A Bayesian Hierarchical Spatial Mapping Approach to Assess Gestational Diabetes Mellitus Risk Among Immigrant Populations in Georgia Restricted; Files Only

Abstract

Gestational diabetes mellitus (GDM) poses significant health risks for both mothers and infants, making monitoring small-area GDM prevalence critical for informing targeted public health interventions. Foreign-born women, despite generally healthier pregnancy profiles, experience higher GDM risk than US-born women. In Georgia, nearly 1 in 4 immigrants face barriers to comprehensive prenatal care and the burden of GDM is likely under-counted. To estimate the true risk of GDM, we developed a Bayesian Hierarchical Immigrant GDM (BHIG) estimation model that accounts for spatial differences and corrects for measurement error. By borrowing strength across counties and adjusting for data sparsity, the BHIG model produces robust county-level estimates of GDM risk among immigrant mothers. Model results suggests that the burden of GDM among immigrant populations in Georgia is higher than currently reported - 12.4% compared to 7.9% according to Georgia Department of Public Health records. Spatial trends reveal clusters of elevated risk and underdiagnosis in central Georgia and the Atlanta metropolitan area. This work demonstrates the importance of integrating spatial and measurement error models to better address maternal health equity.

Table of Contents

1 Introduction 1

2 Methods 5

2.1 Data 5

2.2 Summary of model approach 8

2.3 Data Model for Estimating Gestational Diabetes Mellitus Counts Across All Counties 9

2.4 Process Model for Unobserved Latent Gestational Diabetes Mellitus Logit-Probabilities 10

2.5 Measurement Error Model 11

2.6 Derivation of Corrected GDM Counts and Associated Uncertainty 12

2.7 Simulation Study: Evaluating Measurement Error Correction 13

3 Computation 14

4 Results 15

4.1 Simulation Results 15

4.2 Global Parameter Estimates 15

4.3 Estimated GDM Sensitivity Across Georgia 16

4.4 Posterior Probability and Variance Estimates 17

4.5 Estimated Additional GDM Cases Across Georgia 19

5 Discussion 19

Bibliography 23

Appendix A Notation Table 32

Appendix B Full Model Specification and Conditional Distributions 33

Appendix B.1 Data Model 33

Appendix B.1.1 Observed True Counts 33

Appendix B.1.2 Observed Error-Prone Counts 33

Appendix B.2 Process Model 34

Appendix B.2.1 True GDM Risk 34

Appendix B.2.2 Measurement Error Model 34

Appendix B.3 Full Conditional Distributions 35

Appendix B.3.1 Posterior of θi (True GDM Risk) 35

Appendix B.3.2 Posterior of λi (Sensitivity Parameter) 35

Appendix B.3.3 Posterior of pi (Error-Prone Risk) 35

Appendix B.3.4 Posterior of Spatial Effects (ui and ωi) 35

Appendix B.3.5 Posterior of Regression Coefficients β 35

LIST OF TABLES

1 Births by Nativity in Georgia, 2018-2023 7

2 Summary of model performance across 100 simulated county-level datasets under varying levels of bias imposed on error-prone counties 15

3 Global parameter estimates included in the BHIG model 16

4 Comparison of GDM cases reported by GADPH and estimated by the BHIG model, 2018-2023 19

LIST OF FIGURES

1 Mapped crude prevalence of gestational diabetes (GDM) among immigrants in Georgia, 2018-2023. Gold standard counties (Candler, Decatur, Forsyth, Grady, Hart, and Tift) are outlined in red 6

2 Bayesian hierarchical immigrant gestational diabetes mellitus (BHIG) estimation model 9

3 Mapped posterior median sensitivity estimates for gestational diabetes in Georgia, 2018-2023 17

4 Mapped true posterior median probability (left) and true posterior median variance (right) estimates 18

5 Mapped comparison of posterior median model estimates without and with incorporation of sensitivity. (Left) Posterior median probability estimates for the model without incorporation of sensitivity; (Middle) Posterior median probability estimates for the model with sensitivity adjustment; (Right) Difference between the two models, highlighting areas of change 18

About this Master's Thesis

Rights statement

• Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.

School	Rollins School of Public Health
Department	Biostatistics
Subfield / Discipline	Biostatistics - MPH & MSPH
Degree	M.P.H.
Submission	Master's Thesis
Language	English
Research Field	Health Sciences, Public Health Health Sciences, Epidemiology Statistics
关键词	Bayesian Hierarchical models gestational diabetes risk Disease mapping immigrant health
Committee Chair / Thesis Advisor	Emily N. Peterson, Emory University
Committee Members	Howard H. Chang, Emory University

最新修改

Primary PDF

Thumbnail	Title	Date Uploaded	Actions
	File download under embargo until 15 November 2025	2025-04-16 16:08:50 -0400	File download under embargo until 15 November 2025

Supplemental Files

Thumbnail	Title	Date Uploaded	Actions