Gindi, Renee M. PHD*; Sifakis, Frangiscos PHD†; Sherman, Susan G. PHD†; Towe, Vivian L. PHD†; Flynn, Colin ScM‡; Zenilman, Jonathan M. MD§
Most people select sexual partners who are similar to themselves,1 often referred to as assortative mixing.2,3 Mathematical models indicate that assortative mixing within core groups may be responsible for the continued endemicity of sexually transmitted diseases (STD).4,5 Assortative mixing within geographic areas, or spatial assortativity, may help to maintain the high infection prevalence in certain neighborhoods, as choosing a sex partner from within a high human immunodeficiency virus/sexually transmitted disease (HIV/STD) prevalence neighborhood increases the probability of selecting an infected partner.6–8
Since the 1950s, researchers have found that people tend to choose marriage partners who live nearby.9–12 More recent research has estimated geographic proximity between sexual partners from high-risk populations. At the Baltimore STD clinic, patients from areas with high gonorrhea prevalence lived closer to their partners than patients from lower prevalence areas, on average just a few 100 m apart.13 Individuals with repeat gonorrhea infection were more likely to live near each other than were nonrepeaters.14 Spatial assortativity has been observed in areas with less housing density than Baltimore. Studies of STD clinic patients, commercial sex workers, and injection drug users in Colorado Springs found that people who engaged in exchange sex and injection drug use lived closer to their partners than would be expected in the general population.15
This body of literature is frequently cited as evidence that people choose sex partners from the neighborhoods where they live, but the practices of STD clinic patients may not be generalizable to other populations at risk of HIV/STD. This research is particularly salient in light of the focus on neighborhood-level predictors of high-risk sexual behavior16–21 as spatial assortativity could impact these associations. Using geographical, behavioral, and demographic survey data, we determined prevalence and correlates of spatial assortativity in a non-STD clinic population.
MATERIALS AND METHODS
The Baltimore site of the National HIV Behavioral Surveillance system, locally called the Behavioral Surveillance Research Study (BESURE), is an ongoing collaboration between Centers for Disease Control and Prevention, the Maryland Department of Health and Mental Hygiene, and the Johns Hopkins Bloomberg School of Public Health. The 2007 cycle of BESURE aimed to measure the prevalence of HIV and HIV-related risk behaviors among adults at high risk of heterosexual HIV infection in Baltimore. The study population of the 2007 BESURE cycle has been described elsewhere.22,23 Briefly, census tracts were categorized as being in the top quintile of tracts for poverty and HIV/AIDS rates in the Baltimore-Towson Metropolitan Statistical Area (Baltimore MSA). Within 10 of these census tracts, recruitment venues (e.g., retail stores, public areas) were selected. Venue patrons were approached and screened for study eligibility.24 Participants eligible for the 2007 BESURE cycle were between 18 and 50 years of age; residents anywhere in the Baltimore MSA (not only in the 10 census tracts); male or female (not transgender); reported vaginal or anal sex with a person of the opposite sex in the past 12 months; and had the ability to complete the interview in English. We restricted this analysis to heterosexual partnerships of those participants residing in Baltimore City, a geographic subset of those eligible for the study. Of the 332 participants enrolled, 307 (92%) met the criteria for this analysis, contributing information on 776 heterosexual partnerships. Geographic data were available from 273 individuals (89%) on 510 partnerships (66%). This study was approved by the institutional review boards at the participating agencies.
Participants provided information on their 5 most recent sexual partners in the past 12 months. All data about sexual partnerships (residential, demographic, and behavioral) in this study were provided by participants; partners were not interviewed. Sexual partnerships were classified by participants as main, casual, or exchange (i.e., sex in exchange for money or drugs). Participants were asked whether they had sex with other people during the partnership as a measure of sexual concurrency behavior. Participants were asked whether they suspected that their partner was having sex with other people as a measure of suspected partner concurrency. Unprotected intercourse in a partnership was defined as reporting no condom use during the last vaginal or anal sexual intercourse with that partner.
