Secondary Logo

Journal Logo

Original Articles

The Geography of Sexual Partnerships in Baltimore: Applications of Core Theory Dynamics Using a Geographic Information System


Author Information
Sexually Transmitted Diseases: February 1999 - Volume 26 - Issue 2 - p 75-81
  • Free


THE EPIDEMIOLOGY of sexually transmitted diseases is often characterized by high incidence rates within core groups.1,2 In developing countries, core groups are usually defined by occupational risk-such as long-distance truck drivers or commercial sex workers.3,4 In the United States and Western Europe, core groups are often defined behaviorally, such as commercial sex workers or homosexual men, or geographically-as residents of defined areas that typically are socioeconomically depressed. In the U.S. and British urban environment, core areas have been well documented for gonorrhea,5-11 syphilis, and HIV. Within a core area, disease incidence may be 20 to 30 times that of surrounding areas. For example, in Baltimore we have identified discrete core areas where the reported gonorrhea rate in 1994 for persons age 15 to 39 years was 6,821/100,000 for men and 4,341/100,000 for women.12

The transmission efficiency of gonorrhea through unprotected sexual exposure has been empirically estimated to be 30% to 50% per exposure and treatment effectively breaks the transmission chain. Therefore, under most circumstances, introduction of gonorrhea into a community free of infection should not result in a sustained epidemic. Community-based models of STD transmission therefore postulate that core areas or epidemiologically defined core groups are critical to maintaining high rates of gonorrhea.13,14 In these models, cores are characterized by a high transmission density,15 an empirical function that is dependent on characteristics of the sexual network. In contrast to many other infectious diseases, STD transmission is dependent on behavior-i.e., sexual intercourse. Sexually transmitted disease transmission is therefore limited to sexual networks-i.e., the interrelated sexual connections of a defined social group. Previous work has demonstrated that sexual network members have similar characteristics, such as socioeconomic status and ethnicity.16,17

Analyses of sexual networks provide a rational basis for defining transmissibility to susceptible sexual partners.18 A sexual network is the interrelated sexual connections of a defined social group. In a dense sexual network, there are multiple pathways between sexual partners, leading to multiple sources for disease exposure.19

Sexually transmitted disease field workers have suggested in anecdotal reports that sexual partners within core areas had residences within close geographical proximity,20 (Baltimore City Health Department, unpublished data). This finding, if confirmed, would support the concept of geographic density as a characteristic of sexual networks at least in the core areas.

In 1990 to 1992, we conducted a large prospective study of STD incidence in the Baltimore STD clinics that included data on sexual partner characteristics.21-23 In this analysis, we describe the geographical epidemiology of sexual partnerships and compare the geographic epidemiology of sexual partnerships between core and noncore areas in a high prevalence city. Our hypothesis was that the geographic distance between sexual partners within the core areas would be small. This finding would support the concept of dense sexual networks by imputing that partnerships in core areas were drawn from locally restricted populations.


Patient Population

The STD Transmission and Acquisition (TRAC) study, conducted in 1990 to 1992, was a prospective study of the behavioral and clinical determinants of STD acquisition that has been previously described.21-23 Subjects were systematically recruited from patients attending the two public Baltimore STD Clinics. Briefly, after informed consent, a study clinician administered a comprehensive behavioral questionnaire that included elements of a detailed sexual history, knowledge attitudes and behaviors related to STD and STD prevention, previous medical history, and current symptoms. This was followed by comprehensive clinical and laboratory evaluations for STDs including anogenital culture for N. gonorrhoeae, C. trachomatis, and T. vaginalis and serological evaluation for syphilis. Subjects were encouraged to refer their partners for enrollment.

During the study period, 1,389 persons enrolled. Ninety-five percent were blacks, and the median age was 26 years for men and 23 years for women. Gonorrhea was diagnosed in 21% of subjects, and chlamydia was diagnosed in 21%. An STD was diagnosed in 31% of men and 36% of women at the clinic visit. Of the study participants, 572 were individuals who comprised 286 dyad partnerships. Sexual partnerships were considered only if intercourse was reported within 30 days of the index patient's visit and each partner was named by the other. In situations where both partners came to the clinic together, the index patient was defined as the first person interviewed. The members of these partnerships are considered in the analysis presented here.

