Fichtenberg, Caroline M. PhD*; Muth, Stephen Q. BA†; Brown, Beth MA, MPA‡; Padian, Nancy S. PhD§; Glass, Thomas A. PhD*; Ellen, Jonathan M. MD∥
OVER THE PAST 15 YEARS, there has been growing recognition of the role sexual networks play in the spread of sexually transmitted infections (STIs) and HIV.1–4 Sexual networks necessarily shape the transmission of STIs/HIV because they constitute the paths through which infectious agents move from person to person. Mathematical models have demonstrated that characteristics of network structure such as mixing patterns,5–7 concurrency8–9 and degree distribution2,10–12 affect how quickly and how far a disease can spread through a network. Likewise, the likelihood of getting infected or of transmitting infection has been shown to be theoretically influenced by network position.10,13–15
However, despite the recognition of the importance of sexual networks as determinants of STI/HIV transmission1,4,16 and the multiple theoretical studies of the role of networks in disease transmission, relatively few empirical data exist about the structure of sexual networks in the general population.16 Because of the challenges associated with collecting network data, the majority of empirical studies of sexual networks have relied on data collected for disease control purposes, either as part of routine contact tracing or outbreak investigations.17–21 Further, of the few studies designed specifically to collect network data most have focused on networks of individuals perceived to be at high risk of infection (eg, injection drug users, commercial sex workers).22–26 Although these studies provide crucial insights into the characteristics of networks of infected or ‘high risk' individuals, they are not as informative about the population distribution of network characteristics. Such population-representative information is needed to provide the comparative data (i.e., network structure among the uninfected) necessary for ascertaining how network factors affect disease transmission. Population-representative data are also necessary for determining population susceptibility to STIs/HIV, enabling comparisons between different populations, and providing realistic parameters for mathematical models of STIs/HIV, which are important tools for understanding transmission dynamics and for estimating the impact of interventions.6
To the authors' knowledge only two studies to date have collected population representative sexual network data.27,28 These have begun to shed light on sexual network structure in general population samples by providing pictures of sexual network structure among high school students in a US Midwestern town and among young adults on a small island in Malawi. To complement these studies, we present the results of a population-based study of sexual networks among urban black adolescents. Black adolescents are a subgroup at particularly high risk for STIs in the United States.29–31 Furthermore, there is evidence that sexual network differences may underlie the large racial/ethnic STI and HIV disparities that exist in the US and that have remained unexplained by individual-level behaviors.32–34 Characterizing sexual network structure within black subpopulations may therefore contribute to elucidating these persistent disparities. In this study, snowball sampling was used to map the sexual networks of a household sample of black adolescents living in a neighborhood of San Francisco, CA with relatively high STI rates.35 We describe the structure of the observed sexual networks and the position of individuals in those networks using measures found in prior studies to be associated with STI transmission.
Study Sample Recruitment
Data are from the baseline wave of a population-based longitudinal study of STI risk factors among black adolescents that took place in San Francisco, CA from 2000–2002. The study methods have been described previously.36 Briefly, at baseline in 2000, random digit dialing was used to recruit a household sample of blacks aged 14 to 19 living in the Bayview Hunter's Point neighborhood of San Francisco (zip code 94124). Bayview-Hunter's Point, a predominantly middle- to low-income district, was chosen as the study site as it had some of the city's highest rates of bacterial STIs (1,428 chlamydia and 700 gonorrhea cases per 100,000 in 200035) as well as a high concentration of blacks (the neighborhood was 50% black in 2000 compared to 9% for the city as a whole37). A total of 580 unique eligible teens were identified through random digit dialing; 348 of whom enrolled, for a 60% recruitment rate. Those who did not enroll were younger and more likely to be male. According to the 2000 Census,37 there were approximately 1,700 black 14 to 19 year olds residing in zip code 94124 at the time the study was initiated; random dialing generated about a one-third sample (580/1,700) and approximately one-fifth of the target population were enrolled (348/1,700).
Participants who reported heterosexual sexual activity in the three months preceding the interview became the starting point (‘seeds') for the enumeration of participants' sexual networks. They were asked to provide names, contact information, and permission to contact for up to six people with whom they had had sex in the three months preceding the interview and who were at least 14 years old.
