Objectives: To examine the epidemiology of multiple infections in teenagers and their contacts.
Goal: To demonstrate a network approach to the prevalence of STDs/HIV.
Study Design: A community-based network study of ethnographically representative adolescents.
Results: Of the 512 interviewees, 90% of the boys and 80% of the girls were sexually active with many partners and exhibited age-assortative mixing. Prevalences of gonorrhea, chlamydia, and HSV-2 were high, but syphilis, Hepatitis C and HIV were virtually absent. The prevalence faced by subgroups differed from overall prevalence.
Conclusions: The presence of bacterial STDs and the absence of HIV, HCV, and syphilis may be the result of no IDU and assortative age mixing. Testing for multiple conditions, network assessment of prevalence, and inclusion of social contacts enhance understanding of STD epidemiology.
Average prevalence for a group may be misleading, since persons face disease prevalence within their own network. Such prevalence may differ from the average based on the disease and the characteristics of individuals.
From the *Department of Medicine, Emory University School of Medicine, Atlanta, Georgia; †Quintus-ential Solutions, Colorado Springs, Colorado; and ‡Division of Health Behavior, College of Public Health, University of Kentucky, Louisville, Kentucky
This work was supported by grant U01AI45882 from the National Institute of Allergy and Infectious Disease, National Institutes of Health.
Correspondence: Richard B. Rothenberg, MD, MPH, Department of Medicine, Emory University School of Medicine, Atlanta, Georgia. E-mail: email@example.com.
Received for publication October 17, 2005, and accepted November 29, 2006.
Currently available data suggest a high prevalence of some sexually transmitted diseases among teenagers—for example, a prevalence of 2.4% for chlamydial infection among 15- to 19-year-old women in 2000–20041—and a high level of sexual activity among young people.2 By extension, it is often assumed that these young people are currently at risk for infection with HIV, or may be in the future because of their established pattern of sexual activity.3–5 Virtually all of the cumulative total 37,599 persons diagnosed with HIV infection between the ages 15 and 246 are believed to have been infected as teenagers.
National data, however, are limited in the detail they can provide, and are often unable to offer insights into local circumstances or the specific patterns of risk and disease among subgroups. Young people who may be at high risk because of their sexual activity or drug use are a subgroup that warrants particular attention, because the pattern of acquisition of STDs and HIV may shed light on both short- and long-term consequences of the risks they take. In particular, a disjunction between the acquisition of STDs and HIV may be important in understanding the epidemiology of both and for presaging future health threats to which these young people are subject. To address this issue, we used street-based recruitment to study a group of adolescents, aged 15 to 18, from a working-class community in the metropolitan Atlanta area. Our purposes were to describe the relationship between risk configuration and extant infections, describe the risk for multiple endemic conditions, and provide network perspective on the prevalence of disease faced by given persons and groups.
This study was conducted in a neighborhood in Southwest Atlanta between 1999 and 2003. Using a targeted sampling approach,7,8 we enrolled persons aged 15 to 18 years by approaching them directly in the community. The community we chose was a working-class neighborhood in Southwest Atlanta (locally known as the Swats) that consists primarily of multifamily dwellings, commercial districts, and one large mall. There is a large population of young people and an active street life. After a start-up period of several months, during which field staff became familiar with the target neighborhood and became known to potential participants, we approached young people whom we observed regularly at about 6 venues within the neighborhood (including the mall). We thus attempted to sample representatively from persons who frequented the neighborhood and appeared to be within the designated age range. As it turned out, the neighborhood is cohesive and relatively bounded, so that virtually all of the persons who entered the study, both enrollees and contacts, were from that neighborhood. After an initial startup period, we recognized that our ability to find contacts to young people, primarily because of the evanescent nature of many of their relationships, was limited, and we sought to interview at least one named contact from each interviewed person.
For those under 18, we asked them initially of their interest in participating. For those who expressed interest, we requested an introduction to their parents (or guardians), informed their parents of the nature of the study, and sought the informed consent of the parents and the assent of the subject. It was explained to parents and subjects that the study would ask the latter about their current health, demographic characteristics and living situation, their risk behaviors (both sexual and drug using), and would ask them to identify friends and contacts. They would also be tested for 8 conditions, including HIV, that might be transmitted sexually or parentally, and we would recontact the subject up to 3 times, at 6-month intervals, for repeat evaluations. Parents (guardians) were informed that the results of the interviews and tests would not be shared with them by study personnel. If both consent from the adult and assent from the adolescent were obtained, a face-to-face paper-and-pencil interview and STD/HIV testing were administered in a private site away from the home, with no other persons present. We tested persons by obtaining 10 mL of venous blood and a small specimen of initially voided urine. The specimens were taken daily to the central Emory University laboratory where standard testing procedures were applied. Test results were conveyed in person to the participant, and we provided them with assistance in reaching medical care if needed. We asked participants to identify their sexual, social, and drug-using partners during the preceding 6 months and requested permission to recruit and interview the partners. If permission was granted, we asked the participant to introduce us to the partner; if the participant was hesitant to introduce us, we sought the contact and did not identify the participant. If permission was not granted, we did not attempt to enroll the partner. Those 18 years old and over provided their own consent but were otherwise treated identically.
