Date of Award

Spring 5-15-2015

Degree Type


Degree Name

Master of Public Health (MPH)


Public Health

First Advisor

Richard Rothenberg

Second Advisor

Dajun Dai


INTRODUCTION : Prior research suggests that sexually transmitted disease is not uniformly distributed throughout populations and geographic areas. Several studies of the geography of STI and HIV revealed a consistent core-like distribution of certain infections, such as gonorrhea, syphilis and HIV/AIDS. These studies further theorized that the conditions that precipitate endemic STI and HIV are not bound to populations, but strongly influenced by network-level features of social groups.

METHODS : Two geographic areas in Atlanta, GA were compared – one set of 5 zipcodes in which HIV was highly endemic, and another set of 5 zipcodes in which HIV was only moderately endemic. Two hypotheses were tested in the study area. First, risk variables were selected and composited into a variable representing compound risk, or the presence of multiple risk factors in a single individual, and the distribution of compound risk across the two geographic areas was compared. Second, the correlation between social distance (as geodesic length) and geographic distance (as distance between the centroid of connected individuals) was compared across the two geographic areas.

RESULTS : Compound risk was far more prevalent in the high HIV area than in the moderate HIV area (OR: 3.549; 95% CI: 2.438 -- 5.165), even after controlling for potential confounders. A breakdown of the individual risk variables indicates that involvement in sex work (OR: 2.279; 95% CI: 1.549 – 3.354), history of injection drug use (OR: 4.377; 95% CI: 2.35 – 8.152), and having any disease status disparity (OR: 1.511; 95% CI: 1.113 – 2.086) were each significantly more prevalent in the high HIV area than the moderate HIV area, even when stratifying by gender. The examination of the correlation between social distance and geographic distance revealed markedly different correlations in the two geographic areas. For residents of the high HIV area, the Pearson’s correlation score (CC: 0.17175; 95% CI: 0.154887 – 0.188492) was significantly higher than in the moderate HIV area (CC: 0.07021; 95% CI: 0.050822 – 0.08954).

CONCLUSION : Areas of high HIV endemicity are associated with at least two of the characteristics described by Rothenberg (2005) : a higher prevalence of individual compound risk than observed in low or moderate HIV areas, and a higher correlation between geodesic and geographic distance than observed in low or moderate HIV areas. If the observed higher correlation is true and can be replicated in other study locations and with other demographic groups, then it may be useful to examine whether areas exhibiting a similar correlation are host to higher than expected rates of HIV. The compound risk finding is in line with the kinds of behavior-oriented HIV/STI risk studies that have been historically emphasized, while the difference correlation between geographic and geodesic distance suggests that behavioral factors do not provide a complete explanation for observed differences in endemicity.