SYPHILIS IS AT A low ebb in the United States. In 1997, the prevalence of primary and secondary syphilis, 3.2 per 100,000 population, was the lowest reported since 1941, and is well below the target prevalence of 4.0 per 100,000 population set by the Year 2000 objectives.1 Nevertheless, certain areas, notably in the southeastern United States, continue to experience the substantial occurrence of syphilis. In Georgia, the overall prevalence of primary and secondary syphilis was 7.0 per 100,000 inhabitants in 1997, and the prevalence in Atlanta, 28.4 per 100,000 inhabitants, ranked fourth in the nation among cities with populations greater than 200,000.1 Several zipcode areas within the city of Atlanta have exhibited syphilis rates almost 10 times higher during the 1990s, on the order of 250 per 100,000 inhabitants (Hardy‐Lewis R, unpublished data, 1999).
Areas of high, persistent syphilis endemicity may require nontraditional approaches to sexually transmitted disease (STD) control and prevention. One such approach posits that members of core groups, whose behaviors and relationships involve the active transmission of STDs, are critical to the maintenance of endemicity in definable geographic areas, and are responsible for the spread of STDs outside of such areas.2,3 It has been proposed that identification and intervention with such groups can have a long‐term influence on the overall pattern of disease transmission.4,5
A nontraditional intervention for targeting core groups is the use of social network methods to augment routine partner‐notification activities.6 We previously applied these techniques during an outbreak of syphilis in a suburban Atlanta community.7 Based on these experiences, the Fulton County Department of Health and Wellness conducted a 6‐month pilot project that incorporated social network approaches into routine syphilis‐control activities. The objectives of the study were (1) to examine the feasibility of using social network information for case‐finding activities; (2) to determine the nature and role of groups that may be actively involved in transmission; and (3) to examine the long‐term impact of the pilot effort. In this report, we focus on the first two objectives, and present results from case‐finding efforts and social network analyses.
Based on syphilis case reports (Hardy‐Lewis R, 1999), we chose the Atlanta zipcode area with the highest syphilis prevalence from 1995 to 1997, (Zipcode A) as the study site. The prevalence of syphilis in Zipcode A in 1997, 260 per 100,000 inhabitants, was marginally higher than the combined syphilis rates in five nearby zipcode areas (251 per 100,000 inhabitants from 1995 to 1997), but was the only zipcode area that had not experienced a decline during the period. Zipcode A had a population of 53,783 in 1990 (the most recent year for which accurate data are available), with 46% of its population between the ages of 15 and 44 years. Thirty‐five percent of residents 18 years and older had less than a high school education, and 49% of those 16 years and older were reported as unemployed or not in the labor force. Overall, 32% of the population had incomes below the poverty level, and 56% of children younger than 5 years lived below the poverty level.8
Program Structure and Field Methods
The project began in March 1998 and was completed in October 1998. A field supervisor and two disease investigation specialists were assigned full‐time to the project; a third part‐time investigator joined the group midway through the project. The team was supervised by the STD program director of the Fulton County Department of Health and Wellness, and received consultation from the STD Control Program of the Georgia Division of Human Resources.
The standard approach to syphilis control has evolved over many years, and includes interviewing infected persons at the time of diagnosis (usually in a clinic setting), eliciting contacts, and offering medical evaluation to syphilis‐infected persons. Reinterview is often performed within 2 to 3 days of the initial interview. Sex contacts are the primary focus of these interviews, although occasionally persons are queried about their associates (i.e., nonsexual partners). Preestablished periods for interviewing are used, depending on the stage of syphilis in the infected person. Cases are closed administratively after 30 days of contact tracing. Some programs use the “lot system,” in which cases with contacts in common are managed together.
Standard forms developed by the Centers for Disease Control and Prevention are used to record information about the case interview (CDC 73.54) and for maintaining a field record of activity (CDC 73.2936S). In contrast, the network‐informed approach instituted in this pilot project was street‐centered. The team spent 80% of its time at the locations frequented by clients, and conducted most interviews and follow‐up interviews at those sites; the remaining 20% of the team's time was spent in office‐based and clinic‐related activities.
