DISEASE TRANSMISSION MODELING has demonstrated the importance of sexual mixing patterns in a population for the spread of sexually transmitted diseases (STDs). Sexual networks form the pathways across which these infections are transmitted from one person to another. Unlike casual contact, sexual partnering is nonrandom and has major consequences for the rate and extent to which an infection can spread. 1–4 Strong assortive mating can lead to multiple, decoupled epidemics in a population, 5,6 whereas random mating tends to generate slower but larger epidemics in the long run. 7,8
Hallinan and Williams 9 discussed five factors identified in the social psychology literature that act as bases of interpersonal attraction: propinquity, similarity, complementarity, status, and reciprocity. Propinquity and similarity are believed to have the greatest impact on friendship formation. Individuals are more likely to establish ties with those with whom they have more opportunities to interact. Furthermore, given propinquity, individuals are more likely to form relationships with those similar to themselves in attitudes, values, and behavior. Hence, because ethnic groups are segregated residentially, and because shared cultural values exist within ethnic groups, individuals may be more likely to have relationships with members of their own race or ethnic groups. Relationships with individuals who have like characteristics can be designated as assortative (like with like), and relationships with those who have different ethnicity as disassortative.
Research on sexual networks has identified aspects of sexual behavior that affect the transmission of infection: multiple partnerships, patterns of partner selection, bridging partners, concurrent partners, use of condoms with different types of partners, and other measures of sexual behavior.
The transmission of infections between two different groups depends on the presence of bridge partners, or persons who have sexual intercourse with individuals in both groups. The practice of younger women having sex with older men assisted in the transmission of HIV infection to younger cohorts in Africa, 10 and the practice of older men having sex with younger men facilitated the spread of HIV infection among gay men in the United States. 11,12 Furthermore, in a study on the sexual networks of American adults, it was noted that the higher infection rates for bacterial STDs among blacks can be explained by patterns of sexual networks within and between different racial/ethnic groups. 13
Sex partnerships that overlap in time, or concurrent partnerships, also have been identified as explanations for differences in disease prevalence among different populations. 14–17 Concurrent partnerships increase the opportunity for an infection to spread quickly in a network of sexual partners by allowing infections during a given period to be transmitted to and from each partner who participates in the network. Concurrency in sexual networks has been identified as a factor associated with the spread of HIV infection in Uganda, 15 and with the risk of chlamydia infection in a community study in Colorado. 17
Differences the characteristics of partners, particularly differences in partner age, may have an effect on sexual behavior and the use of condoms. Only a small number of analyses have considered the effects of partner characteristics on condom use and the use of contraceptives. Differences in age or other characteristics of partners may reflect differences in the power to make decisions in a relationship, and these differences in power may affect the degree to which condoms or other contraceptive methods are used. For example, in a study on the degree of control that young women have over their first intercourse, Abma et al 18,19 found that women whose first sexual partners were significantly older than themselves were more likely to report that the intercourse was involuntary. These women also were less likely to report that contraceptives were used.
American adolescents experience high rates of STD infection. 20 As compared with older adults, adolescents (10- to 19-year-olds) and young adults (20- to 24-year-olds) are at higher risk for acquiring STDs because they may be more likely to have multiple partners, to engage in unprotected intercourse, and to select higher risk partners. In addition, adolescent women may have an increased risk of infection with Chlamydia trachomatis because of cervical ectopy. 21–24
Although a large literature exists on adolescent sexual behavior in the United States, relatively little information is available on the sexual mixing patterns of adolescents. The objectives of this report are to describe the mixing patterns of American adolescents by age and ethnicity, to describe the patterns of bridging partners and concurrency by age and ethnicity, and to examine the association of bridging and concurrency in this population with condom use.
Data for this study was drawn from the National Longitudinal Study of Adolescent Health (AddHealth). The AddHealth data were designed to assess the health status of adolescents, and to explore the causes of their health-related behaviors, focusing on the effects of multiple contexts or environments (both social and physical) in which they live. 25 A sample of schools was selected by Quality Education Data, Inc. The sample of schools was stratified by region, urbanicity (urban, suburban, or rural), school type (public, private, or parochial), ethnic mix, and size. Consequently, 80 schools were selected by probability proportional to size. More than 70% of the originally sampled high schools participated in the study. If a school refused to participate, a replacement school was selected from the same stratum.