We accumulated data in separate modules (a demographic profile; medical history; psychosocial factors; behavioral characteristics; laboratory results) that were assembled into a single data set for analysis purposes. Data were password protected and access was limited to selected study personnel. We made considerable effort during and at the conclusion of the study to deduplicate persons who had been named in the study, first using interviewer recall, then employing an algorithm for identifying records of individuals with similar characteristics, manually judging the probability that these records described a single person. The study was approved and monitored yearly by the Emory University Institutional Review Board.
We used information from the initial interviews with enrolled respondents (N = 242) and the contacts they named (N = 270) (together designated here as “interviewees” [N = 512]) to describe the demographic characteristics, risk-taking behavior, and test results of subjects in the study, maintaining the distinction between respondents and contacts. Taken together, these 512 interviewees formed 991 pairs for which we have interview information from both members. Because this group was interconnected, each interviewee could appear in more than one individual’s personal network (displayed in Fig. 2, with accompanying explanation). Thus, without regard to the distinction between respondents and contacts, each person has an egocentric (personal) network, from which the prevalence that he or she faces is calculated. In addition, that person may appear in the egocentric network of multiple others. It is evident that the prevalence faced by an individual, or by persons who share given characteristics, will be different from the prevalence of the overall deduplicated group. This network approach elucidates the heterogeneity of prevalences faced by different persons or groups.
We used the interviewed population of 512 persons to assess associations (using univariate comparisons, stratified analyses, and logistic regression), of important exposure variables and the outcomes of interest (gonorrhea and chlamydial infection). For the univariate comparison of respondents and contacts, statistical testing was not performed, primarily because the sampling procedure sought representativeness but was not necessarily random. We observed, however, less than a 10% point difference between respondents and contacts on almost all comparisons; such differences would not be statistically significant in general at the observed factor probabilities and sample size. We used the 991 dyads to calculate the frequency with which diagnoses were made in the persons to whom an interviewee was connected and examined those frequencies for both infected and uninfected persons.
Demographics and Social Setting
Of the 242 adolescents recruited into the study, 119 (49%) were boys and 123 (51%) were girls (Table 1). The recruited respondents were 15 to 18 years old (79% were 15–17). The 270 contacts, in contrast, were older, a result permitted by the study design (Fig. 1) Almost 98% of the total population were black. About 74% (179/242) of respondents were presently enrolled in school. That proportion was highest for 15-year-olds (91%) and lowest for 18-year-olds (44%) (data not shown). Among the respondents, 66% of the girls and 46% of the boys had actually been in school during the current year; of those not in school, 21% and 19%, respectively had been expelled. Fewer contacts were currently enrolled (119/270, 44%), but the proportions currently attending or ever expelled were similar. Almost three-quarters of respondents lived with at least 1 parent, though only one-quarter had been at their current address for at least 5 years (average time at a current address was about 2 years). Contacts were similar but reported a longer tenure at their current address.
Over 80% of respondents reported receiving income from their family, usually some form of allowance. Twenty-four percent of the boys and 8% of the girls said that they derived income from selling drugs or from stealing. About 44% of boys and 24% of girls had spent time in jail (27% of boys and 11% of girls had stayed at least 1 night in jail during the preceding 6 months). Seventeen percent of boys and 11% of girls were currently on probation. Nearly three-quarters of all respondents had a family member who had spent time in jail (Table 1). Contacts derived less of their income from their family, and a greater proportion had been incarcerated, but they were otherwise similar to respondents (virtually all comparative measures were within 10% points of each other, as noted).
Prior History of STDs
Ninety percent of respondents reported having excellent or good health. Among respondents, 12% of females and 5% of males had been previously diagnosed with gonorrhea or chlamydial infection, though there is some uncertainty that subjects were able to report these diagnostic distinctions accurately. Less than 1% of the cohort reported a prior diagnosis of herpes, hepatitis, or genital warts. No one in the study population had been diagnosed with HIV or syphilis. Contacts reported a similar experience (data not shown in tabular form).