When those partners were interviewed, the same process was repeated, and repeated again when their partners were interviewed. In this way partners up to two relationships away from the seeds were interviewed. Study staff, four individuals with experience recruiting hard-to-reach individuals and training in contact elicitation and contact tracing, made extensive efforts to identify and locate named partners, using word-of-mouth, school yearbooks, public records, jail and prison records, community leaders and key informants.
More than half (62%) of the 348 enrolled household sample reported ever having had heterosexual sex, and 166 (48%) reported sex in the last three months. These 166 network seeds named 269 partners, of whom 11 were other seeds, 149 were not interviewed either because they could not be located, refused to participate, or the referring partner refused permission to interview them, and 109 were interviewed. These 109 ‘first-order' partners named 143 second-order partners, of whom 50 were naming partners (‘reciprocal nominations'), 60 could not be interviewed, and 33 were interviewed. These 33 second-order partners named 48 third-order partners, of whom 11 were reciprocal nominations and the remainder were not recruited for participation (by design). On average, participants named 1.6 partners. Of 388 named first-, second-, and third-order partners who were not reciprocal nominations, 142 were interviewed (37%). Considering only partners who were eligible to be recruited (i.e., where permission to recruit was granted and who were not reciprocal nominations or third-order partners), the recruitment rate was 51% (142/281). Overall, 308 individuals (166 seeds and 142 network members) were interviewed out of 554 unique individuals identified through the study (56%).
Demographic and behavioral data were collected by computer-assisted telephone interviews (CATI) for the majority of participants (71%) and by computer-assisted personal interview (CAPI) for those who could not be reached by phone and were instead interviewed in person (29%). Biologic samples for STI testing were collected in person, on average 10 days after the interview (SD = 16 days). Among interviewed network seeds and partners, 79% provided a specimen for STI testing. Participants testing positive were notified of their test results and received assistance with locating appropriate treatment. Data collection lasted from July 2000 to October 2001. Participants were paid $25 for completing the interview, and $10 for providing a specimen. Informed consent was obtained directly for those 18 and older, whereas guardian consent with participant assent was obtained for those younger than 18. Participants were assured that partners who were contacted would not be told that they had been nominated by someone. All study procedures were approved by the Institutional Review Boards at the University of California, San Francisco, the University of California, Berkeley, and the Johns Hopkins Medical Institutions.
The following demographic characteristics were collected: age, race, gender, and school enrollment status. Participants were asked to report the following behaviors for the three-month period preceding the interview: condom use, number of recent partners, use of alcohol and illegal drugs. For condom use, participants were asked to report their frequency of condom use with each partner: never, a few times, most of the time, every time. These partner-specific reports were averaged at the nominating individual level and categorized into never, sometimes, most of the time, and always.
Current infection with chlamydia or gonorrhea was measured by ligase chain reaction (LCR) nucleic acid amplification test (NAAT) (LCx Probe System; Abbott Laboratories, Abbott Park, IL.) on self-collected vaginal swabs for females and urine samples for males. The LCR test has been found to have a sensitivity of 90 to 100%, and a specificity of 95 to 100%.38,39
Sexual Network Characteristics
The first step in the analysis of the network data were the identification of individuals who were in the same network component, i.e., connected to each other directly or indirectly. Two important assumptions were made in this process. First, a partnership was considered to exist as long as one partner reported it, even if the other partner was interviewed and did not report it (in network terms, arcs were symmetrized). Second, exact relationship timing was not taken into account because of the difficulty of measuring relationship timing as well as evidence that cumulative network measures are more relevant for disease transmission than instantaneous measures.14 Instead, network components included all reported partnerships regardless of whether the partnerships overlapped in time. Therefore, whereas each partnership represents a relationship that occurred within 3 months of the interview of the person reporting it, components could indirectly link people over periods longer than 3 months.
We characterized the sexual networks using metrics found in prior studies to be associated with STI transmission10,13–15,40 or found to be theoretically relevant to STI transmission.41 Two types of metrics were computed: centrality variables that describe the position of an individual in a component, and structural variables that describe the morphology of network components. Centrality measures included: degree (number of direct partners), reach (number of partners 2 steps away, or number of partners' partners), information centrality (how close a person is on average to other component members), betweenness centrality (the likelihood of being on the shortest path between other component members), and eigenvector centrality (a measure of importance favoring connections to more highly connected partners). Structural characteristics included component size (number of people in a connected group), component density (proportion of all possible links that are observed), component diameter (longest shortest path within a component), and component centralization (extent to which a component is centralized around the node with highest centrality) according to degree, information and betweenness centrality. Only size, diameter, density, degree, and reach are meaningful in components of size two, therefore these were the only measures computed for components of size two.