Participants were asked to classify their own race and race of each partner. Race was classified into mutually exclusive categories. A binary variable indicated whether participant's and partner's races were the same. Participants were asked whether partners were older, younger, or the same age. If participants selected “older/younger,” they were asked for the partner's age. Partners were considered to be of the same age if (1) participants reported that partner was the “same age” or (2) partner age was within 5 years of participant age (after Laumann et al25). Participants were considered to live in poverty based on their reported income and the number of dependents supported.26
Area-level measures of poverty were created using data from the US Census. Poverty core areas were classified as those census tracts (n = 48) in the top quartile of poverty for Baltimore City in 2006.27 Area-level measures of HIV risk were created using HIV/AIDS case rates. HIV core areas were classified as those census tracts (n = 48) in the top quartile of heterosexually transmitted HIV/AIDS case rates for Baltimore City in 2006 (Maryland Department of Health and Mental Hygiene, unpublished data) (Figure 1).
Study protocols did not allow for the collection of geocodable addresses of participants. Participants were asked to point to their residences on detailed Baltimore City maps with census boundaries. The census tract of each participant's residence was ascertained and recorded. The same procedure (point, ascertain, record) was repeated for each of the participant's 5 most recent partners. Using ArcGIS, participant and partner census tracts were plotted using Baltimore City shapefiles from the US Census.29 Spatial assortativity was defined as partner residence within the same or adjacent census tracts. In a more subjective measure, participants were also asked whether they met each partner “in the neighborhood where they live,” with the definition of “in the neighborhood” left to participants.
We calculated the frequency, mean, standard deviation, median, and interquartile range of study measures among the 307 individuals and 776 partnerships. The partnership was the primary unit of analysis, though participant-level statistics are also presented.
As “Don't Know” was an acceptable response option for partner's residence, we could only calculate frequency of spatially assortative partnerships among those 510 partnerships with data on partner residence. We calculated associations with completeness of tract data and spatial assortativity using the chi-squared test, Cochran-Armitage trend test, and Fisher exact test for significance where appropriate. We calculated assortativity as percent within-group choice, expressed as the proportion of people with a certain characteristic whose sexual partners have the same characteristic.25
Regression models used partnerships as the unit of analysis, with spatial assortativity as the outcome variable. We calculated adjusted prevalence ratios (PR) and 95% confidence intervals (CI) in multivariate models in light of the fact that our primary outcome was not rare.30 We were primarily interested in the effects of area-level measures (HIV core, poverty core) on spatial assortativity and identifying factors that may confound these relationships. To account for correlations between multiple partnerships within a single participant, we used generalized estimating equations. We explored interactions between gender and partnership types, and considered the impact of defining spatial assortativity as partners residing in the same census tract. Using multiple data imputation in SAS v.9.1 (Cary, NC), we conducted a sensitivity analysis to determine the impact of missing partner spatial data on our study results.
Of the 307 participants in this analysis, a majority were black (Table 1). Participants were evenly divided by gender and had a mean age of 33 years. One-third (34%) lived in the census tracts in the top quartile of HIV/AIDS case rates in Baltimore City. Almost half (43%) reported 3 or more partners in the past 12 months (median: 2, interquartile range: 1–4). Casual partnerships were most common (44%), with 39% of partnerships classified as main and 16% as exchange (Table 2). Participants reported concurrency in 67% of partnerships and suspected partner concurrency in 56% of partnerships.
Two-thirds of partnerships had geographical partner data (n = 510) and 273 participants (89%) provided geographical data on at least 1 partner. Almost half of these partnerships (238, or 47%) were spatially assortative with 69% of these within the same census tract. Almost two-thirds (65%) of the 273 participants providing partner data reported at least 1 spatially assortative partnership, and 40% of participants reported that all partnerships were spatially assortative.
Spatial, racial, and age assortativity were calculated by gender and partnership type (Table 3). Spatial assortativity was most common in women's exchange partnerships (74%) but less common among all types of men's partnerships. Partnerships in this population were highly assortative by race. Approximately half of partnerships were assortative by age within 5 years.