The daily work pattern included an early morning discussion of the day's work, with group discussion of the ongoing epidemiologic patterns and directions for further investigation. The field supervisor made final decisions on work activities, but field workers had considerable discretion in choices for follow‐up activities. The use of beepers and cellular phones was an important component in maintaining team cohesion and focus. While in the field, investigators divided their time between cultivation of personal connections to the area and searches for particular individuals. Some time was devoted to “passing time” with persons at sites of epidemiologic importance (e.g., crack and alcohol houses, street sites used for congregation, and make‐shift “home” arrangements of homeless persons). All syphilis‐infected persons who reported to the project and uninfected persons who were thought to be potentially important in transmission were interviewed. Most syphilis serologies were drawn in the field.
During interviews, an attempt was made to elicit information regarding all major partners (e.g., sex partners, drug‐using partners, important social contacts). Most persons were queried about partners they may have had during the past year or longer, and all cases were kept open during the project so that persons named as partners would continue to be sought. The traditional standards for confidentiality were maintained; one deviation was that persons who were not known to be direct sexual contacts of a syphilis‐infected person were not told that they had been exposed to infection, and instead were told that infection was highly prevalent in their area and that they could benefit from a “check.”
The standard interview and field record forms were used, but these data were transferred to an electronic format using an entry program constructed in EpiInfo,9 and were processed with SAS10 to examine epidemiologic features of the population and to produce data sets that were then used by programs for social network analysis.11,12 In addition to standard epidemiologic tables and standard measures of program performance, we examined the network properties of the groups identified, including the number of components, the centrality of individuals and the centralization of the network, and the frequency of occurrence of microstructures (i.e., small groups of highly interactive individuals). Network diagrams were constructed at intervals during the project, and were used to inform the team about current structure and to identify subgroups whose exploration might be fruitful.
Population Characteristics and Interviewing Results
The group identified by this project was primarily composed of African Americans (97%). Seventy percent of participants were between 25 and 44 years, with a modal age among those interviewed of 35 to 39 years (27%). Similarly, 62% of contacts identified were 25 to 44 years. Age‐mixing analysis revealed that approximately 50% of all respondents were within 5 years of their contact's age (43% for men, and 59% for women).
Forty‐eight syphilis‐infected persons and 50 uninfected persons were interviewed during the course of the project, and these persons named 396 contacts. The 98 respondents and 396 contacts together represented 381 persons; three infected persons and one uninfected person named no contacts, and three of the infected persons were not named as contacts. At the termination of the project, the cumulative prevalence of syphilis in this population was 12.6%.
On average, infected persons were associated with 5.7 overall contacts and 3.0 sexual contacts. Uninfected persons were associated with 2.4 overall contacts and 1.5 sexual contacts. Thirty‐two percent of the contacts named were identified as sex partners only, 14% as drug partners only, and 10% as both drug and sex partners. The remainder of those named were considered social contacts (i.e., close friends who may share resources) or encounters (i.e., people who were identified as part of the street scene, though not necessarily connected directly to other participants). As expected, the greatest proportion of infected contacts resulted from examination of the sexual contacts of infected persons (23.1%) (Table 1). Between 5% and 6% of nonsexual contacts of infected persons, sexual contacts of uninfected persons, and nonsexual contacts of uninfected persons were also infected. All sexual contacts of persons known to be infected with syphilis and selected persons who were not known to be direct contacts of syphilis‐infected persons received epidemiologic treatment; however, the majority did not, because such treatment is not established as a routine health department procedure.
During routine counseling and HIV testing, which is offered to all clients by the Fulton County Department of Health and Wellness, 24 cases of HIV were found in this population. Seven of the 24 persons infected with HIV were not previously aware of their infection, and 9 were coinfected with syphilis.
