimer’s Disease and Related Dementias in South Carolina: Causal Inference And Machine Learning Approaches. (Doctoral dissertation). Retrieved from https://scholarcommons.sc.edu/etd/8869