Participants residing in HIV core areas were significantly more likely to have spatially assortative partnerships than participants from the noncore areas (Table 4). Participants reporting 1 or 2 sexual partners were significantly more likely to have spatially assortative partnerships compared to participants reporting 3 or more partners. Exchange partnerships were significantly more likely to be spatially assortative than main or casual partnerships, though this differed by gender. Spatial assortativity was significantly more common among women's exchange partnerships than men's exchange partnerships (74% and 43%, respectively, P = 0.01). Unprotected intercourse was marginally more likely in spatially assortative partnerships (P = 0.09).
After accounting for the correlation between multiple partnerships per participant with generalized estimating equation, and adjusting for partnership type, number of partners, and unprotected intercourse, participants from HIV core areas were marginally more likely to have spatially assortative partnerships than participants from noncore areas (PR: 1.2, 95% CI: 1.0–1.4). Unprotected intercourse was significantly more likely among spatially assortative partnerships. Concurrency, racial assortativity, and age assortativity were not significantly associated with spatial assortativity. When we considered spatial assortativity as partners residing in the same census tract, we found that HIV core area residence remained associated with spatial assortativity (PR: 1.3, 95% CI: 1.0–1.7), though unprotected intercourse and exchange sex were no longer strongly associated with spatial assortativity.
Partners' census tracts were missing for 34% of the 776 partnerships eligible for this analysis. Several factors were associated with missing partner data. Exchange partnerships were the most likely to have missing data, followed by casual and main partnerships (51%, 42%, and 19%, respectively). When participants met their partners in their neighborhoods, only 24% of partner census tracts were missing, compared to 46% when they did not meet in the neighborhood. Less recent partnerships were more likely to have missing data than most recent partnerships. We imputed the missing assortativity data using 3 variables identified above, plus gender. HIV core area residence remained associated with spatial assortativity (PR: 1.2, 95% CI: 1.0–1.5), though unprotected intercourse was no longer strongly associated with spatial assortativity.
In a population recruited through venue-based sampling, almost half reported choosing spatially assortative partners. Participants who lived in the HIV core areas were more likely to choose spatially assortative partners than residents of non core areas after adjusting for partnership type, gender, and number of partners. This relationship persisted in sensitivity analyses. Women who engaged in exchange sex were most likely to report spatial assortativity. Our study confirms Zenilman et al previous work, which was limited to STD clinic patients.13 This remained the case even when sampling from a non-STD clinic population and using census tract data rather than geocodable address.
Choosing a sex partner from a high HIV prevalence area puts an individual at high risk of selecting an infected partner, with exchange sex and unprotected sex increasing the probability of HIV acquisition even further. Spatial assortativity was common but not universal in this population, indicating that partners' residential exposures should be measured when studying the effect of neighborhood-level factors on sexual behavior.
Contextual factors that may impact geography of partner selection include access to public transportation, segregation of residential housing, natural boundaries like rivers and roads, and access to specific social institutions that create opportunities for meeting.31 Some of the differences in spatially assortative partnering between the HIV core and noncore areas could reflect differential access to transportation, with car ownership allowing more latitude in partner selection. Although we did not ask about transportation access, Census data indicate that household car ownership in the HIV core census tracts was significantly lower than in the noncore census tracts (45% vs. 69%, P < 0.0001).27 Although differences in car ownership may be related to poverty, residence in high poverty areas does not confound the observed relationship between HIV core area residence and spatially assortative partnerships in this analysis. Residents living in census tracts in the top quartile of poverty in Baltimore City were no more likely than the residents of other census tracts to report spatially assortative partnerships (52% vs. 48%, P = 0.14).
