Date of Award

2017

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

College

College of Nursing

First Advisor

Elaine J. Amella

Second Advisor

Martina Mueller

Third Advisor

Patricia W. Slattum

Fourth Advisor

Andrew Singleton

Fifth Advisor

Mike A. Nalls

Abstract

Cognitive frailty is considered a potentially reversible age-related condition characterized by the simultaneous presence of both physical frailty and cognitive decline. The concept of cognitive frailty existing in older adults is indisputable, although the mechanisms and the directional relationship behind the dynamic association remain unexplained. Mechanisms have been suggested, often linking cognitive frailty to cognitive impairment or as a component of frailty but without an understanding of the biological bases for these associations we cannot not move forward with intervention trials. This dissertation examines the biological mechanisms for cognitive frailty. The study is the first to use a large number of protein and genetic markers identified by a systematic review to define the underlying pathology for cognitive frailty. We use an innovative Boosted trees machine learning technique for developing a population based predictive model. Xgboost is based in boosted trees and provides more efficient and accurate predictive modeling with large datasets and a rapid / robust framework for feature selection. Statistical modeling is used to design, test, and validate an accurate method for and identifying and classifying the features that predict individuals with cognitive frailty. The tree boosting model is used for the evaluation of multiple variables simultaneously and provides a high predictive value with low bias. The results presented within this dissertation create a foundation of understanding for a new aging condition and encourage translational research focused on the detection and prevention of cognitive frailty.

Rights

All rights reserved. Copyright is held by the author.

Share

COinS