Participating schools provided a roster of students for use in the project. In most cases, schools also agreed to administer the in-school questionnaire to each student during one class period. The in-school sample was completed by 90,118 students in 1994–1995.
Once a high school was selected, a feeder school (one that included seventh grade students and sent its graduates to that high school) was recruited. A feeder school then was selected in each of 80 communities.
Next, a sample of students on school rosters was selected to participate in an in-home interview. Students were grouped into 12 grade and gender strata within a school, and an average of 17 students were selected randomly from each within-school strata. This method yielded approximately 200 students for in-home interviews from each of 132 schools. In 16 small schools, all the students were selected for in-home interviews.
Respondents selected through this two-stage sampling procedure constituted a nationally representative sample of adolescents. Additional students were chosen to represent a number of ethnic groups and disabled students, and to create genetic samples of related individuals. Only students selected from the nationally representative sample were included in this study. This nationally representative sample provided data on 18,924 respondents interviewed in their homes in wave 1.
In a second wave conducted 2 years later, 13,570 wave 1 respondents were interviewed in their homes. Data from both the Wave 1 and the Wave 2 interviews are included in this report.
This analysis focuses on the sexually active portion of the sample. During the in-home interview, information on sexual activity was obtained through two different sequences of questions. Respondents were asked a direct question on whether they ever had sexual intercourse. In wave 1, of the 18,924 respondents, 7508 reported ever having sexual intercourse. Of the 13,570 respondents in wave 2, 5989 reported that they had sexual intercourse. Merging the two waves of data, 9303 respondents distinctly reported ever having sexual intercourse.
The unit of analysis in this report consists of both individual records and sexual relationships. For the analysis related to sexual partners, the respondent level data file had to be converted from one with a record for each individual to a file with a record for each sexual partner. The 9,303 individuals who answered the direct question affirmatively reported having 26,657 sexual partners.
In each of the two waves, the respondents were asked to report on relationships in the preceding 18 months. Some telescoping might have occurred, such that respondents reported relationships in wave 2 that began more than 18 months before. Relationships reported by the same respondent in waves 1 and 2 were compared on five variables including demographic and relationship variables. This procedure identified 724 relationships (2.7% of the sample relationships) that likely were duplicates. These relationship records were deleted from the relationship data set. Finally, relationships for which the respondent did not report sexual intercourse with that partner were deleted from the file.
Examination of the data on partners also showed that some individuals who answered the direct question by saying they never had sexual intercourse later reported intercourse when asked about specific partners. From these cases, 363 additional relationships were added to the relationship file. Only 20 of the reported relationships were homosexual. These were deleted from the file. The total number of heterosexual relationships used in the analysis was 17,266, which were reported by 8,024 individuals.
In the 16 schools wherein all the respondents were eligible for the in-home interview, a large number of relationships (but not all) were reported by both partners, causing those relationships to be overrepresented in the analysis. Tabulations provided by the AddHealth staff at the University of North Carolina on the frequency of relationships reported by both partners allowed adjustments to the sampling weights that compensated for the overrepresentation of these relationships.
Cross-tabulations were developed to examine respondent and partner agreement on variables such as age and ethnicity. To test the significance of differences between distributions, χ2 statistics were used. The SUDAAN statistical package was used to estimate standard errors of percentages, taking the clustered design into account. 26
To examine the association between individual characteristics and condom use, a logistic regression analysis was used. The SUDAAN statistical package was used to generate logistic models using sampling clusters to group observations. 23 The results are reported as important if the P values of a test statistic were smaller than 0.05.
ADDHealth data includes weights designed to compensate for unequal probabilities of selection and nonresponse. Each relationship was assigned the corresponding wave 1 or 2 weight for the respondent. Weights were used in all analyses to provide national estimates of relationship characteristics for adolescents.
The data on age were coded in single years for both respondent and sexual partner. Differences between ages of the partners were coded as more than 2 years older, within 2 years, and more than 2 years younger.
The data on respondent and partner ethnicity were coded as white, black, Latino, or other. These data were collected by the respondent's report of his or her most important ethnicity. Patterns of mixed ethnicity were not analyzed because only 2.7% of the respondents reported a second ethnicity.
Condom use refers to whether the respondent ever reported using condoms with sexual partners.