Racial assortativity was common in this population, with black women consistently displaying the most highly assortative partner selection.25,32–38 The near universality of racial assortativity in all age groups and partnership types indicates that it should be taken into account in studies of partner selection. We found that this population was moderately assortative by age, somewhat lower than the 75% to 83% assortativity observed by Laumann et al among men and women and 75% assortativity calculated by Darroch et al among women nationwide.25,39 Age assortativity may not play a large role in HIV/STD risk among adults; whereas disassortative mixing by age increases risks of STD among adolescent girls,7,39–44 the same association is not seen among adult women.7,43,45
This study is subject to several limitations, including missing data and imprecise measurement. We collected data on only the 5 most recent sexual partners. Of the 246 participants who knew where their partners lived, 11% had more than 5 partners. A large proportion of these (46%) had 6 to 7 partners. Given that participants with more partners are less likely to have spatially assortative partnerships (Table 4) and that spatial assortativity is less likely for less recent partners (data not shown), we expect that partnerships 6 and 7 would not be spatially assortative. Most (69%) participants with 6 to 7 partners lived outside of HIV core areas. These missing partnerships may have caused us to underestimate the true association between HIV core area and spatial assortativity.
Also of note is the proportion of participants who were unable or unwilling to report where their sex partners lived. This may be due to not knowing where their partners lived (particularly in the case of casual and exchange partnerships) or not remembering where sex partners lived (particularly in the case of less recent partners or greater number of recent partners). Explicit refusals constituted 1% of missing data on partnerships, with the remainder presumably from participants who did not know their partners' residences. Although the high proportion of missing data are consistent with other estimates in the literature on missing partner data in partner notification,46,47 the missing outcome data increases the uncertainty in our estimates. We were, however, able to assess the impact of the missing data in our regression models using multiple imputation and other sensitivity analyses. The sensitivity analyses generally confirmed the relationship of HIV core areas and spatially assortative partnerships, though associations with other factors (e.g., unprotected intercourse, exchange sex) were less consistent. Given previous work on the biases because of missing data in sexual network analysis, further research on missing spatial data in the analysis of partner selection data are needed.48
Without geocodable address, we could not calculate Euclidian distances between partners to check consistency of our results with the previous literature.13,15 Study protocols required the use of census tract rather than census block group to protect participant confidentiality. Self-reported data on sensitive subjects may be subject to social desirability bias. Although participants may have felt that they should not report risk behaviors like concurrent partnerships, the prevalence of these behaviors was so high that this bias is unlikely.
Our study used a venue-based sampling strategy to locate the general population at high risk for HIV/STD, increasing its generalizability. We were able to examine the relationships between several types of assortativity in a high-risk population, including demographic and spatial assortativity. We were able to link spatial proximity, and reported risk behaviors, and found that residence in an HIV core area was independently associated with increased likelihood of spatial assortativity. This study therefore has implications for future public health practice and research on geographical and contextual factors in HIV/STD prevention.
Evidence that high HIV prevalence areas also have geographically denser sexual networks could help health departments decide to continue targeting screening to these core areas, rather than taking a generalized screening approach. Prevention programs could also capitalize on the spatial proximity of partners in high-prevalence areas by supplementing their individual-focused prevention messages with social marketing campaigns (e.g., billboards, leaflets).49
Shared geographical space may be the underlying mechanism by which social norms are developed in a network.50 New studies assessing the impact of social norms and attitudes on sexual behavior should take geographical context into account. Structural influences on local sex partner availability (i.e., because of high levels of incarceration) may strongly impact partner selection patterns (i.e., concurrency) in areas where spatial assortativity is common.51 We would expect to see particularly strong associations between contextual factors and behavior when both partners are influenced by the same environment.
These data were collected in the Baltimore City-specific questionnaire of the National HIV Behavioral Surveillance system. Future waves of this national survey should add items on partner's residence in regards to comparing spatial assortativity across regions and populations at high risk for HIV/STD infection. In 1999, Zenilman et al recommended targeted neighborhood approaches to screening and intervention in urban core areas. It appears that these recommendations are still warranted.
1.Watson D, Klohnen EC, Casillas A, et al. Match makers and deal breakers: Analyses of assortative mating in newlywed couples. J Pers 2004; 72:1029–1068.