Eighty percent of syphilis‐infected persons were interviewed on the day that their case was assigned to a disease‐intervention specialist. Only three people were interviewed more than 1 week after case assignment; one person was found and interviewed 2 months after originally assigned. The majority of clients were reinterviewed formally or in the course of daily street contacts. In accordance with standard guidelines, epidemiologic treatment was given to 54 persons (15.6% of the group). One half of the syphilis‐infected persons in this group (24 persons) were newly brought to treatment (i.e., cases were discovered through the efforts of the project team).
Development and Characteristics of the Network
The team chose to begin with a 42‐year‐old woman who had been diagnosed with syphilis in January 1998, for whom one contact had been identified but not found, and whose case was closed administratively in February 1998. The team reinterviewed this client in March 1998, and identified a small network of four syphilis‐infected persons (including the previously identified contact) and their associated drug and sexual partners (see Figure 1, inset A). The remainder of the network was delineated over the next 6 months using the approach previously described, but “discovery” of the network did not proceed outward from the margins. Rather, groups (i.e., subgraphs in the parlance of social network analysis) were identified and connected over the course of the project. When the project was completed, the 381 persons in the population were grouped in 17 separate connected components, the largest of which contained 278 persons (73% of the group; Figure 1, left). There were two subgroups of 10 persons each, and 14 subgroups were never connected to other components (Figure 1, right).
Network Contributions to Case Discovery
One grouping within the network (Figure 1, inset B) is noteworthy for its demonstration of several features of the network approach. The key figure, the “S” within inset B of Figure 1, was a 45‐year‐old woman known through ethnographic information as a heavy drug user and a commercial sex worker. After syphilis was diagnosed, she was interviewed six times by various team members, but was reluctant to discuss any sexual or drug contacts. Interviews of persons known to be associated with her and of others led to the discovery of an additional seven persons infected with syphilis, none of whom were documented to be directly related to her sexually. This subcluster of cases was also linked, although not by direct sexual contact, with the original group of four cases (Figure 1, inset A).
Most persons were named by multiple sources. Of the 15 syphilis‐infected persons identified by network processes as nonsexual contacts or as contacts to uninfected persons (12.5% of total cases; Table 1), 6 were uniquely identified by these nontraditional means, and 9 were also named as sexual contacts to infected persons. Of the six uniquely identified cases of syphilis, three involved persons who were positive contacts (i.e., persons who were positive contacts to positive contacts of uninfected persons), and in the next generation an additional six persons were positive contacts (i.e., positive contacts to positive contacts of positive contacts of uninfected persons). These nine contacts have been counted among the 30 cases of sexual contacts of infected persons (Table 1); however, these cases originated from uninfected persons one to three generations before. Thus, a network method contributed wholly or partially to the discovery of 24 of the 48 cases.
Using the most conservative definition for network attribution (i.e., cases involving syphilis‐infected persons who were not sexual contacts to infected persons), the proportion of syphilis cases found by nontraditional methods rose to a maximum of 30% (10 cases, 96 contacts), and was 12.5% at the termination of the project (48 cases, 396 contacts).
The network depicted in Figure 1 demonstrates a high degree of centralization, as measured by degree of network centralization (94%) and by Bonacich power centralization (12,228) (Table 2).13 The largest Seidman 2‐core for the network contained 33 interconnected persons.14 Microstructures were common (Table 2); for example, 26 cliques of three persons each (triangles) contained 53 different persons (14% of the total population). More than 2,000 groupings of open triads (2‐plex, n = 3) were found in the group (see Table 2 for definitions of structures). Comparison with previously published data8,15,16 on the occurrence of microstructures in different epidemiologic settings (Table 3) reveals the intensity of interaction within this group. Cliques of three persons are virtually absent from the HIV studies, but are prominent in the Atlanta syphilis‐outbreak studies and in the current pilot project. Similarly, the number of 2‐cliques (n = 3) as a proportion of sample size is dramatically higher in the syphilis studies.