The majority of respondents were sexually active: nearly 90% of males and 80% of females had engaged in vaginal intercourse (Table 2). The average age at first intercourse for boys was 13 years, and the average age at first intercourse for girls was 14 years. Sexually active males had an average of 2.4 partners in the past month, while sexually active females had an average of 1.1 partners. Males reported 13.9, and females 4.1, lifetime partners on average. The values for contacts were similar.
Both boys and girls had sexual contact largely within their age group. Ninety percent of sexual contacts named by participants were 21 years of age or younger. Boys named a greater proportion of contacts younger than themselves, and the majority of their contacts were within 3 years of their age (Fig. 1). The pattern for girls differed somewhat: Although the majority of their contacts were within 3 years of their own age, those contacts were generally somewhat older. Only 3% of contacts to either boys or girls were over the age of 25 years.
Among respondents, the percentages of males and females ever using a condom were the same as the percentages ever engaging in vaginal intercourse. More males than females reported ever using condoms: 88% of males and 79% of females. One-third of females reported ever using oral contraceptives, but only 6% had used oral contraceptives at most recent sexual intercourse. About 40% of boys and girls had ever used penile withdrawal as a form of sexual protection, and 42% had ever engaged in sexual intercourse without using contraception. Among female respondents, 36% reported having ever been pregnant, while 20% of males reported ever getting someone pregnant (Table 2). The equivalent percentages for contacts were higher: 52% for girls and 42% for boys.
Among male respondents, 86% agreed that condoms protect against pregnancy, but only 68% of female respondents believed this to be true. Eighty-two percent of boys and 63% of girls stated that condoms protect against STDs. Almost 40% of both boys and girls agreed that penile withdrawal is protective against pregnancy, and 9% felt that this method is protective against STDs. Approximately 90% stated that they considered themselves to be at low risk for chlamydial infection, herpes, hepatitis, HIV, syphilis, and gonorrhea. The results were similar for contacts (data not shown).
In the study group, 58% of respondents stated that they had ever smoked a cigarette. Almost 75% had ever used alcohol, and 65% had ever used marijuana. Approximately 50% of the respondents had used alcohol in the past 6 months, and 50% had used marijuana during the same time period. Less than 2% of respondents reported use of powdered cocaine, PCP, hallucinogens, or inhalants. None of the respondents reported smoking crack cocaine, or injecting cocaine, heroin, amphetamines, or any other drug. Contacts reported a higher use of marijuana and powdered cocaine, but none had smoked cocaine or injected drugs.
Among the 242 respondents, 27% of girls and 13% of boys were positive for chlamydia infection. Gonorrhea was also higher among girls: 11% compared to 2% of boys. Herpesvirus type 2 (HSV-2) was highly prevalent in the study population. Forty-two percent of boys and 52% of girls were positive for HSV-2 antibody. For Hepatitis B, 16% of boys and 21% of girls tested positive, but in many instances respondents were uncertain of their vaccination status. Approximately 7% of girls had trichomoniasis (males were not tested). Only 1 boy (and no girls) was positive for Hepatitis C (1 additional infection with Hepatitis C was detected in a contact late in the study). There were no cases of HIV or syphilis identified in participants, but 1 social contact, a 19-year old woman, had previously undiagnosed infection with HIV. With minor variations, the overall prevalence of these conditions was similar for the contacts, with no cases of syphilis identified.
Prevalence of Infection in a Person’s Environment
As noted, the 991 dyads formed by 512 interviewees permits assessment of the prevalence of infection in the environment (i.e., personal network) of any individual, or of groups of individuals that share characteristics. The prevalence of infection faced by persons in these dyads varied by the type of infection and whether the person was positive or negative for a given pathogen. For all contacts (social, sexual, or drug using), scanning the columns of Table 3 reveals a general uniformity in the presence of infection, regardless of the infection status of the respondents. The presence of infection in contacts to interviewees who were either positive or negative for a given condition reflected the overall prevalence of infection in this group (Table 3, bottom row). For example, gonorrhea was present in 4.4% of the 574 persons who were connected to interviewees without HSV-2, and in 6.5% of the 417 persons connected to interviewees with HSV-2; the parallel figures for chlamydial infection were 18.6% and 19.9%. In general, though not in every case, infection was more prevalent in persons connected to positive interviewees than in persons connected to negative respondents. The outstanding example was the well-known association of gonorrhea and chlamydial infection. The prevalence of chlamydia infection was 37% in personal networks of those with gonorrhea, and 18% of those without gonorrhea. The prevalence of gonorrhea was 9% in the personal networks of those with chlamydia, and 4.5% in those without chlamydia (Table 3). This general pattern was also observed for males and females examined separately (data not shown). Finally, the same pattern was observed when only sex contacts were considered, but with more pronounced separation between contacts to positive and negative persons. For example, 20% of those connected to chlamydia-positive persons had gonorrhea and 33% had chlamydia; 9% of those connected to chlamydia-negative persons had gonorrhea and 16.5% had chlamydia (Table 4). As was the case with all contacts (Table 3), persons connected sexually to those without disease had disease frequencies that reflected those of the overall population. In fact, persons connected to boys and girls who were negative for HSV-2, gonorrhea, and chlamydia had disease frequencies that reflected the general population prevalences (data not shown).