Gender refers to the reported gender of the respondent.
Concurrency was determined by examining the dates for the first and last times the respondents had sex in the relationship. Persons with concurrent relationships had overlapping dates.
A respondent was classified as a bridge person for age if he or she had partners in more than one age group. Age groups were classified as more than 2 years younger than the respondent, within 2 years of the respondent's age, or more than 2 years older than the respondent.
Bridge race was defined as a respondent with a partner in more than one ethnic group.
Table 1 shows the demographic characteristics of the respondents in the study who reported sexual relationships. Of the approximately 8000 persons, 11% were age 14 years or younger, 36% were ages 15 to 16 years, and 54% were age 17 years or older. More than half of the sample (62%) was white, 19.2% were black, 11.7% were Latino, and 7% were of other races.
Table 2 shows descriptive data on relationship partners by age, gender, and ethnicity. In the total sample, 78% of the partnerships were with someone of the same ethnicity, and 22% were with someone of a different ethnicity. Age differences between partners were common in the sample. Approximately 55% of the relationships included a partner within 2 years of age. However, older partners were reported in nearly one third of the relationships, and younger partners in the remainder (12.6%).
The age of the respondent was associated with several of the partner characteristic variables. The older respondents were somewhat more likely to have a partner of the same ethnicity (80%) than the youngest respondents (74%). The age of the respondent also was related to age differences between partners. The youngest respondents (age 14 years or younger) were more likely to have partners two or more years older (41%), and few partners two or more years younger (3%). The oldest age group (age 17 years or more) had a substantially larger proportion of partners who were two or more years older (28%) and two or more years younger (17%).
Gender was not associated significantly with ethnicity of partner. However, the girls were more likely than the boys to have partners two or more years older (49% of the girls versus 14% of the boys), whereas the boys were more likely to have partners two or more years younger (22% of the boys versus 4% of the girls).
Ethnicity was related to several characteristics of partner choice. Both white and black respondents had a low percentage of partners of other races (13% of the whites versus 15% of the blacks), whereas Latinos and others had a higher percentage (42% and 77%). The age of the partner did not differ among ethnic groups.
Table 3 shows the proportion of respondents who reported more than one partner. Approximately 44% of the total sample reported one partner, and 56% reported more than one partner.
Small ethnic differences were observed in the proportion of respondents reporting multiple partners. A higher proportion of black (61%) and other respondents (59%) reported multiple partners than Latino (55.8%) and white respondents (54%) (P < 0.05).
Gender differences in the reporting of multiple partners differed by ethnic group. Among white respondents, the girls reported a slightly higher proportion of multiple partners (52.4% of the boys versus 55.9% of the girls, P < 0.05). However, among black and Latino respondents, more of the boys reported more than one partner (blacks: 63.1% of the boys versus 58.5% of the girls, P < 0.05; Latino: 59.3% of the boys versus 51% of the girls, P < 0.05).
The percentage of adolescents with bridge and concurrent relationships is shown in Table 4 for adolescents with two or more partners. Slightly more than half (54%) of this subsample of respondents reported concurrent partners. Among ethnic groups, the proportion of persons reporting concurrent partners did not differ significantly.
Gender differences were present in the reporting of concurrent partners. Among all the respondents, the girls reported concurrent partnerships more often than boys (58% of the girls versus 50% of the boys, P < 0.05). This difference was found in all the ethnic groups.
Two types of bridge partners were considered in this analysis: bridge by age and bridge by race. Bridge by age, or the reporting of partners from two different age groups, was very common in this population. Approximately 69% of the respondents reported partners in two different age groups. The percentages were slightly lower for black and Latino respondents (66% of the blacks versus 65% of the Latinos).
The girls were much more likely to report partners in two different age groups than the boys (61% of the boys versus 77% of the girls, P < 0.01). The girls in each ethnic group also reported partners in two different age groups more often than the boys.
Bridge by ethnicity, or respondents with partners in two different ethnic groups, was less common in this sample than respondents with partners in different age groups. Approximately 35% of the respondents reported that they had sexual partners from two different ethnic groups. The four ethnic groups differed on this variable. Members of Latino and other groups were much more likely to report partners in two ethnic groups (Latino 60%, other 90%).
