Stephen Q. Muth

Social networks and infectious disease: The Colorado Springs study

The social network paradigm provides a set of concepts and methods useful for studying the structure of a population through which infectious agents transmitted during close personal contact spread, and an opportunity to develop improved disease control programs. The research discussed was a first attempt to use a social network approach to better understand factors affecting the transmission of a variety of pathogens, including hepatitis B virus (HBV) and human immunodeficiency viruses (HIV), in a population of prostitutes, injecting drug users (IDU) and their personal associates in a moderate-sized city (Colorado Springs, CO). Some of the challenges of studying large social networks in epidemiological research are described, some initial results reported and a new view of interconnections in an at risk population provided. Overall, for the first time in epidemiologic research a large number of individuals (over 600) were found connected to each other, directly or indirectly, using a network design. The average distance (along observed social relationships) between persons infected with HIV and susceptible persons was about three steps (3.1) in the core network region. All susceptibles in the core were within seven steps of HIV infection.

Social network dynamics and HIV transmission.

Objective:To prospectively study changes in the social networks of persons at presumably high risk for HIV in a community with low prevalence and little endogenous transmission. Methods:From a cohort of 595 persons at high risk (prostitutes, injecting drug users, and sexual partners of these persons) and nearly 6000 identified contacts, we examined the social networks of a subset of 96 persons who were interviewed once per year for 3 years. We assessed their network configuration, network stability, and changes in risk configuration and risk behavior using epidemiologic and social network analysis, and visualization techniques. Results:Some significant decrease in personal risk-taking was documented during the course of the study, particularly with regard to needle-sharing. The size and number of connected components (groups that are completely connected) declined. Microstructures (small subgroups of persons that interact intensely) were either not present, or declined appreciably during the period of observation. Conclusions:In this area of low prevalence, the lack of endogenous transmission of HIV may be related in part to the lack of a network structure that fosters active propagation, despite the continued presence of risky behaviors. Although the relative contribution of network structure and personal behavior cannot be ascertained from these data, the study suggests an important role for network configuration in the transmission dynamics of HIV.

Network structural dynamics and infectious disease propagation

We aimed to relate dynamic changes in risk-network (sex and/or injecting drug) structure to observe STD/HIV transmission. We analysed macroand micro-structural elements in 2 heterosexual networks, augmented by ethnographic observations. In a Colorado cohort of injecting drug users (n=595), measures of subgroup formation and of density of activity show decrease of network cohesion over time; only one HIV transmission was observed in 3 years. In a group of adolescent heterosexuals in Georgia (n=99), the reverse process (increase in structural cohesion) was associated with efficient syphilis transmission: 10 cases were observed. Changes in personal risk behaviours over time were modest. STD/HIV transmission patterns were associated with intensification or diminution of network cohesion. Network and ethnographic data suggest that enhanced connectivity facilitates transmission while segmentation impedes it, suggesting opportunities for interventions. These data also emphasize the need to re-evaluate purely behavioural explanations of STD/HIV transmission.

Mapping a social network of heterosexuals at high risk for HIV infection

Objective:To determine how heterosexuals at risk for HIV infection interconnect in social networks and how such relationships affect HIV transmission. Design:Cross-sectional study with face-to-face interviews to ascertain sociosexual connections; serologic testing. Participants:Prostitute women (n=133), their paying (n=129) and non-paying (n=47) male partners; injecting drug users (n= 200) and their sex partners (n=41). Participants were recruited in sexually transmitted disease and methadone clinics, an HIV-testing site, and through street outreach in Colorado Springs, Colorado, USA. Main outcome measures:Reported behaviors, risk perceptions, sociosexual linkages, and HIV prevalence. Results:Respondents were well informed, but reported engaging in high-risk behaviors frequently. Nevertheless, over 70% of respondents perceived themselves to be at low risk for HIV infection. The 595 respondents identified a social network of 5162 people to which they belonged. Network analytic methods indicated 147 separate connected components of this network; eight of the 19 HIV-positive individuals in the network were located in smaller components remote from the largest connected component. Conclusion:The isolated position of HIV-positive individuals may serve as a barrier to HIV transmission and may account for the lack of diffusion of HIV in heterosexual populations in this region. Network analysis appears useful for understanding the dynamics of disease transmission and warrants further development as a tool for intervention and control.

Pathways to prostitution: The chronology of sexual and drug abuse milestones

To assess the sequence, timing, and prevalence of sexual and illegal drug use milestones in prostitute women, we interviewed 237 prostitutes in the community and 407 comparison women at an STD clinic. Drug use was more commonly reported by prostitutes than comparisons (86% vs. 23%), as was non‐consensual prepubertal sex (32% vs. 13%). Sexual‐ and drug‐related milestones occurred in the same order in both groups, with drug use preceding sexual activity and injecting drug use preceding prostitution. Ninety‐four percent of prostitutes who injected drugs reported noninjectable drug use before prostitution, and 75% of prostitutes who injected drugs reported doing so before beginning prostitution. The age distributions at critical events were similar for prostitutes and comparison women who reported regular drug use. Comparison women who did not report regular drug use were in general older than both these groups at the time of early sexual experience and drug experimentation. However, the ordering of these eve...