Risk Factors Associated With Gonorrhea and Chlamydial Infection
Although, as noted, this population exhibited a high level of sexual risk, only a few such risks were significantly associated with bacterial infections. The strongest association with chlamydial infection was concurrent gonorrhea (OR = 10.9; 95% CI, 4.4–27.3). Chlamydial infection was also associated with reporting condom use at last intercourse (OR = 2.7, 95% CI, 1.5–5.0); having more than 1 sex partner during the last 6 months (OR = 2.0; 95% CI. 1.1–3.6); and reporting an inability to seek medical treatment (OR = 2.6; 95% CI 1.5–4.5). The strongest association with gonorrhea was concurrent chlamydial infection (OR = 11.7; 95% CI 4.6–30.1). The presence of gonorrhea was also associated with spending at least one night in jail during the last 6 months (OR = 3.1; 95% CI 1.04–9.2). Other point estimates, though largely meeting expectations, were nonsignificant.
In this group of adolescents who reside within a defined neighborhood in Atlanta, we found substantial occurrence of the bacterial STDs (gonorrhea and chlamydia infection), very high prevalence of several chronic viral STDs (hepatitis B, HSV-2), but virtually no infection with Hepatitis C, HIV, or syphilis. Several factors may be protective. First, the high level of sexual activity (early age at sexual debut, frequent sexual encounters, numerous partners) was not accompanied by injecting drug use or crack smoking (although such drugs were sold by our participants) (Table 2). Second, there was little disassortative mixing by age, and older persons were not commonly named as sexual contacts (Fig. 1). Blanket risk for HIV assigned to sexually active adolescents should be tempered by knowledge of the specific risks to which they are subject. Current behavior does not preclude future risk, but the long-term trajectory of these adolescents is not known, and whether they will adopt behaviors that put them at greater risk is uncertain. Nonetheless, evaluation of HIV prevention programs should include an assessment of a priori risk that goes beyond the catalog of sexual acts. (In this group, for example, an HIV prevention program would be wildly successful.)
On the other hand, facile generalization is unwarranted. It is unclear to what extent this group is representative of youth at risk. Though investigators may be tempted to dichotomize groups into “core” and “noncore” based on their propensity to sustain transmission, there is more likely a continuum of “coreness.” If so, the group described here occupies the end of the spectrum with intense transmission and is probably important in the overall maintenance of STD endemicity. Again, evaluation of STD (and HIV) interventions requires an understanding of the frequency distribution along the continuum of “coreness.”
The significant risk factors for chlamydial and gonococcal infection described in this study confirm several well-established associations, as well as the divergence of risk profiles for the 2 conditions.9 The joint presence of these 2 infections has long been recognized,10–14 though the association noted here is stronger than that usually reported. There was a strong association between gonorrhea and time spent in jail, previously reported by Stoner et al.,9 and an increased risk for gonorrhea associated with disassortative age mixing (though not statistically significant in these data), an observation that confirms inferences drawn from national age-specific incidences for gonorrhea.15 Though several of these observations may be of interest, the contribution of “risk-factor epidemiology” to an understanding of this group is uncertain, reflective of controversies in the approach.16–18
The customary comparison of respondents and contacts revealed that they were similar on many counts, suggesting that we had entered a peer group with high sexual activity and bounded social interactions. The major difference—in age distribution—was an important observation facilitated by the study design. We recruited persons 15 to 18 years of age but placed no restriction on the age of contacts. Had we begun with the slightly older group (and assuming equal success in interviewing), the roles of respondent and contact would have been reversed. These observations suggested the potential utility of a network-informed approach.
