The Association Between Neighborhood-level Poverty and HIV Virologic Failure Differs by Gender in Durban, South Africa: A Multi-level Analysis Público

Coker, Daniella Frances (2015)

Permanent URL:


Objective: If left inadequately diagnosed or untreated, HIV virologic failure (VF) can lead to acquired drug resistance or early mortality. Gender differences in predictive risk factors for VF, such as individual-level socioeconomic status (SES), have been found. In order to characterize the mechanisms through which SES impacts VF, it is important to also consider the role of neighborhood-level SES.

Methods: This is a secondary analysis of the Risk Factors for Virologic Failure (RFVF) case-control study of HIV positive patients residing in eThekwini and attending McCord Hospital in Durban, South Africa between October 2010 and June 2012, after at least 5 months of their first ART regimen. Cases were those with virologic failure (VL > 1,000 copies/mL) and controls were those without virologic failure (VL ≤ 1,000 copies/mL). Multilevel logistic regression (GEE) models incorporating interaction by gender were used to assess the gender-specific associations between neighborhood-level poverty and VF after controlling for individual-level demographic, clinical, and SES factors.

Results: 152 cases and 286 controls representing 52 neighborhoods (Main Places) were included in this analysis. Most patients in the sample were female (64.6%), came from high poverty neighborhoods (60.7%), were employed (72.4%), and had at least some secondary-level education (80.6%). In a GEE model only containing gender, neighborhood-level poverty, and their interaction term (Model 1) the OR for the effect of residence in low versus high poverty neighborhoods for men was 1.42 (95% CI: 0.79, 2.53) and for women was 0.76 (95% CI: 0.48, 1.22). Adjusting for individual-SES and clinical factors (Model 3: ORmen = 1.23; 95%CI = 0.63, 2.43 and ORwomen = 0.75; 95%CI = 0.49, 1.15) provided similar results for both men and women compared to when only CD4 count was adjusted for (Model 4: ORmen = 1.28; 95%CI = 0.64, 2.55 and ORwomen = 0.74; 95%CI = 0.48, 1.16).

Conclusions: After controlling for individual-level socioeconomic (SES) and clinical factors, men living in richer areas and women living in poorer areas had greater tendency for VF, perhaps partly due to gender-specific HIV-related stigma. Future studies should employ mediation analyses to further characterize potential mechanisms through which neighborhood-level SES effects may impact VF differently for men and women.

Table of Contents

Chapter I: Background/Literature Review. 1

Introduction. 2

HIV in South Africa. 3

HIV Epidemiology. 3

Development of the HIV epidemic in South Africa. 3

HIV Treatment. 4

Virologic Failure (VF). 7

Acquired Drug Resistance: Effect of Virologic Failure (VF). 9

Poor Adherence: Cause of Virologic Failure. 11

Risk Factors for Virologic Failure. 13

Demographic risk factors: Gender and Age. 13

Clinical risk factors: Duration of Antiretroviral Therapy. 16

Psychosocial risk factors. 17

Transportation risk factors: Distance and Travel time. 17

Socioeconomic risk factors: Education, Income, and Employment. 19

Neighborhoods and Health. 24

Neighborhood-Level Socioeconomic Status. 26

Role of Spatial Analysis: HIV in South Africa. 28

Causal Analyses. 30

Study Site and Geographic Considerations. 32

Durban. 33

Cato Manor. 34

Inanda, Ntuzuma, and KwaMashu (INK). 37

Umlazi. 41

Chatsworth. 42

Study Site. 44

Chapter II: Manuscript. 45

Title/Authors/Abstract. 46

Introduction. 47

Methods. 52

Study Location. 52

Study Design. 52

Data Sources. 53

Causal Analysis. 57

Statistical Analysis. 58

Results. 60

Patient Characteristics. 60

Neighborhoods. 61

Regression Models. 62

Discussion. 65

Strengths. 72

Limitations. 73

Future Directions. 74

Conclusion. 75

References. 76

Tables and Figures. 90

Table 1. Frequency distributions of RFVF patient characteristics by Case status. 90

Table 2. Frequency distributions of RFVF patients (n) within Main Places (N) by Neighborhood-level Exposures. 92

Table 3. Crude Odds Ratios (ORs) between each Covariate and Case status or Neighborhood-level Exposures (Normal Logistic Regression). 96

Table 4. Summary of Gender-specific effects (Low Neighborhood Poverty vs. High Neighborhood Poverty on Virologic Failure) of Multivariable GEE Logistic Regression Models. 99

Figure 1. Directional acyclic graph (DAG) relating study exposure (Neighborhood SES), outcome (VF), and covariates. 100

Figure 2. Hierarchy of graphic frames used by Statistics South Africa (Census 2011). 101

Figure 3. Exclusion scheme for current analysis from total cohort of RFVF study. 102

Figure 4. Spatial distribution of RFVF study participants in eThekwini Municipality (n=438) - Durban, South Africa. 103

Figure 5. Spatial distribution of RFVF cases and controls in eThekwini municipality (n=438) - Durban, South Africa. 104

Figure 6. Spatial distribution of RFVF study participants in eThekwini municipality (n=438) by neighborhood type - Durban, South Africa. 105

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.
Subfield / Discipline
  • English
Research field
Palabra Clave
Committee Chair / Thesis Advisor
Committee Members
Última modificación

Primary PDF

Supplemental Files