2.Doherty IA, Padian NS, Marlow C, et al. Determinants and consequences of sexual networks as they affect the spread of sexually transmitted infections. J Infect Dis 2005; 191(suppl 1):S42–S54.
3.Renton A, Whitaker L, Ison C, et al. Estimating the sexual mixing patterns in the general population from those in people acquiring gonorrhoea infection: Theoretical foundation and empirical findings. J Epidemiol Community Health 1995; 49:205–213.
4.Hethcote HW, Yorke JA. Gonorrhea Transmission Dynamics and Control. Lecture Notes in Biomathematics, Vol 56. Heidelberg, Germany: Springer-Verlag, 1984.
5.Newman MEJ. Mixing patterns in networks. Phys Rev 2003; 67:1–13.
6.Rothenberg R. Maintenance of endemicity in urban environments: A hypothesis linking risk, network structure and geography. Sex Transm Infect 2007; 83:10–15.
7.Aral SO, Hughes JP, Stoner B, et al. Sexual mixing patterns in the spread of gonococcal and chlamydial infections. Am J Public Health 1999; 89:825–833.
8.Jennings J, Glass B, Parham P, et al. Sex partner concurrency, geographic context, and adolescent sexually transmitted infections. Sex Transm Dis 2004; 31:734–739.
9.Peach C. Homogamy, propinquity and segregation: A re-evaluation. Am Sociol Rev 1974; 39:636–641.
10.Ramsøy NR. Assortative mating and the structure of cities. Am Sociol Rev 1966; 31:773–786.
11.Eckland BK. Theories of mate selection. Eugen Q 1968; 15:71–84.
12.Katz AM, Hill R. Residential propinquity and marital selection: A review of theory, method, and fact. Marriage Fam Living 1958; 20:27–35.
13.Zenilman JM, Ellish N, Fresia A, et al. The geography of sexual partnerships in Baltimore: Applications of core theory dynamics using a geographic information system. Sex Transm Dis 1999; 26:75–81.
14.Bernstein KT, Curriero FC, Jennings JM, et al. Defining core gonorrhea transmission utilizing spatial data. Am J Epidemiol 2004; 160:51–58.
15.Rothenberg R, Muth SQ, Malone S, et al. Social and geographic distance in HIV risk. Sex Transm Dis 2005; 32:506–512.
16.Hogben M, Leichliter JS. Social determinants and sexually transmitted disease disparities. Sex Transm Dis 2008; 35(suppl 12):S13–S18.
17.Ellen JM, Jennings JM, Meyers T, et al. Perceived social cohesion and prevalence of sexually transmitted diseases. Sex Transm Dis 2004; 31:117–122.
18.Holtgrave DR, Crosby RA. Social capital, poverty, and income inequality as predictors of gonorrhoea, syphilis, chlamydia and AIDS case rates in the United States. Sex Transm Infect 2003; 79:62–64.
19.Adimora AA, Schoenbach VJ, Martinson FE, et al. Social context of sexual relationships among rural African Americans. Sex Transm Dis 2001; 28:69–76.
20.Latkin CA, Curry AD, Hua W, et al. Direct and indirect associations of neighborhood disorder with drug use and high-risk sexual partners. Am J Prev Med 2007; 32(suppl 6):S234–S241.
21.Cohen D, Spear S, Scribner R, et al. “Broken windows” and the risk of gonorrhea. Am J Public Health 2000; 90:230–236.
22.Towe V, Sifakis F, Gindi R, et al. Prevalence of HIV infection and sexual risk behaviors among individuals having heterosexual sex in low income neighborhoods in Baltimore, MD: The BESURE Study. J Acquir Immune Defic Syndr 2010; 53:522–528.
23.Gallagher KM, Sullivan PS, Lansky A, et al. Behavioral surveillance among people at risk for HIV infection in the US: The national HIV behavioral surveillance system. Public Health Rep 2007; 122(suppl 1):32–38.
24.MacKellar D, Valleroy L, Karon J, et al. The young men's survey: Methods for estimating HIV seroprevalence and risk factors among young men who have sex with men. Public Health Rep 1996; 111(suppl 1):138–144.