All analyses were carried out in SAS v.9.1.42 Network component membership and network measures were calculated in SAS, by adapting James Moody's SPAN modules.43 The network visualization software Pajek44 was used to create network images. Differences between subgroups were assessed using t-tests for normally distributed continuous measures, Wilcoxon-Mann-Whitney tests for nonnormally distributed continuous variables and chi-squared tests or Fisher's exact tests for categorical variables. Results were stratified by gender because of the strong association between STIs and STI risk factors and gender.45,46
On average, network seeds were 17.5 years old, and 60% were female (Table 1). Three quarters (74%) reported being in school. Substance use was prevalent, with 56% of participants reporting alcohol use within the last three months, and 58% reporting recent illegal drug use (mostly marijuana).
Among network seeds, the average number of partners reported for the past three months was 1.6, with more than two thirds of seeds reporting only one partner (Table 1). Only three individuals reported more than six partners in the last three months, suggesting that the six partner limit for network elicitation did not substantially bias the observed network structure. Almost half of the sample reported always using condoms with all their recent partners, whereas one-fifth reported never using them with any partners. Females were more likely than males to report only one partner (p-value <.0001), had almost half as many partners as males (p-value <0.0001), and were less likely than males to report that condoms were used during sex (p-value = 0.0001). Prevalence of either chlamydia or gonorrhea was 9% among network seeds, with females more than three times as likely as males to be infected (14% vs. 4%) although the difference did not reach statistical significance, p-value = 0.07). Those who did not provide specimens were demographically similar to those who did but had slightly fewer recent partners and were slightly more likely to use condoms consistently, suggesting that we may have oversampled those who were more likely to be infected. The prevalence we observed in the household sample is consistent with the recorded incidence rates for black 14 to 20 years old in San Francisco in 2000: 7696.8/100,000 chlamydia cases and 2453.8/100,000 gonorrhea cases.35
Except for race, female network members did not differ from female seeds in terms of the demographics, behaviors, or STIs described in Table 1. Male network members, in contrast, were three years older than male seeds (p-value <0.0001), half as likely to be in school (p-value <0.0001), and less likely to use condoms (p-value = 0.0002). Overall, prevalence of STIs in network members was twice that in seeds (18% vs. 9%, p-value = 0.06), with most of the difference among males (17% in male network members vs. 4% in male seeds, p-value = 0.02).
Sexual Network Structure
The 166 seeds were connected, directly or indirectly, to 388 network members in 159 separate sexual network components (Fig. 1). Overall the components were small and acyclic. Three quarters involved three people or fewer and the average component size was 3.5 (SD = 3.1) (Table 2). Although the majority of components were small, five involved more than ten people. The largest of these comprised 25 individuals and accounted for five percent of the unique individuals interviewed or named in the study (Table 3, Fig. 1). Visually, all components, even the largest, were characterized by a linear branching structure (Fig. 1). We found no cycles in any of the sexual network components.
Reflecting the large number of dyads, mean diameter was relatively low (1.8 steps) (Table 2). Mean density was high, reflecting the fact that density is 100% for dyads; among components larger than two, density was 49%. Components larger than two were highly centralized around the node with the highest degree (80%), and slightly more so around the node with the highest betweenness centrality (88%). Degree and betweenness centralization decreased with increasing component size, reflecting the linear branching structure of larger components (Table 3).
Components involving male seeds were less dense (p-value = 0.003), smaller (p-value = 0.003), and less centralized than those involving female seeds (p-value = 0.09) (Table 2), reflecting the fact that female seeds were more likely to be in dyads (64% of components originating from female seeds were dyads vs. 42% among component originating from male seeds).
Sexual Network Position
Mirroring the number of reported partners, network seeds were directly linked to 1.6 partners on average (SD = 1.1) and 69% were linked to only one partner (Table 4). Nearly 30% were linked to someone who was linked to someone else (reach >0).
Male seeds were more central in their networks than females in terms of degree (p-value <0.0001) and betweenness (p-value = 0.0003) but occupied similar positions in terms of reach, information centrality and eigenvector centrality (p-value = >0.15).
