Missing Data Imputation and Multilevel Conditional Autoregressive Modeling of Spatial End-Stage Renal Disease Incidence

TL;DR

This paper introduces a Bayesian multilevel spatial model to analyze end-stage renal disease incidence in Florida, incorporating a novel method for imputing missing spatial data to improve healthcare strategies.

Contribution

It presents a new spatial state space imputation method and applies a multilevel Bayesian model to analyze renal disease incidence and social factors affecting dialysis facilities.

Findings

Quantified effects of social and health indicators on dialysis hospitalization ratios

Developed a novel spatial data imputation technique for missing data

Provided insights for healthcare policy targeting disadvantaged populations

Abstract

End-stage renal disease has many adverse complications associated with it leading to 20-50% higher mortality rates in people than those without the disease. This makes it one of the leading causes of death in the United States. This article analyzes the incidence of end-stage renal disease in 2019 in Florida using a multilevel Conditional Autoregressive model under a Bayesian framework at both the Zip Code Tabulation Area and facility levels. The effects of some social factors and indicators of health on the standardized hospitalization ratio of dialysis facilities are quantified. Additionally, as kidney research studies are posed with a great burden due to missing data, we introduce a novel method to impute missing spatial data using spatial state space modeling. The outcomes of this study offer potentially valuable insights for policymakers aiming to develop strategies that enhance healthcare and service quality for disadvantaged populations.