Mapping Spatial Patterns

As part of study enrollment subjects and partners had to give a valid residential address to the interviewer, which was used to contact the subjects periodically for follow-up visits. These addresses were entered into a geographic information system (GIS). The GIS software package used was Map Info (version 4.0) (Map Info Corporation, Troy, NY), a commercially available GIS software package. The digitized Baltimore regional street maps for use with the software package were obtained from the Baltimore Metropolitan Council and are similar to the U.S. Census TIGER files. Addresses were validated using the database of all known residential addresses in the Baltimore region.

Enrollment of Partnerships

The spatial relationships of partnerships were evaluated by converting the locations of geocoded addresses to standardized State Plane Coordinates (NAD 83). Using the GIS software, the distances between partner residences was calculated as the Euclidean straight line distance between each pair of points and distance was expressed in meters (Figure 1). In Baltimore City, there are few natural barriers (e.g., mountain, large rivers), and therefore the distance between two points is a good approximation of accessibility. Partners who lived together by definition were assigned a distance of zero.

Fig. 1
Fig. 1:
Examples using the geographic information system to determine scalar distances between cases in Baltimore. Three examples are presented. Cases are initially plotted on a digitized map, and then scalar distances are computed. In the three examples, the distances shown are 4.4 kilometers, 12.8 kilometers, and 2.6 kilometers. The 4.4 kilometer partnership has a partner within the Baltimore city limits and a partner within the adjacent Baltimore County.

The patterns of residential distances for males and females were examined separately. We compared the cumulative distribution of distances between residences of each sex living in and outside the core using one-tailed Kolmogorov-Smirnov two-sample tests,24 because under the theory of core maintenance we anticipated a priori that the distances between residences of partners in which the referent individuals lived in the core should be smaller than for those residing outside the core.

To determine if the spatial distribution of partnerships was random relative to the recruited population, a series of Monte Carlo simulations were used to generate the expected distribution of distances. The simulation algorithm was written in a spreadsheet program (Microsoft Excel, Redmond, WA) and is available from the authors upon request. The expected distribution of the distances between members of dyadic partnerships was determined by assigning the residences of the males in the dyad to a randomly selected female partner, until all partnerships were assigned. Distances were calculated between these assigned partnerships and the cumulative distribution function of distances between partnerships calculated. The random assignments were repeated 500 times to generate the expected range of partnership distances. The models were based on selecting random partners for index cases from any of the enrolled members. These distributions were ranked and compared with the observed cumulative distribution function of the distances between the residences of partners.

Defining the Core

The core area for gonorrhea was defined using the reported public and private-sector reported gonorrhea in 1994. This is described in detail elsewhere,11 and has had a stable epidemiologic pattern since 1990 (Baltimore City Health Department, unpublished data). Briefly, the 7,330 cases for 1994 were geocoded as described earlier. The 1990 census data and census tract borders were imported into the software, and census-tract specific rates were calculated. Of 202 census tracts in Baltimore City, 90 had ≥30 cases of gonorrhea reported in 1994. The gonorrhea core was defined as the 13 census tracts with upper quartile gonorrhea rates for persons age 15 to 39 years, corresponding to rates of 4,370 to 6,370 per 100,000. When compared with the overall citywide incidence, the rate ratio of the core rate/citywide was 2.2 to 3.2.


Data from the 286 dyad partnerships with addresses within the Baltimore metropolitan area for both individuals were used for this study. Seventy-two (25.2%) partnerships reported living at the same address. Based on the 1994 gonorrhea incidence rate, the core areas were defined in two inner-city regions of Baltimore City, including six census tracts on the east side and seven on the west side.11

A total of 40 men and 44 women reported residences within the 13 census tracts identified as the core area for gonorrhea transmission in Baltimore,11 whereas 246 males and 242 females resided outside the core. There were no significant differences in the proportion of partners reporting that they lived together inside and outside the core area (Table 1).

Scalar Distance Between Reported Residential Address of Sexual Partners By Gender of Index Case

There was a significant difference for both males and females in the distance to their partner's residence depending on whether they lived in the core area (Figure 2). Individuals who lived in the core area tended to reside significantly closer to their partners, on average, than individuals who lived outside the core. Males living in the core resided within 338.7 meters (median) of their partner (0-1,077.2 meters; 1st-3rd quartiles), with a maximum distance of 12.2 kilometers. Males residing outside the core had a median distance to their partner's residence of 1,956.5 meters (0-5,178.9 meters; 1st-3rd quartiles), with a maximum distance of 24.0 kilometers. This difference was statistically significant (K-S statistic = 0.35; p = 0.0002). Similarly, females residing in the core tended to live closer to their partners (median = 547.4 meters; 1st-3rd quartiles = 0-2,631.9 meters; maximum = 23.1 kilometers) than did females living outside the core (median = 1,805.9 meters; 1st-3rd quartiles = 0-5,101.1 meters; maximum = 21.4 kilometers). However, the difference was only marginally significant (K-S statistic = 0.19; p = 0.076). Spatial patterns did not differ by presence or absence of an STD diagnosis in the index subject or partner.