Male network members were more central than female network members in terms of information (p-value = .02) and eigenvector centrality (p-value = <0.0001); however female network members had more than twice the reach centrality (p-value = <0.0001), indicating that they were partners of high degree males.
Female seed participants were less central than interviewed female network members in terms of degree centrality (p-value = <0.0001), reach centrality (p-value = <0.0001) and betweenness centrality (p-value = 0.0047), but more central in terms of information (p-value = <0.0001) and eigenvector centralities (p-value = <0.0001) (Table 4). In contrast, male seeds did not significantly differ from interviewed male network members. Noninterviewed network members were less central than interviewed network members by all measures except reach (2.1 vs. 1.7) suggesting that noninterviewed network members were partners of individuals with higher degree centrality.
Figure 1 shows the distribution of infected individuals in the network.
Based on three months' worth of accumulated relationships, the sexual networks of a population-based sample of urban black adolescents living in a neighborhood with moderate endemic rates of STIs were disconnected and linear, with a preponderance of small network components. Three quarters of components were dyads or triads, and 85% of network seeds were in components involving 5 or fewer individuals. At the same time, a few large components were observed: five components comprised more than ten people, with the largest component including 25. No cycles were observed. This low overall network connectivity was observed despite the fact that 13% of participants were infected with chlamydia or gonorrhea or both.
Compared to male seeds, interviewed male network members were older, more likely to be out of school and not to use condoms, and almost twice as likely to be infected with chlamydia or gonorrhea. Besides reflecting the age discordance that is often observed within heterosexual relationships,47–49 the differences between these two groups illustrate the way in which a lower-risk group (the household sample) may be exposed to infection through sexual linkage to a higher risk group (network members), perhaps an example of how infection may spread from the core to the rest of the population.50 However, because snowball sampling will necessarily produce a sample of partners biased toward those connected to seeds with larger degree, the observed differences also may reflect the differences between the characteristics of a random sample of sexually active adolescents and a sample skewed toward those with higher network connectivity.
Consistent with the gender differences in individual-level sexual behaviors observed elsewhere,51 females had lower degree and betweenness centrality than males, and female network seeds led to smaller components than male seeds. However, female and male seeds had similar reach centrality, as well as information and eigenvector centrality. Similarly, male network members had higher degree centrality than female network members, whereas female network members had more than twice the reach centrality. This reflects the fact that although female adolescents may have fewer direct partners, they occupy network positions that are as or more risky as those of males, because of their male partners' network connections. This highlights how network structure may capture STI and HIV risk better than individual-level behaviors alone.1
Although it is difficult to directly compare our results to those of previous sexual network studies because of differences in network sampling methodology and study settings, the structures we observed are very similar to those from previous studies of STI and HIV contact tracing data in endemic settings.17–20,52–54 We found an average component size of 3.5, compared to a range of 2.1 to 3.6 in these contact tracing studies. Fifty-five percent of our components were dyads and 20% triads, compared to 50 to 60% dyads and 20 to 30% triads among the contact tracing studies. Also, like most contact tracing studies, the structures we observed were characterized by linear branching with few to no cycles.
In contrast, the structures we observed differed significantly from those seen in outbreak settings. For example, our participants had much lower degree centrality than adolescents who were part of a syphilis outbreak investigation (1.8 vs. 7.4).21 Similarly, the components we observed were much more linear and acyclic than those involved in a gang-associated chlamydia outbreak.20
The fact that we observed similar structures as those seen in contact tracing studies in endemic settings, despite the fact that we mapped the networks of a household sample of adolescents, may be the result of similar network sampling methodologies. Sampling method can indeed have a large impact on observed network characteristics.55 Furthermore, the technique we used, snowball sampling, is similar to contact tracing in that it depends on partner referrals to delineate networks. Also, the restriction to infected individuals in contact tracing, and the limitation to two waves of snowball sampling in our study, would reduce the likelihood of observing cycles and dense interconnections in both types of studies.