The first objective of this project, to examine the feasibility of using social network information for case‐finding activities, was met. Field personnel, spending the major proportion of their time in the communities of concern, established trust and credibility and were able to elicit network information that facilitated case finding. Because these procedures represent a departure from traditional syphilis‐intervention techniques, many administrative details remain unresolved (e.g., monitoring of work assignments, supervision of field time, evaluation of field workers, processes for assessing day‐to‐day progress, alternative record keeping). Though the computer basis for network analysis is straightforward, it is largely unfamiliar to program staff and to many epidemiologists who work in the STD field, thereby raising the need for additional training and redeployment of staff. In addition, many STD control programs are also responsible for HIV control and prevention activities, and depending on local circumstances, may have additional duties. Although this project appears to be a redirection of currently available resources, program realignment is rarely straightforward, and is never easy. Thus, the feasibility of these activities is well established by this pilot project, but its actual implementation requires further programmatic development.
The second objective, to determine the nature and role of groups that may be actively involved in transmission, has largely been met. We succeeded in identifying an interconnected group of people whose behavior put them at high risk for the transmission of syphilis and, because of our side‐stream activities, of HIV. This group has network characteristics consistent with the active transmission of STDs: a high level of network centralization, a large connected component that includes nearly three fourths of the group, and the frequent occurrence of microstructures that facilitate transmission. The cumulative prevalence of 12.6% cases reinforces the notion that as we move from national to local areas, the focal nature of syphilis transmission becomes strikingly apparent. If we were to suggest, for example, that this method of case discovery has actually provided an estimate of incidence density, the rate in this population would be 25,200 per 100,000 persons per year, which can be compared with the previously cited national estimate of 3.2 per 100,000 persons for 1997. Unfortunately, in an actively evolving endemic situation, it is difficult to distinguish between prevalent cases, which include cases contracted prior to the study period, and new incident infections.
Similarly, it is difficult to determine what proportion of syphilis discoveries can be attributed directly to the network methods. The situation regarding cases found in Figure 1 (inset B) is illustrative: seven additional syphilis cases were uncovered through nonsexual network connections in the vicinity of this person. Clearly, uninfected persons did not transmit the infection to others, but they served to direct us to persons who, by their current and past associations, were at risk. The association with an extremely high risk person (Figure 1, inset B, “S”) helped position the field team for discovery. Although direct attribution is difficult, the nontraditional approach was instrumental in uncovering this subcluster.
Counting and attribution are made difficult by the nature of network activity. For example, we reported the mean number of sexual and nonsexual contacts associated with each person interviewed. These figures are different from those that would be reported using only the interviewee's report of contacts. Our figures are augmented by associations identified from network information, which provide fuller information on the intensity and complexity of interactions. Similarly, the mean degree of 2.05 (Table 2) differs from the contact index because it includes network information on all participants, regardless of whether they were interviewed.
Counting difficulties can be illustrated by our observation that between 5% and 6% of sexual and nonsexual contacts of uninfected persons were infected with syphilis. These 15 infected persons constituted 31% of the syphilis cases uncovered during the project. (Thirty persons were sex contacts who presumably would have been found using traditional methods, and 3 syphilis‐infected persons were never named as contacts, for a total of 48 syphilis diagnoses; Table 1). But the network approach provides multiple sources of information on most participants. Nine of the 15 cases were also identified as sex partners, and 6 of the 15 cases were identified exclusively by the network process. These six cases subsequently led to nine additional persons with syphilis. Thus, network methods contributed to the discovery of 50% of cases, but attribution to traditional or network methods is arguable. Nevertheless, even the most conservative final estimate of 12.5% (the 6 cases found exclusively by nontraditional methods) suggests an important impact of network information.
The overall contribution of network methods to the understanding of the configuration and dynamics of disease transmission is less debatable. Network information relieves the program of its case‐by‐case approach and permits bona fide community evaluation. Program managers are in a position to use this information to determine if they are targeting resources appropriately, and can use network methods as an evaluative tool to determine if the boundaries of a group have been reached (e.g., from the falling impact of network‐based information), or when sexual activity and the types of interactions and microstructures begin to decrease in intensity. The process suggests that a group‐by‐group versus case‐by‐case approach may have substantial value in dealing with areas of persistent syphilis endemicity.