Risk network structure in the early epidemic phase of HIV transmission in Colorado Springs

This study describes the risk network structure of persons with HIV infection during its early epidemic phase in Colorado Springs, USA, using analysis of community-wide HIV/AIDS contact tracing records (sexual and injecting drug partners) from 1985 to 1999. Paired partner information from other STD/HIV programme records was used to augment network connections. Analyses were conducted with and without this supplemental information. The results suggest that a combined dendritic and cyclic structural network pattern is associated with low to moderate HIV propagation in Colorado Springs, and may account for the absence of intense propagation of the virus.

Choosing a centrality measure: Epidemiologic correlates in the Colorado Springs study of social networks☆

Abstract In a continuing analysis of a large network of persons who practice risky behaviors in an area of low prevalence for HIV transmission, we compared eight measures of centrality. Although these measures differ in their theoretical formulation and their distributional forms, they demonstrated substantial concordance in ranking as noncentral all but one of the HIV-positive persons in a large connected component of 341 persons, providing further support for the role of network structure in disease transmission.

Social and geographic distance in HIV risk.

Objective: The objective of this study was to examine the relationship between social distance (measured as the geodesic, or shortest distance, between 2 people in a connected network) and geographic distance (measured as the actual distance between them in kilometers [km]). Study: We used data from a study of 595 persons at risk for HIV and their sexual and drug-using partners (total N = 8920 unique individuals) conducted in Colorado Springs, Colorado, from 1988 to 1992—a longitudinal cohort study that ascertained sociodemographic, clinical, behavioral, and network information about participants. We used place of residence as the geographic marker and calculated distance between people grouped by various characteristics of interest. Results: Fifty-two percent of all dyads were separated by a distance of 4 km or less. The closest pairs were persons who both shared needles and had sexual contact (mean = 3.2 km), and HIV-positive persons and their contacts (mean = 2.9). The most distant pairs were prostitutes and their paying partners (mean = 6.1 km). In a connected subset of 348 respondents, almost half the persons were between 3 and 6 steps from each other in the social network and were separated by a distance of 2 to 8 km. Using block group centroids, the mean distance between all persons in Colorado Springs was 12.4 km compared with a mean distance of 5.4 km between all dyads in this study (P <0.0001). The subgroup of HIV-positive people and their contacts was drawn in real space on a map of Colorado Springs and revealed tight clustering of this group in the downtown area. Conclusion: The association of social and geographic distance in an urban group of people at risk for HIV provides demonstration of the importance of geographic clustering in the potential transmission of HIV. The proximity of persons connected within a network, but not necessarily known to each other, suggests that a high probability of partner selection from within the group may be an important factor in maintenance of HIV endemicity.

Sexual networks and sexually transmitted infections: A tale of two cities

Research on risk behaviors for sexually transmitted infections (STIs) has revealed that they seldom correspond with actual risk of infection. Core groups of people with high-risk behavior who form networks of people linked by sexual contact are essential for STI transmission, but have been overlooked in epidemiological studies. Social network analysis, a subdiscipline of sociology, provides both the methods and analytical techniques to describe and illustrate the effects of sexual networks on STI transmission. Sexual networks of people from Colorado Springs, Colorado, and from Winnipeg, Manitoba, Canada, infected with chlamydia during a 6-month period were compared. In Winnipeg, 442 networks were identified, comprising 571 cases and 663 contacts, ranging in size from 2 to 20 individuals; Colorado Springs data yielded 401 networks, comprising 468 cases and 700 contacts, ranging in size from 2 to 12 individuals. Taking differing partner notification methods and the slightly smaller population size in Colorado Springs into account, the networks from both places were similar in both size and structure. These smaller, sparsely linked networks, peripheral to the core, may form the mechanism by which chlamydia can remain endemic, in contrast with larger, more densely connected networks, closer to the core, which are associated with steep rises in incidence.

The Atlanta Urban Networks Study: a blueprint for endemic transmission

ObjectiveTo study prospectively social networks and behavior in a group of persons at risk for HIV because of their drug-using and sexual practices, with particular emphasis on the interaction of risks and concomitant network structure. MethodsA longitudinal study was conducted of 228 respondents in Atlanta, Georgia in six inner-city community chains of connected persons, interviewing primary respondents and a sample of their contacts every 6 months for 2 years. Ascertained were: HIV and immunologic status; demographic, medical, and behavioral factors; and the composition of the social, sexual, and drug-using networks. ResultsThe prevalence of HIV in this group was 13.3% and the incidence density was 1.8% per year. Substantial simultaneity of risk-taking was observed, with a high level of both non-injecting (crack, 82%) and injecting (heroin, cocaine or both, 16–30%) drug use, the exchange of sex or money for drugs by men (approximately 35%) and women (57–71%), and high frequency of same-sex sexual activity by men (9.4%) and women (33%). The intensity of interaction, as measured by network features such as microstructures and concurrency, was significantly greater than that observed in a low prevalence area with little endemic transmission. ConclusionThe traditional hierarchical classification of risk for HIV may impede our understanding of transmission dynamics, which, in the setting of an inner-city population, is characterized by simultaneity of risk-taking, and moderately intense network interactions. The study provides further evidence for the relationship of network structure to transmission dynamics, but highlights the difficulties of using network information for prediction of individual seroconversion.