Fig. 2
Fig. 2:
Cumulative proportion of partners' distances from index case by gender. Note that the core index cases (▪) were much more likely to have a short distance to partners compared with the partners of index cases who lived outside the core (⧖). The pattern was similar for partners of both male index cases (A) and female index cases (B).

Although the distances between residences of core residents and their partners were short, there was substantial contact between core residents and those outside the core. Among males residing in the core, 55.0% (22/40) of their partners also had residences within the core, but 12 of these lived together. If individuals with the same address were excluded, 64.3% (18/28) of women with male partners in the core lived outside the core themselves. The trend was even stronger for women residing in the core (n = 44). Of these, excluding the 12 living with their partners, 71.9% (23/32) of their male partners lived outside the core area.

This high proportion of core individuals with partners outside the core resulted in a relatively low proportion of individuals outside the core whose partners resided inside the core-12.0% (22/184) of males who did not live with their partners and lived outside the core area had a female partner living in the core. Similarly, 10.6% (19/180) of females living outside the core whose partners lived at a different address had male partners residing in the core.

Partnerships were not randomly distributed in space. Compared with the Monte Carlo simulation models, the observed distances between the residents of partnerships was significantly shorter than expected if the partners were selected at random. The cumulative distribution function of the distances between partnerships was much shorter than even the most extreme random pattern generated (Figure 3). In the observed data, including all partnerships (with core and noncore index subjects), the median distance between partner residences was 1,699 meters (0-5,023 meters; 25% and 75% quartiles). In the Monte Carlo models, which were based on selecting random partners for index cases from the enrolled members' addresses, the median ranked model (i.e., of the 500 model curves generated) had a median distance between residences of 5,779 meters, whereas the series with the lowest median distance had a median distance of 4,889 meters, significantly different from the observed data.

Fig. 3
Fig. 3:
Distances between partners under a variety of Monte Carlo simulations. The observed pattern from the actual partner data results in the highest proportion of partners living within close proximity (Curveser_mtch). Of the 500 simulations, the one that is closest to the observed data is Curve mtc_no0. The other models include the simulation that is at the 90th percentile (Curve s70), the median simulation (Curve allrand), and a 10th percentile simulation (Curve s70) (see text). We conclude that partners are not spatially distributed randomly.


Our study demonstrates that within the Baltimore STD core area, sexual partners are recruited from within the immediate neighborhood. Sexual network analysis has become an important tool in studying the epidemiology of sexually transmitted diseases and intravenous drug use. Most analyses have focused on describing sociometric measures of "network closeness" in a traditional sociological context, with each step defined as either a sexual partner interaction (sexual networks),18 a needle sharing interaction (needle sharing networks), or as a friend or associate (social network). In this framework, the number of social or behavioral contacts of an individual and their potential interrelatedness solely defines "dense" networks. For example, in a study of inner-city Baltimore injecting drug users, Latkin25,26 found that the mean size of drug and sex partner networks was 10.3, and the number of ties in the network (density) was associated with more frequent injecting. In Klovdahl's Colorado Springs study,19 a similar analysis found that the size of injecting-sexual networks was an average of 11.3 members. Therefore, social networks may provide a good index of the daily routine for injecting drug users. Friedman27 and colleagues in New York found that the sociometric location (i.e., location within a social network) was associated with risky drug use behaviors, although not condom use. However, because of the expense and training required, social network interviewing is not practical as a widespread intervention strategy.

Klovdahl also described the intervention challenge as finding reliable markers or proxies for the structural or sociometric location of individuals in large social networks to develop network-informed control strategies, a sentiment which was echoed by Friedman28 in reviewing approaches to gonorrhea contact tracing.

Network analyses are extremely valuable in describing risk behaviors and in understanding the social determinants of those behaviors, especially in situations where behavioral interventions are the only potential control strategies. However, for STDs we make here an important distinction between the theoretical definition of the core group, which is a conceptual entity, and an operational definition, which allows the targeting of control interventions. Sexually transmitted disease dynamics are directly related to sexual partner mixing patterns, which in the field intervention setting are particularly difficult and expensive to discern. Garnett17 notes that spatial location is undoubtedly linked to socioeconomic indicators-it provides a useful starting point to crudely estimate mixing patterns between groups. For gonorrhea and syphilis, incident diseases that have well-described focal distributions, this approach may be appropriate.