25.Laumann EO, Gagnon JH, Michael RT, et al. The Social Organization of Sexuality: Sexual Practices in the United States. Chicago, IL: University of Chicago Press; 1994.
26.Department of Health and Human Services. Annual Update of the HHS Poverty Guidelines. Fed Regist 2007; 72:3147–3148.
30.McNutt LA, Wu C, Xue X, et al. Estimating the relative risk in cohort studies and clinical trials of common outcomes. Am J Epidemiol 2003; 157:940–943.
31.Laumann E, Ellingson S, Mahay J, et al. The Sexual Organization of the City. Chicago, IL: University of Chicago Press, 2004.
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; 26:250–261.
33.Morris M, Kurth AE, Hamilton DT, et al. Concurrent partnerships and HIV prevalence disparities by race: Linking science and public health practice. Am J Public Health 2009; 99:1023–1031.
34.Ford K, Norris A. Patterns of union formation among urban minority youth in the United States. Arch Sex Behav 2000; 29:177–188.
35.Stoner BP, Whittington WL, Hughes JP, et al. Comparative epidemiology of heterosexual gonococcal and chlamydial networks: Implications for transmission patterns. Sex Transm Dis 2000; 27:215–223.
36.Lee JK, Jennings JM, Ellen JM. Discordant sexual partnering: A study of high-risk adolescents in San Francisco. Sex Transm Dis 2003; 30:234–240.
37.Ford K, Sohn W, Lepkowski J. American adolescents: Sexual mixing patterns, bridge partners, and concurrency. Sex Transm Dis 2002; 29:13–19.
38.Ford K, Sohn W, Lepkowski J. Characteristics of adolescents' sexual partners and their association with use of condoms and other contraceptive methods. Fam Plann Perspect 2001; 33:100–105.
39.Darroch JE, Landry DJ, Oslak S. Age differences between sexual partners in the United States. Fam Plann Perspect 1999; 31:160–167.
40.Manlove J, Terry-Humen E, Ikramullah E. Young teenagers and older sexual partners: Correlates and consequences for males and females. Perspect Sex Reprod Health 2006; 38:197–207.
41.Begley E, Crosby RA, DiClemente RJ, et al. Older partners and STD prevalence among pregnant African American teens. Sex Transm Dis 2003; 30:211–213.
42.Harawa NT, Greenland S, Cochran SD, et al. Do differences in relationship and partner attributes explain disparities in sexually transmitted disease among young white and black women? J Adolesc Health 2003; 32:187–191.
43.Kraut-Becher JR, Aral SO. Patterns of age mixing and sexually transmitted infections. Int J STD AIDS 2006; 17:378–383.
44.Ford K, Lepkowski JM. Characteristics of sexual partners and STD infection among American adolescents. Int J STD AIDS 2004; 15:260–265.
45.Stein CR, Kaufman JS, Ford CA, et al. Partner age difference and prevalence of chlamydial infection among young adult women. Sex Transm Dis 2008; 35:447–452.
46.Mathews C, Coetzee N, Zwarenstein M, et al. A systematic review of strategies for partner notification for sexually transmitted diseases, including HIV/AIDS. Int J STD AIDS 2002; 13:285–300.
47.Peterman TA, Toomey KE, Dicker LW, et al. Partner notification for syphilis: A randomized, controlled trial of three approaches. Sex Transm Dis 1997; 24:511–518.
48.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; 17:2079–2097.
49.Cohen DA, Wu SY, Farley TA. Structural interventions to prevent HIV/sexually transmitted disease: Are they cost-effective for women in the Southern United States? Sex Transm Dis 2006; 33(supp l7):S46–S49. doi: 10.1097/1001.olq.0000221015. 0000264056.ee.
50.Latane B. The psychology of social impact. Am Psychol 1981; 36:343–356.
51.Adimora AA, Schoenbach VJ, Martinson F, et al. Concurrent sexual partnerships among African Americans in the rural south. Ann Epidemiol 2004; 14:155–160.