Toronto, Ontario

Biopsychosocial Characterization of Cognitive Decline and Late Life Depression Trajectories Using Bayesian Consensus Clustering and Machine Learning

MSc Thesis Defense

Read Post

Presenter: Mu Yang

Supervisory Committee: Dan Felsky (Supervisor), Osvaldo-Espin Garcia and Divya Sharma
Date and Time: Thursday, August 31st , 2023 at 10am EST

Location: Health Sciences Building (155 College St), Room HS734


Abstract: Cognitive decline and late-life depression (LLD) often co-occur and impact elderly’s wellbeing. However, their intersection and antecedents are understudied. We analyzed 2,992 participants from the Religious Orders Study and Memory and Aging Project, integrating symptoms and diagnostic records with Bayesian consensus clustering to define latent subtrajectories of cognitive decline and LLD. Logistic regression, elastic-net regression and XGBoost were used to build models of these trajectories including n=57 biopsychosocial predictors. Associations of subgroups with postmortem neuropathologies were assessed for 1,721 deceased participants. Three subgroups were identified for cognitive decline and two for LLD, which overlapped significantly (chi-square p=8.1x10-26). Elastic-net regression performed best overall, while XGBoost excelled at predicting moderate subgroups. High neuroticism and low physical health at baseline predicted unhealthy subgroups for both cognition and LLD. There were no neuropathologies associated with LLD trajectories. Our results suggest potential values of targeting neuroticism and physical health for healthy aging and patient screening.

All are welcome!