Our data confirm that sexual partner recruitment in dense urban areas with high STD rates is truly a neighborhood phenomenon. In Baltimore, where inner-city communities are highly segregated on the basis of economic and racial characteristics, this pattern produces assortative mixing of sexual partners within the core. This concurs with Garnett's17 analysis of the Colorado Springs contact tracing data; using census tract definitions, he found that assortative mixing is a key factor in the endemic persistence of gonorrhea.

Nevertheless, core members often had substantial contact outside the core-especially females. This pattern supports Rothenberg's original hypothesis5 that maintaining a community-wide epidemic is dependent on core areas functioning as a reservoir for infecting other areas.

No previous study to our knowledge has evaluated the sociogeographical characteristics of sexual contacts. Potterat,6 also in Colorado Springs, reported that 51% of gonorrhea in the city came from four census tracts, in other words, a core pattern. Although their further work extensively documented the social network characteristics of this population, they did not describe the geography of sexual partnerships.

Baltimore provides a unique setting to study this phenomenon because of its high population density and geographic compactness. Population density in Baltimore is 9,109 per square mile (1990 census),29 and within the core areas, the density is higher because much of the housing stock is either row houses or apartment complexes. This density in part explains the remarkably short distances between the residences of sexual partners. Because there are no geographic barriers in the central city that would impede movement, such as waterways or mountains, we believe that the scalar distance measures represent accurate representations of ranging distances used to recruit sexual partners. For example, this type of situation would need to be modified in New York or Pittsburgh, where natural barriers such as waterways or mountains may make the scalar distance inaccurate when used alone.

Our data are subject to a number of biases. For example, because index patients and partners were recruited from the STD clinic, the population served by those clinics will be overrepresented. These persons are predominantly poor, minority residents of inner-city Baltimore. In our data, this bias cannot be controlled. However, the overall epidemiology of both syphilis and gonorrhea in Baltimore, including cases reported from the private sector, closely follow the catchment patterns of the STD clinic. Therefore, we do not believe that this bias would markedly affect our results. Second, the core area represents the most impoverished sections of Baltimore. Persons living in the adjacent and peripheral areas may have higher income and therefore increased access to transportation, especially automobiles. This pattern also correlates closely with housing and population density.

Nevertheless, our data provide intriguing insights into the microepidemiology of STDs and sexual partner recruitment patterns in a dense urban center that offer opportunities for new intervention strategies. Rothenberg18 and Potterat30 have suggested that, in contrast to the current "disease specific approach," intervention programs adopt a "neighborhood approach" using an ethnographically directed evaluation of the local setting. These approaches can be further strengthened by the incorporation of noninvasive, urine-based nucleic acid diagnostics for gonorrhea and chlamydia testing into a field setting that will use sociogeographic risk as the trigger for interventions. For example, approaches could use mobile clinics or vans that could collect population-based samples from areas identified to be at high risk for disease. Geographic approaches to disease intervention may be viable alternatives to traditional case-by-case contact tracing and warrant further, more formalized evaluation.


