Publication Date

Spring 4-8-2021

Abstract

Chronic kidney disease (CKD) in children is known to result in poor growth and quality of life, and frequently results in kidney failure. The Chronic Kidney Disease in Children study (CKiD) is a prospective cohort study enrolling children ages 1 to 16 to assess health outcomes in children with CKD including the effects of declining glomerular filtration rate and the resulting consequences of growth failure on morbidity. Quantification of the magnitude of the risk for decreased kidney function and, ultimately, failure has been achieved through a variety of studies, often including cohort studies such as the CKiD study. Longitudinal studies are widely known to suffer from dropout and missing data. In this dissertation, we present an extension to the Two-Step Approach for nonignorable missing outcomes originally proposed for a linear mixed model framework that we have extended to a survival analytic endpoint. We argue that this Extended Two-Step approach provides a test for informative dropout and adjusts the risk estimates to reduce bias. Through simulations and an application to decreased kidney function in children enrolled in the CKiD study, we demonstrate the benefit of the method and the ease with which it can be applied.

Degree Name

Statistics

Level of Degree

Doctoral

Department Name

Mathematics & Statistics

First Committee Member (Chair)

James Degnan

Second Committee Member

V. Shane Pankratz

Third Committee Member

Erik Erhardt

Fourth Committee Member

Li Li

Project Sponsors

Partial support for this work was received via a pilot grant from the Mountain West Clinical Translational Research Infrastructure Network (NIH/NIGMS # 1U54GM104944, Subaward #17-746Q-UNM-PG56-00)

Language

English

Keywords

Informative dropout, informative censoring, missing outcomes, Two-Step Approach, CKiD, frailty model

Document Type

Dissertation

Download

Share

COinS