Document Type

Thesis - Open Access

Award Date

2025

Degree Name

Master of Science (MS)

Department / School

Mathematics and Statistics

First Advisor

Semhar Michael

Abstract

Traditional survival analysis techniques assume that all individuals will eventually experience the event of interest. This is as opposed to allowing for a subset of cured individuals. Hence, we consider a mixture cure model that can allow a subset of people to be cured of the disease and will never experience the event of interest. This gives a more accurate interpretation of survivability by analyzing both the shortand long-term effects of a set of variables on the individuals’ survival probabilities. However, the cure status is considered an unknown variable. Thus, the expectationmaximization algorithm is required to estimate this latent variable as well as the parameters of the model. This process may take an infeasible amount of time when working with large data due to the sequential bootstrap sampling procedure. Several computational techniques, including parallel computing, are applied to circumvent this situation. This study used a United States Renal Data System (USRDS) dataset that contains 2,228,693 people with incident end-stage kidney disease (ESKD) from 2000 through 2020, including those who either are on dialysis or had at least one transplant. Numerous predictors were included when modeling, such as demographic, co-morbid, and geographic factors. In Chapter 1, the results of an MCM are compared to a traditional Cox proportional hazards model. For instance, the Cox model indicates that White individuals had the highest adjusted hazard of all-cause mortality compared to any other race. The MCM confirmed this trend but also showed that White individuals were more likely to be classified as “cured” than both American Indians and Black individuals, each. In Chapter 2, survival scores were developed and validated using the MCM. Many spatial, survival trends across the United States were observed that could be validated using the USRDS data and current literature. For example, the Great Plains regions of the United States contained individuals who mostly had lower survivability whereas individuals residing in the Southeast section of the United States had higher survivability. The MCM developed within this thesis may benefit decision makers in more effectively addressing ESKD disparities.

Library of Congress Subject Headings

Chronic renal failure.
Kidneys -- Diseases -- Risk factors.
Proportional hazards models.
Survival analysis (Biometry)
Mixture distributions (Probability theory)

Publisher

South Dakota State University

Download

Share

COinS
 

Rights Statement

In Copyright