The Joint Effects of Diabetes and Depression on Stroke in the Atherosclerosis Risk in Communities (ARIC) Study 公开

Abstract

1 Abstract

1.1 Background and Objectives

Although diabetes and depression are well recognized risk factors for stroke, their joint

effects on stroke risk remain underexplored. This study used data from the Atheroscle-

rosis Risk in Communities (ARIC) Study to quantify combined associations of prestroke

depression and diabetes with stroke risk, and to investigate the predictive potential of

depression-diabetes multimorbidity on stroke risk using logistic regression and machine

learning models.

1.2 Methods

We analyzed 5,459 ARIC participants who attended Visit 5 (2011–2013), were free of

prior stroke, and had complete data on diabetes and depressive symptoms (CES-D ≥9).

Participants were classified into four exposure groups: no diabetes or depression, dia-

betes only, depression only, and both diabetes and depression. Incident stroke events

were ascertained through 2020 via adjudicated hospitalization records. Cox proportional

hazards models estimated hazard ratios and 95% confidence intervals across three adjust-

ment levels: unadjusted, adjusted for age, sex, and race, and fully adjusted. We then

applied XGBoost classifier and stepwise logistic regression (AIC-based). Model discrim-

ination was assessed by AUC.

1.3 Result

Over a median 7.8-year follow-up, 233 incident strokes occurred. In the fully adjusted

Cox model, individuals with both depression and diabetes had the highest risk of stroke

(HR 1.95; 95% CI 1.05–3.60), compared to reference. Diabetes only (HR 1.22; 95% CI

0.91–1.63) and depression only (HR 1.12; 95% CI 0.55–2.30) showed weaker, non-significant

associations. The XGBoost model achieved moderate discrimination, improving recall for

stroke cases from 0.40 to 0.62 through threshold adjustment. The final logistic regression

model included age, prevalent CHD, the depression×diabetes interaction (p=0.025), and

hypertension, yielding a C-statistic of 0.625 (95%CI 0.59–0.66).

1.4 Conclusion

Prestroke depression-diabetes multimorbidity suggests a synergistic increase in stroke

risk beyond individual effects. However, the stroke risk predictive performance of the

models was constrained by class imbalance and limited predictor scope. These findings

demonstrates the importance of integrated metabolic and mental health management to

mitigate stroke risk.

Table of Contents

Contents

1 Abstract 2

1.1 Background and Objectives . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.3 Result . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

2 Introduction 3

3 Methods 4

3.1 Data Source . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

3.2 Study Population and Inclusion/Exclusion Criteria . . . . . . . . . . . . 4

3.3 Exposure definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

3.3.1 Diabetes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

3.3.2 Depression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

3.4 Outcome definition: Stroke . . . . . . . . . . . . . . . . . . . . . . . . . . 5

3.5 Covariates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

3.5.1 Sociodemographic Variables . . . . . . . . . . . . . . . . . . . . . 5

3.5.2 Lifestyle Variables . . . . . . . . . . . . . . . . . . . . . . . . . . 6

3.5.3 Clinical and Anthropometric Variables . . . . . . . . . . . . . . . 6

3.5.4 Medical history variables . . . . . . . . . . . . . . . . . . . . . . . 6

3.6 Statistical Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

3.6.1 Cox Proportional Hazards Model . . . . . . . . . . . . . . . . . . 7

3.6.2 Machine Learning Prediction Model . . . . . . . . . . . . . . . . . 7

3.6.3 Logistic Regression Model . . . . . . . . . . . . . . . . . . . . . . 8

3.7 Variables included in each model . . . . . . . . . . . . . . . . . . . . . . 9

3.7.1 Cox Proportional Hazards Model . . . . . . . . . . . . . . . . . . 9

3.7.2 Machine Learning Prediction Model . . . . . . . . . . . . . . . . . 9

3.7.3 Logistic Regression Model . . . . . . . . . . . . . . . . . . . . . . 10

4 Results 10

4.1 Baseline Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

4.2 Cox Proportional Hazards Model . . . . . . . . . . . . . . . . . . . . . . 13

4.3 Machine Learning model: Prediction of stroke risk . . . . . . . . . . . . . 17

4.4 Logistic Regression Model . . . . . . . . . . . . . . . . . . . . . . . . . . 20

5 Discussion 20

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	Epidemiology
Subfield / Discipline	Epidemiology - MPH & MSPH
Degree	M.P.H.
Submission	Master's Thesis
Language	English
Research Field	Biology, Biostatistics Health Sciences, Epidemiology
关键词	stroke Cox proportional hazards model Multimorbidity Atherosclerosis Risk in Communities (ARIC) Study Machine learning (XGBoost) depression diabetes Risk prediction
Committee Chair / Thesis Advisor	Alvaro Alonso, Emory University

最新修改

Primary PDF

Thumbnail	Title	Date Uploaded	Actions
	The Joint Effects of Diabetes and Depression on Stroke in the Atherosclerosis Risk in Communities (ARIC) Study ()	2025-07-28 16:41:19 -0400	Press to Select an action Download

Supplemental Files

Thumbnail	Title	Date Uploaded	Actions