Gender influenced the patterns of bridging by ethnicity. Among white respondents, the girls were more likely to report partners of different ethnicity than the boys (28% of the girls versus 24% of the boys, P < 0.05). For black and Latino respondents, the situation was reversed, with the boys more likely to report partners of different ethnicity than the girls (black: 35% of the boys versus 19% of the girls; Latino: 62% of the boys versus 57% of the girls, P < 0.01).
The percentage of respondents who reported using a condom with at least one partner is shown in Table 5. Approximately one third of the respondents had used a condom with at least one partner. Those with bridge or concurrent partners were less likely to have used a condom with their partners.
Table 6 shows a logistic regression with condom use as the dependent variable. In the first model, not including the number of partners, concurrency was significant and related to increased condom use (odds ratio [OR], 1.20; CI, 1.01–1.44;P < 0.05). Ethnicity was not related to condom use. Girls were less likely to report condom use than boys (OR, 0.79; CI, 0.65–0.94;P = 0.01) Respondents who reported partners in two different age groups also reported reduced condom use (OR, 0.80; CI, 0.65–0.94;P < 0.05). Among those who reported partners in more than two ethnic groups no significant differences in condom use were observed. The number of partners reported was associated with reduced condom use (OR, 0.45; CI, 0.41– 0.49). Two-way interactions between independent variables were tested, but not found to be significant.
Data from a national survey of American youth, the ADDHealth study, provided an opportunity to examine the characteristics of sexual networks among American youth. The analysis showed some significant features of these sexual networks.
Existing theory predicted that similarity and propinquity should influence the demographic characteristics of partners. The data indicate that whereas some similarity exists among partner characteristics, partners with different characteristics also are quite common. Similarity can be found in the ethnicity of the partners associated with both white and black adolescents. Most partners come from the same ethnic group. However, Latinos and respondents of other ethnicity have partners of different ethnicity more frequently. These results on ethnicity are consistent with those from studies of U.S. adults 27 and urban youth. 28 In contrast to the results for ethnicity, partners of different ages often were observed in all ethnic groups. This was observed most often for female respondents.
A large proportion of the sample respondents reported more than two sexual partnerships, concurrent partnerships, and partners in two different age groups. The reporting of partners in two different ethnic groups varied strongly by ethnicity. The reporting of multiple partners and partners in two or more age or ethnic groups was very common in this population. Bridging by age was very common among girls, and bridging by ethnicity was very common among persons in Latino and other ethnic groups. Concurrent partnerships were common in all age and ethnic groups.
Condom use was reduced for persons with multiple partners and for those with partners in different age groups. When control was used for these factors, condom use was found to have a positive relationship to concurrency, although the bivariate relationship was negative.
In a network study of chlamydia infection, Potterat et al 17 found that both concurrency and a disassortative mixing pattern by ethnicity were associated with chlamydia transmission. Although the dependent variable in the current study was condom use rather than chlamydia infection, other variables including number of partners and bridging by age were more important predictors of reduced condom use. Furthermore, in the multivariate analysis of the current study, concurrency had a positive association with condom use. Differences in results may have resulted from differences between adolescent and adult populations, and in the study selection criteria.
In a study of sexual networks among American adults, Laumann and Youm 13 found that black adults usually selected blacks as sexual partners, and that there was much more mixing between core and periphery groups in the black population. The current study found a high degree of within-group choice among American adolescents, although lack of information about the sexual behavior of adolescent partners prevents a similar analysis of these data.
These findings from this study are important for STD clinics, family planing clinics, and other health providers who work with sexually active adolescents. Multiple relationships and relationships with persons of different age and ethnic groups increase the vulnerability of adolescents to sexually transmitted infections. Adding to this risk is the lower use of condoms in these relationships. The counseling of sexually active adolescents should include discussion of differences in power or communication that may occur where a partner's personal characteristics may differ.
1. Anderson R, May R, Boily M, Garnett G, Rowley J. The spread of HIV-1 in Africa: sexual contact patterns and the predicted demographic impact of AIDS. Nature 1991; 352: 581–589.
2. Gorbach PM, Sopheab H, Phalla T, et al. Sexual bridging by Cambodian men: potential importance for general population spread of HIV/STD epidemics. Sex Transm Dis 2000; 1: 1–8.