It is also likely however that the similarities between our findings and those of other contact tracing studies in endemic settings reflect an underlying principle about network structure necessary for endemic transmission. For instance, the lack of cycles we observed is considered a hallmark of endemic, as opposed to epidemic, transmission.20,21,23,24,56 Also, the larger components we observed are exactly the kind of large linear components posited to function as core groups in a previous study using chlamydia and gonorrhea contact tracing data.17
Further suggesting that the networks structures we observed may be characteristic of adolescent sexual networks, is the similarity between our findings and those of another general adolescent population network study using Add Health data.27 The studies are difficult to compare because they differed in significant ways: in the Add Health study, networks included both sexual and nonsexual romantic links, were not constrained by a two-wave snowball methodology, and included relationships from the past 18 months (as opposed to 3 in our study). Not surprisingly that study found more large scale connectivity: 50% of the students were part of the same large romantic/sexual network loop, whereas in our study, the largest component comprised only 5% of individuals. However, although the large-scale structure may have differed, the small-scale structure was similar: components exhibited mostly linear branching, and the degree centrality distributions were similar.
Interpretation of our results is complicated by the biases inherent to empirical network studies, namely nonrandom sampling and missing data.55 Because we sampled individuals, not networks, large components were more likely to be observed, leading us to possibly overestimate the proportion of large components. On the other hand, despite the extensive efforts of study staff to locate partners, only 56% of the unique individuals in the network were interviewed. Of note, although this is a low recruitment rate, it does exceed that obtained in two other studies that used contact tracing to interview sex partners, where recruitment rates of nominated partners were 38%54 and 35%.57 However, because of poor recall or reluctance to disclose certain partnerships, an unknowable number of partners may have never been disclosed to study staff. In Addition, by design, only two waves of snowball sampling were carried out. Nondisclosure of partners, inability to recruit partners, and limited snowball sampling waves will all have contributed to underestimates of network connectivity and component size, and may explain the observed lack of cycles.
An additional limitation of our methods is that interviews were not conducted over a span of 16 months. As a result, large components could have been the result of some partners being harder to locate than others because this would enable partners to accumulate additional partners. However, this is unlikely to have played a role as we found no correlation between the time it took to interview partners and component size (p-value for correlation = 0.88 (Table 3)).
Despite these limitations, this study constitutes, to the best of our knowledge, the first nonegocentric study of the sexual networks of a household sample of individuals in the US. One important strength of this study is that sexual network data were collected by interviewing partners and partners' partners, rather than by simply relying on egocentric reports about partner characteristics, as in many other studies of mixing patterns and sexual networks. A further strength is the use of biologic tests for STI infection, instead of relying on self-report. As such, our results confirm, in a different setting, previous studies showing that endemic transmission of chlamydia and gonorrhea may not require highly interconnected and cyclic sexual networks.20,17 Instead, a small number of large linear network structures, formed from the joining together of smaller components, may provide enough connectivity to maintain endemic infection. Furthermore, the differences in STI prevalence between the network seeds and network members may illustrate the way infection may spread from a higher prevalence ‘core' to lower prevalence noncore populations. This suggests that interventions aiming to limit the occurrence of structural elements with high disease transmission potential should focus on the kinds of simple network structures observed in this study.
1. Aral SO. Sexual network patterns as determinants of STD rates: paradigm shift in the behavioral epidemiology of STDs made visible. Sex Transm Dis 1999; 26(5):262–4.
2. Watts D, Strogatz S. Collective dynamics of ‘small-world' networks. Nature 1998; 393:440–442.
3. Liljeros F, Edling CR, Amaral LAN, Stanley HE, Aberg Y. The web of human sexual contacts. Nature 2001; 411(6840):907–908.
4. Doherty IA, Padian NS, Marlow C, Aral SO. Determinants and consequences of sexual networks as they affect the spread of sexually transmitted infections. J Infect Dis 2005; 191 Suppl 1:S42–S54.
5. Garnett GP, Swinton J, Brunham RC, Anderson RM. Gonococcal infection, infertility, and population growth: II. The influence of heterogeneity in sexual behaviour. IMA J Math Appl Med Biol 1992; 9(2):127–144.
6. Garnett GP. An introduction to mathematical models in sexually transmitted disease epidemiology. Sex Transm Infect 2002; 78(1):7–12.
7. Anderson RM, Gupta S, Ng W. The significance of sexual partner contact networks for the transmission dynamics of HIV. J Acquir Immun Defic Syndr 1990; 3(4):417–429.
8. Morris M, Kretzschmar M. Concurrent partnerships and the spread of HIV. AIDS 1997; 11:641–8.
9. Kretzschmar M, Morris M. Measures of concurrency in networks and the spread of infectious disease. Math Biosci 1996; 133(2):165–195.
