Back to search

South Dakota State University

Modeling Heterogeneous Trajectories of Chronic Kidney Disease Progression with Finite Mixture Models Using the All of Us Dataset

Abstract

dc:description.abstract

Background: Chronic kidney disease (CKD) does not progress uniformly across patients, and a single average trajectory often fails to capture this heterogeneity. Understanding variation in disease progression is important for improving risk stratification and clinical management. Objectives: This study aims to characterize the heterogeneity of chronic kidney disease (CKD) progression using longitudinal estimated glomerular filtration rate (eGFR) data and to identify factors associated with distinct progression trajectory patterns. Methods: A longitudinal cohort study was conducted using data from the All of Us} Research Program collected between April 23, 2002, and October 1, 2023. Adult participants with repeated eGFR measurements were included, resulting in a final analytic cohort of (n = 6, 424) unique patients. Gaussian mixture models (GMM) were applied to transition times between CKD stages to examine variation in the timing of disease progression. In addition, latent class mixed models (LCMM) were used to identify subgroups with distinct eGFR trajectories over time. Chi-square tests were used to assess associations between demographic variables and trajectory class membership. Multivariable logistic regression was then performed to identify independent predictors of trajectory membership, comparing the improving trajectory group to the declining trajectory group. Results: The Gaussian mixture model (GMM) captured heterogeneity in the rate of CKD progression across stages. The latent class mixed model (LCMM) identified three distinct eGFR trajectory classes: a stable class, a declining class, and an increasing class. In univariate analyses, race and ethnicity were associated with trajectory membership, but these associations became weaker after adjustment for clinical covariates. In the multivariable logistic regression model, older age (AOR =0.465, 95% CI: 0.322--0.654, p< 0.001), Black race (AOR =0.315, 95% CI: 0.115--0.838, p=0.023), higher baseline eGFR (AOR =0.832, 95% CI: 0.801--0.860, p< 0.001), hypertension (AOR =0.327, 95% CI: 0.150--0.691, p=0.004), cardiovascular disease (CVD) (AOR =0.338, 95% CI: 0.111--0.983, p=0.050), and end-stage renal disease (ESRD) (AOR = 0.161, 95% CI: 0.061--0.399, p< 0.001) were independently associated with lower odds of belonging to the improving trajectory group. Although log(UACR) was not statistically significant after adjustment (p=0.180), the interaction between diabetes and log(UACR) was significant (AOR = 0.497, 95% CI: 0.281--0.857, p=0.014). Other demographic, socioeconomic, behavioral, and healthcare utilization variables were not independently associated with trajectory membership. The model demonstrated good explanatory performance, with a Nagelkerke R^2 of 0.793. Conclusions: CKD progression is highly heterogeneous and appears to be associated primarily with baseline kidney function and clinical disease severity in addition to demographic characteristics. Mixture-based modeling approaches provide a more detailed understanding of progression patterns and may support improved risk stratification and targeted clinical management.

Degree

thesis:*
Name thesis:degree_name
Master of Science (MS)
Level thesis:degree_level
Thesis - Open Access
Discipline thesis:degree_discipline
Mathematics and Statistics

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Musah, Yahaya Mumuni
Contributors dc:contributor
  • Semhar Michael

Rights

dc:rights

Identifiers

dc:identifier.*
Repository record dc:identifier
https://openprairie.sdstate.edu/etd2/2174

Chain of custody

source
Harvested from
South Dakota State University
Base URL
openprairie.sdstate.edu/do/oai/
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

Musah, Yahaya Mumuni. Modeling Heterogeneous Trajectories of Chronic Kidney Disease Progression with Finite Mixture Models Using the All of Us Dataset. Thesis - Open Access thesis, https://openprairie.sdstate.edu/etd2/2174