The relationship of network structure to disease transmission remains an area of intense investigation. The findings of this project, when compared with the findings of other network evaluations, suggest that structure plays an important role, and makes a contribution that is distinguishable from the contribution of personal behavior alone. Each of the studies cited (Table 3) included persons with high levels of risk taking, but the results regarding disease transmission are strikingly different. Such conclusions must be tempered by a recognition of the differences in clinical epidemiology between HIV and syphilis. Striking structural differences have been noted when comparing high‐prevalence and low‐prevalence HIV areas17 and in areas with lower syphilis endemicity18 than the area described in this study.
The third objective, to examine the longer term impact of the pilot effort, has not been met. Continued surveillance of Zipcode A with renewed application of more traditional methods may provide insight into the impact of this project. Such information will be crucial in determining the overall value of these methods, and whether the group‐by‐group approach has the potential to eliminate syphilis. If the core‐group hypothesis holds true, intervention in multiple groups of the type uncovered in Zipcode A should have a substantial impact on transmission in these focal areas of persistent endemicity, and may lead to the same disappearance of infection that most areas of the United States now enjoy. This pilot project provides some evidence that the network‐informed approach is worth pursuing.
1. US Department of Health and Human Services, Public Health Service, Division of STD Prevention. Sexually transmitted disease surveillance, 1997. Atlanta: Centers for Disease Control and Prevention, 1998.
2. Thomas JC, Tucker MJ. The development and use of the concept of a sexually transmitted disease core. J Infec Dis 1996; 174(suppl 2):S134-S143.
3. Yorke JA, Hethcote WHJ, Nold A. Dynamics and control of the transmission of gonorrhea. Sex Transm Dis 1978; 5:51–56.
4. Rothenberg RB, Potterat JJ, Woodhouse DE. Personal risk taking and the spread of disease: beyond core groups. J Infec Dis 1996; 174(suppl 2):S144-S149.
5. Wasserheit JN, Aral SO. The dynamic topology of sexually transmitted disease epidemics: implications for prevention strategies. J Infec Dis 1996; 174(suppl 2):S201-S213.
6. Rothenberg RB, Narramore J. The relevance of social network concepts to sexually transmitted disease control. Sex Transm Dis 1996; 23:24–29.
7. Rothenberg RB, Sterk C, Toomey KE, et al. Using social network and ethnographic tools to evaluate syphilis transmission. Sex Transm Dis 1998; 25:154–160.
9. Dean AG, Dean JA, Coulombier D, et al. Epi Info, Version 6: A Word-Processing, Database, and Statistics Program for Public Health on IBM-compatible Microcomputers. Atlanta, GA: Centers for Disease Control and Prevention, 1995.
10. SAS Institute Inc. SAS Language and Procedures: Usage, Version 6. Cary, NC: SAS Institute Inc., 1989.
11. Borgatti SP, Everett M, Freeman L. Ucinet for Windows: Software for Social Network Analysis. Natick, MA: Analytic Technologies, 1999.
12. Krackhardt D, Blythe J, McGrath C. KRACKPLOT: 3.0. Natick, MA: Analytic Technologies, 1999.
13. Wasserman S, Faust K. Social Network Analysis. Melbourne: Cambridge University Press, 1994.
14. Seidman SB. Network structure and minimum degree. Soc Networks 1983; 5:269–287.
15. Rothenberg RB, Sterk C, Long D, Pach A. The interaction of personal risk taking and network structure: why HIV does not propagate. Paper presented at: XIX Sunbelt Social Networks Conference; February 17–20, 1999; Charleston, SC.
16. Rothenberg RB, Potterat JJ, Woodhouse DE, Muth SQ, Darrow WW, Klobdahl AS. Social network dynamics and HIV transmission. AIDS 1998; 12:1529–1536.
17. Rosenberg D, Moseley K, Kahn R, et al. Networks of persons with syphilis and at risk for syphilis in Louisiana: evidence of core transmitters. Sex Transm Dis 1999; 26:108–114.
© Copyright 2000 American Sexually Transmitted Diseases Association
18. Reference not provided.