Modeling and Statistical Analyses Probing the Heterogeneity in Individual Change Points in Longitudinal Cognitive Outcomes Open Access

Abstract

Trail Making Test Part B (Trail-B) score is sensitive to the executive function of a person and has been frequently used in longitudinal cognitive studies. In this thesis project, we use the longitudinal Trail-B scores from the Uniform Data Set (UDS) to help understand the course of cognitive decline in a cohort of study participants with mild cognitive impairment (MCI) at their initial visits. Our investigations attend to the presence of a potential change point in the trajectory of Trail-B Score. Unlike in existing studies, we allow the change point to vary across individuals. Moreover, we develop an intuitive exploratory approach to determining the timing of the individual change points. We further apply parametric and semi-parametric Acceleration Failure Time (AFT) models to assess how patients’ baseline characteristics (age, gender, education level in years, Trail-B score at the initial visit) affect the timing of the change points. We then fit a set of linear mixed-effects models to study the longitudinal patterns of Trail-B scores while taking into account the presence of a change point in the Trail-B score. In these models, we consider the interaction effects between the time and each individual characteristic. Based on our analyses, a more advanced age at baseline is associated with more rapid progression to the change point. In contrast, receiving more education at baseline or having a larger baseline Trail-B score are related to later occurrence of the change point. And participants with different genders do not have significantly different change points at the 0.05 level. With the final linear mixed model being chosen based on a stepwise selection procedure, we find out that only the baseline trail-B score shifts the change rate of Trail-B score over the entire period. Age and education significantly affect the slope of the time-score relationship before the change points but not after them. Our results shed useful insight regarding the timing of cognitive decline change point in MCI patients and its relation to patient characteristics. The knowledge can contribute to the development of early interventions to improve patient outcomes.

Table of Contents

Introduction 1

Methods 5

Data Collection and Cleaning 5

Statistical Analysis 5

Individual Change Point Detection Algorithm 5

Acceleration Failure Time Models 6

Longitudinal Data Analysis 7

Bootstrapping 12

Results 13

Discussion 24

References 26

About this Master's Thesis

Rights statement

• Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.

School	Rollins School of Public Health
Department	Biostatistics
Subfield / Discipline	Biostatistics - MPH & MSPH
Degree	M.S.P.H.
Submission	Master's Thesis
Language	English
Research Field	Health Sciences, Public Health Biology, Biostatistics
Keyword	Change-point Longitudinal
Committee Chair / Thesis Advisor	Limin Peng, Emory University
Committee Members	John Hanfelt, Emory University

Last modified

Primary PDF

Thumbnail	Title	Date Uploaded	Actions
	Modeling and Statistical Analyses Probing the Heterogeneity in Individual Change Points in Longitudinal Cognitive Outcomes ()	2024-04-08 11:19:08 -0400	Press to Select an action Download

Supplemental Files

Thumbnail	Title	Date Uploaded	Actions