3. Aral SO. Behavioral aspects of sexually transmitted diseases, core groups, and bridge populations. Sex Transm Dis 2000; 27: 327–328.
4. 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.
5. Koopman J, Simon C, Jacquez JA, et al. Sex partner selectiveness effects on homosexual HIV transmission rates. J Acquir Immune Defic Syndr Hum Retrovirol 1998; 1: 486–504.
6. Hyman J, Stanley E. Using mathematical models to understand the AIDS epidemic. Math Biosci 1988; 90: 415–473.
7. Gupta S, Anderson R. Networks of sexual contacts: implications of the patterns for the spread of HIV. AIDS 1989; 3: 807–817.
8. Anderson R. Some aspects of sexual behavior and the potential demographic impact of AIDS in developing countries. Soc Sci Med 1992; 34: 271–280.
9. Hallinan M, Williams RA. Interracial friendship choices in secondary school. Am Soc Rev 1989; 54: 67–78.
10. Knolle H. Age preference in sexual choice and the basic reproduction number of HIV/AIDS. Biometrics 1990; 32: 243–256.
11. Morris M, Zavisca J, Dean L. Social and sexual networks: their role in the spread of HIV among young men. AIDS Educ Prev 1995; 7: S24–S35.
12. Service SK, Blower SM. HIV transmission in sexual networks: an empirical analysis. Proc R Soc Lond B Biol Sci 1995; 260: 237–244.
13. Laumann EO, Youm Y. Racial/ethnic differences in the prevalence of sexually transmitted diseases in the United States: a network explanation. Sex Transm Dis 1999; 26: 250–261.
14. Hudson CP. Concurrent partnerships could cause AIDS epidemics. Int J STD AIDS 1993; 4: 349–353.
15. Morris M, Kretzschmar M. A microsimulation study of the effect of concurrent partnerships on the spread of HIV in Uganda. 2000 (unpublished manuscript).
16. Watts CH, May RM. The influence of concurrent partnerships on the dynamics of HIV/AIDS. Math Biosci 1992; 108: 89–104.
17. Potterat JJ, Zimmerman-Rogers H, Muth SQ, et al. Chlamydia transmission: concurrency, reproduction number, and the epidemic trajectory. Am J Epidemiol 1999; 150: 1331–1339.
18. Abma JC, Chandra A, Mosher WD, Peterson L, Piccinino L. Fertility, family planning, and women's health: new data from the 1995 National Survey of Family Growth. Vital Health Stat 1997; 23: 19.
19. Abma J, Driscoll A, Moore K. Young women's degree of control over first intercourse: an exploratory analysis. Fam Plann Perspect 1998; 30: 12–18.
20. Division of STD Prevention. Sexually Transmitted Disease Surveillance, 1999. Atlanta, GA: Centers for Disease Control and Prevention, September 2000.
21. Centers for Disease Control. Premarital sexual experience among adolescent women: United States 1970–1988. MMWR Morb Mortal Wkly Rep 1991; 39: 929–932.
22. Centers for Disease Control. Pregnancy, sexually transmitted diseases, and related risk behaviors among U.S. adolescents. Atlanta: Centers for Disease Control and Prevention, 1004. Adolescent Health: State of the Nation Monograph Series. No. 2. CDC Publication No. 099–4630, 1994.
23. Santelli J, Brener ND, Lowry R, Bhatt A, Zabin LS. Multiple sexual partners among U.S. adolescents and young adults. Fam Plann Perspect 1998; 30: 271–275.
24. Sonenstein FL, Ku L, Lindberg LD, Turner C, Pleck JH. Changes in sexual behavior and condom use among teenaged males: 1988 to 1995. Am J Public Health 1998; 88: 956–959.
25. Bearman J, Jones J, Udry R. The National Longitudinal Study of Adolescent Health. Chapel Hill, NC: Carolina Population Center, 1997.
26. Shah BV, Barnwell BG, Bieler GS. SUDAAN software for statistical analysis of correlated data: user's manual. Research Triangle Park, NC: Research Triangle Institute, 1999.
27. Laumann EO, Gagnon JH, Michael RT, Michaels S. The social organization of sexuality. Chicago: University of Chicago Press, 1994.
28. Ford K, Norris AE. Sexual networks of African American and Hispanic youth. Sex Transm Dis 1997; 24: 327–333.