A Multilevel Spatial Survival Analysis of Patients in Texas with End-Stage Renal Disease

This study investigates end-stage renal disease cases in Texas using a multilevel spatial survival modeling framework. The objective is to evaluate a multilevel model specification that incorporates regional as well as individual factors, and that can be extended with random effects capturing unexplained variation in the independent variables; these random effects can be partitioned into simultaneous spatially structured and spatially unstructured components.

The analysis uses data from 109,018 adult patients who initiated end-stage renal disease treatment between 2009 and 2018, obtained from the United States Renal Data System. This paper presents this model structure for survival analysis using Moran eigenvector spatial filtering, providing an alternative way to conduct advanced spatial survival analysis.

Clinical variables, particularly age, cardiovascular comorbidities, and transplant status, are dominant predictors of survival. Racial disparities are observable, with Asian and Black patients exhibiting lower mortality risk relative to White patients. Socioeconomic indicators (poverty, urbanicity, and unemployment rate) show attenuated significance after adjusting for spatial and aspatial random effects, indicating their impact is partly mediated through unobserved regional heterogeneity and spatial autocorrelation.

These findings underscore the necessity of accounting for spatial dependencies and multilevel structures in survival analysis to avoid potentially biased inferences. The devised approach can offer a robust framework for guiding geographically targeted health interventions and resource allocation aimed at improving end-stage renal disease patient outcomes and reducing health disparities across diverse regions.