1. Thomas JC, Tucker MJ. The development and use of the concept of a sexually transmitted disease core. J Infect Dis 1996; 174(Suppl 2):S134-S143.
2. Woodhouse MEJ, Dye C, Etard J-F, et al. Heterogeneities in the transmission of infectious agents: implications for the design of control programs. Proc Natl Acad Sci USA 1997; 94:338-342.
3. Bwayo J, Plummer FA, Omari M, et al. Human immunodeficiency virus infection in long-distance truck drivers in east Africa. Arch Intern Med 1994; 154:1391-1396.
4. Moses S, Plummer FA, Ngugi EN, Nagelkerke NJD, Anzala AO, Ndinya-Achola JO. Controlling HIV in Africa: effectiveness and cost of an intervention in a high-frequency STD transmitter core group. AIDS 1991; 5:407-411.
5. Rothenberg RB. The geography of gonorrhea: empirical demonstration of core group transmission. Am J Epidemiol 1983; 117:688-693.
6. Potterat JJ, Rothenberg RB, Woodhouse DE, Muth JB, Pratts CI, Fogle JS II. Gonorrhea as a social disease. Sex Transm Dis 1985; 12:25-32.
7. Zenilman JM, Bonner M, Sharp KL, Rabb J, Alexander R, and The Disease Intervention Staff of The Dade County Public Health Unit. Penicillinase-producing Neisseria gonorrhoeae in Dade County, Florida: evidence of core-group transmitters and the impact of illicit antibiotics. Sex Transm Dis 1988; 15:45-50.
8. Rice RJ, Roberts PL, Handsfield HH, Holmes KK. Sociodemographic distribution of gonorrhea incidence: implications for prevention and behavioral research. Am J Public Health 1991; 81:1252-1257.
9. Lacey CJN, Merrick DW, Bensley DC, Fairley I. Analysis of the sociodemography of gonorrhoea in Leeds, 1989-93. Br Med J 1997; 314:1715-1718.
10. Low N, Daker-White G, Barlow D, Pozniak AL. Gonorrhoea in inner London: results of a cross sectional study. Br Med J 1997; 314:1719-1723.
11. Ellen JM, Hessol NA, Kohn RP, Bolan GA. An investigation of geographic clustering of repeat cases of gonorrhea and chlamydial infection in San Francisco, 1989-1993: evidence for core groups. J Infect Dis 1997; 175:1519-1522.
12. Becker KM, Glass GG, Brathwaite W, Zenilman JM. The geographic epidemiology of gonorrhea in Baltimore using a geographic information system. Am J Epidemiol 1998; 147:709-716.
13. Yorke JA, Hethcote HW, Nold A. Dynamics and control of the transmission of gonorrhea. Sex Transm Dis 1978; 5:51-56.
14. Anderson RM, May RM. Infectious Diseases of Humans: Dynamics and Control. Oxford: Oxford University Press, 1992.
15. Rothenberg RB, Potterat JJ. Temporal and social aspects of gonorrhea transmission: the force of infectivity. Sex Transm Dis 1988; 15:88-92.
16. Catania JA, Binson D, Stone V. Relationship of sexual mixing across age and ethnic groups to herpes simplex virus-2 among unmarried heterosexual adults with multiple partners. Health Psychol 1996; 15:362-370.
17. Garnett GP, Anderson RM. Contact tracing and the estimation of sexual mixing patterns: the epidemiology of gonococcal infections. Sex Transm Dis 1993; 20:181-191.
18. Rothenberg RB, Narramore J. The relevance of social network concepts to sexually transmitted disease control. Sex Transm Dis 1996; 23:24-29.
19. Klovdahl AS, Potterat JJ, Woodhouse DE, Muth JB, Muth SQ, Darrow WW. Social networks and infectious disease: the Colorado Springs study. Soc Sci Med 1994; 38:79-88.
20. New York City Health Department STD Control Bureau. Final report of the PPNG Targeted Intervention Program (TIP) study, 1988.
21. Ellish N, Weisman CS, Celentano D, Zenilman JM. Reliability of partner reports of sexual histories in a heterosexual STD clinic population. Sex Transm Dis 1996; 23:446-452.
22. Thomas DL, Zenilman J, Shih JW, et al. Evidence of sexual transmission of hepatitis C to females attending Baltimore sexually transmitted disease clinics-an analysis of 309 sexual partnerships. J Infect Dis 1995; 171:768-775.
23. Zenilman JM, Weisman CS, Rompalo AM, et al. Condom use to prevent incident STDs: the validity of self-reported condom use. Sex Transm Dis 1995; 22:1-7.
24. Conover WJ. Practical Nonparametric Statistics. New York: John Wiley & Sons, 1971:462.
25. Latkin C, Mandell W, Oziemkowska M, et al. Using social network analysis to study patterns of drug use among urban drug users at high risk for HIV/AIDS. Drug Alcohol Depend 1995; 38:1-9.
26. Suh T, Mandell W, Latkin C, Kim J. Social network characteristics and injecting HIV-risk behaviors among street injection drug users. Drug Alcohol Depend 1997; 47:137-143.
27. Friedman SR, Neaigus A, Jose B, et al. Sociometric risk networks and risk for HIV infection. Am J Public Health 1997; 87:1289-1296.
28. Friedman SR. Network methodologies, contact tracing, gonorrhea, and human immunodeficiency virus. Sex Transm Dis 1996; 23:523-525.
29. Baltimore City Department of Planning. 1990 Community Profiles. Baltimore City Department of Planning, Baltimore, 1992.
30. Potterat JJ. Sociogeographic space and sexually transmitted diseases in the 1990s. Todays Life Sci 1992; December: 16-23.
© Copyright 1999 American Sexually Transmitted Diseases Association