10. Christley RM, Pinchbeck GL, Bowers RG, et al. Infection in social networks: using network analysis to identify high-risk individuals. Am J Epidemiol 2005; 162(10):1024–1031.
11. Schneeberger A, Mercer CH, Gregson SA, et al. Scale-free networks and sexually transmitted diseases: a description of observed patterns of sexual contacts in Britain and Zimbabwe. Sex Transm Dis 2004; 31(6):380–387.
12. Barabasi A-L, Bonabeau E. Scale-free networks. Sci Am 2003; 288(5):60–69.
13. Ghani AC, Garnett GP. Risks of acquiring and transmitting sexually transmitted diseases in sexual partner networks. Sex Transm Dis 2000; 27(10):579–587.
14. Ghani AC, Swinton J, Garnett GP. The role of sexual partnership networks in the epidemiology of gonorrhea. Sex Transm Dis 1997; 24(1):45–56.
15. Bell DC, Atkinson JS, Carlson JW. Centrality measures for disease transmission networks. Soc Networks 1999; 21(1):1–21.
16. Liljeros F, Edling CR, Amaral LAN. Sexual networks: implications for the transmission of sexually transmitted infections. Microbes Infect 2003; 5(2):189–196.
17. Wylie JL, Jolly A. Patterns of chlamydia and gonorrhea infection in sexual networks in Manitoba, Canada. Sex Transm Dis 2001; 28(1):14–24.
18. Potterat JJ, Phillips-Plummer L, Muth SQ, et al. Risk network structure in the early epidemic phase of HIV transmission in Colorado Springs. Sex Transm Infect 2002; 78 Suppl 1:i159–i163.
19. De P, Singh AE, Wong T, Yacoub W, Jolly AM. Sexual network analysis of a gonorrhoea outbreak. Sex Transm Infect 2004; 80(4):280–285.
20. Potterat JJ, Muth SQ, Rothenberg RB, et al. Sexual network structure as an indicator of epidemic phase. Sex Transm Infect 2002; 78 Suppl 1:i152–i158.
21. Rothenberg RB, Sterk C, Toomey KE et al. Using social network and ethnographic tools to evaluate syphilis transmission. Sex Transm Dis 1998; 25(3):154–160.
22. Rothenberg RB, Potterat JJ, Woodhouse DE, Muth SQ, Darrow WW, Klovdahl AS. Social network dynamics and HIV transmission. Aids 1998; 12(12):1529–1536.
23. Potterat JJ, Rothenberg RB, Muth SQ. Network structural dynamics and infectious disease propagation. Int J STD AIDS 1999; 10(3):182–185.
24. Rothenberg RB, Long DM, Sterk CE, et al. The Atlanta Urban Networks Study: a blueprint for endemic transmission. Aids 2000; 14(14):2191–2200.
25. Friedman SR, Neaigus A, Jose B, et al. Sociometric risk networks and risk for HIV infection. Am J Public Health 1997; 87(8):1289–1296.
26. Rothenberg R, Baldwin J, Trotter R, Muth S. The risk environment for HIV transmission: results from the Atlanta and Flagstaff network studies. J Urban Health 2001; 78(3):419–432.
27. Bearman PS, Moody J, Stovel K. Chains of affection: the structure of adolescent romantic and sexual networks. Am J Sociol 2004; 110(1):44–91.
28. Helleringer S, Kohler HP. Sexual network structure and the spread of HIV in Africa: evidence from Likoma Island, Malawi. Aids 2007; 21(17):2323–2332.
29. Cates W. Estimates of the incidence and prevalence of sexually transmitted diseases in the United States. Am Social Health Association Panel. Sex Transm Dis 1999; 26(4 Suppl):S2–S7.
30. Weinstock H, Berman S, Cates W Jr. Sexually transmitted diseases among Am youth: incidence and prevalence estimates, 2000. Perspect Sex Reprod Health 2004; 36(1):6–10.
31. Centers for Disease Control and Prevention. Tracking the hidden epidemics. Trends in STDs in the United States 2000–2001. Atlanta, GA, Division of STD Control.
32. Laumann EO, Youm Y. Racial/ethnic group differences in the prevalence of sexually transmitted diseases in the United States: a network explanation. Sex Transm Dis 1999; 26(5):250–61.
33. Kottiri BJ, Friedman SR, Neaigus A, Curtis R, Des J. Risk networks and racial/ethnic differences in the prevalence of HIV infection among injection drug users. J Acquir Immun Defic Syndr 2002; 30(1):95–104.
34. Adimora AA, Schoenbach VJ. Social context, sexual networks, and racial disparities in rates of sexually transmitted infections. J Infect Dis 2005; 191 Suppl 1:S115–S122.
35. STD Control Section. San Francisco Sexually Transmitted Disease Annual Summary, 2004–2005. San Francisco, CA, San Francisco Department of Public Health.
36. Ellen JM, Brown BA, Chung SE. et al. Impact of sexual networks on risk for gonorrhea and chlamydia among low-income urban African Am adolescents. J Pediatr 2005; 146(4):518–522.
38. Schachter J, McCormack WM, Chernesky MA et al. Vaginal swabs are appropriate specimens for diagnosis of genital tract infection with Chlamydia trachomatis. J Clin Microbiol 2003; 41(8):3784–3789.
39. Zenilman JM, Miller WC, Gaydos C, Rogers SM, Turner CF. LCR testing for gonorrhoea and chlamydia in population surveys and other screenings of low prevalence populations: coping with decreased positive predictive value. Sex Transm Infect 2003; 79(2):94–97.
40. Rothenberg R, Potterat J, Woodhouse DE, Darrow WW, Muth SQ, Klovdahl AS. Choosing a centrality measure: Epidemiologic correlates in the Colorado Springs study of social networks. Soc Networks 1995; 17:273–297.
41. Borgatti SP. Centrality and network flow. Soc Networks 2005; 27:55–71.
42. SAS. Cary, NC: SAS Institute Inc.
45. Centers for Disease Control and Prevention. STD Surveillance 2006–2007. Atlanta, GA. 2007.
46. Aral SO. Sexually transmitted diseases: magnitude, determinants and consequences. Int J STD AIDS 2001; 12(4):211–215.
47. Begley E, Crosby RA, DiClemente RJ, Wingood GM, Rose E. Older partners and STD prevalence among pregnant African Am teens. Sex Transm Dis 2003; 30(3):211–213.
48. Boyer CB, Shafer MA, Teitle E, Wibbelsman CJ, Seeberg D, Schachter J. Sexually transmitted diseases in a health maintenance organization teen clinic: associations of race, partner's age, and marijuana use. Arch Pediatr Adolesc Med 1999; 153(8):838–844.
49. DiClemente RJ, Wingood GM, Crosby RA, et al. Sexual risk behaviors associated with having older sex partners: a study of black adolescent females. Sex Transm Dis 2002; 29(1):20–4.
50. Yorke JA, Hethcote HW, Nold A. Dynamics and control of the transmission of gonorrhea. Sex Transm Dis 1978; 5(2):51–56.
51. Grunbaum JA, Kann L, Kinchen S, et al. Youth risk behavior surveillance–United States, 2003. MMWR Surveill Summ 2004; 53(2):1–96.
52. Jolly AM, Muth SQ, Wylie JL, Potterat JJ. Sexual networks and sexually transmitted infections: a tale of two cities. J Urban Health 2001; 78(3):433–445.
53. Ghani AC, Ison CA, Ward H, et al. Sexual partner networks in the transmission of sexually transmitted diseases. An analysis of gonorrhea cases in Sheffield, UK. Sex Transm Dis 1996; 23(6):498–503.
54. Day S, Ward H, Ghani A, et al. Sexual histories, partnerships and networks associated with the transmission of gonorrhoea. Int J STD AIDS 1998; 9(11):666–671.
55. Ghani AC, Donnelly CA, Garnett GP. Sampling biases and missing data in explorations of sexual partner networks for the spread of sexually transmitted diseases. Stat Med 1998; 17(18):2079–2097.
56. Cunningham SD, Michaud JM, Johnson SM, Rompalo A, Ellen JM. Phase-specific network differences associated with the syphilis epidemic in Baltimore city, 1996–2000. Sex Transm Dis 2004; 31(10):611–615.
57. Stoner BP, Whittington WL, Hughes JP, Aral SO, Holmes KK. Comparative epidemiology of heterosexual gonococcal and chlamydial networks: implications for transmission patterns. Sex Transm Dis 2000; 27(4):215–223.