The analyses presented in Tables 3 and 4 are derived from a network perspective. We adopted the premise that the distinction between respondents and contacts is an artifact of case finding. That we happen to find one or the other first, and refer to that person as a “respondent,” and the persons that he or she names as “contacts,” is a simple function of the moment we intervene. It is more important that this respondent–contact pair is embedded in a larger network of persons who share similar characteristics (Fig. 2). The nonsexual, nondrug-using partners of both infected and uninfected persons had a high prevalence of disease. This observation suggests that traditional case finding methods, usually limited to the sexual partners of uninfected persons, will miss considerable disease transmission in a group such as this one.
The observation of multiple risks, multiple infections, and a high level of interactivity might then engender a reconsideration of some standard STD control practices. A high prevalence of disease in a highly active population suggests that contact tracing need not focus exclusively on the sexual partners of infected persons. From these data, it would appear that interviewing anyone (infected or not with any disease) for any partners (sexual or otherwise) would discover substantial infection in contacts. Certainly, interviewing those with gonorrhea or chlamydia infection for their sexual partners would produce the highest “yield,” but one would also expect that the nonsexual partners of uninfected persons would have, in this population, a chlamydial prevalence of nearly 20%. The approach used here—enrolling respondents, eliciting contacts, and seeking out the contacts for interview and testing—is derived from traditional contact tracing procedures.19 The finding of substantial prevalence of infection in nonsexual contacts to uninfected persons (“suspects” and “associates” in the older parlance of STD control) corroborates a similar phenomenon observed in a formal contact tracing approach to persons with syphilis,20 and suggests the value of a nontraditional approach.
Like all studies that assess sexual and drug-associated risk factors, our findings are limited by the validity of adolescents’ self-reported behaviors. Also, the use of nonprobability sampling and the choice to conduct the study in an urban area of the south limits generalizability. With these 2 issues in mind, it is informative to consider similarities and differences between adolescents in this study and a nationally representative sample of adolescents participating in the Youth Risk Behavioral Survey (YRBS) conducted by the Centers for Disease Control and Prevention.2 For example, adolescents in our study were about equally likely to ever smoke cigarettes (58%) when compared with adolescents in the YRBS (54.3%) and they were equally likely to report ever using alcohol (75% vs. 74% in the YRBS). However, our sample was much more likely to report marijuana use (65%) when compared with adolescents in the YRBS (38%). Yet, the nonexistent use of crack and injectable drugs in our sample was quite distinct when compared with the 7.6% of YRBS participants reporting any form of cocaine use and the 2.1% reporting injection drug use. The greatest differences were relative to sexual behavior. In our sample, 9 of every 10 males and 8 of 10 females were sexually experienced, as opposed to YRBS data suggesting that 47% of those surveyed in 2005 were sexually experienced. In contrast to YRBS data indicating that 14% reported having sex with 4 or more partners, we found that 4 partners was the average for females, with 14 partners being the average for males. Similarly, YRBS data suggest that 6% sexually debuted before age 14 whereas this age was the average age of debut in our sample. Thus, it appears that youth in our sample were quite different from typical youth primarily about the use of illicit drugs (especially their nonuse of crack and injectable drugs) and key sexual behaviors. Finally, it should also be noted that the pattern of selecting sex partners within a 3-year age discrepancy is relatively consistent with reports from other studies.21,22
In summary, a network-informed approach, coupled with testing for multiple pathogens, provides a more comprehensive view of the multiple STD endemicities that are often examined in isolation. In particular, this approach demonstrates that the overall prevalence of disease in a group may not be fully informative, and the prevalence faced by individuals or groups of people that share characteristics may differ substantially from overall prevalence. We propose that testing for “the STDs”—which, as in this study, can now be accomplished from small samples of blood and urine taken in the field—should become the standard of practice in areas suspected of high endemicity, and that the friends and acquaintances of persons in such social milieus are important for understanding the epidemiology and for effective disease control.
2. Centers for Disease Control and Prevention. Youth behavioral risk factor surveillance—United States, 2005. MMWR Morb Mortal Wkly Rep 2005; 55 (SS-5):1–112.
3. Edgardh K. Sexual behaviour and early coitarche in a national sample of 17-year-old Swedish girls. Sex Transm Infect 2000; 76:98–102.
4. Edgardh K. Sexual behaviour and early coitarche in a national sample of 17-year-old Swedish boys. Acta Paediatr 2002; 91:985–991.
5. O’Donnell BL, O’Donnell CR, Stueve A. Early sexual initiation and subsequent sex-related risks among urban minority youth: The reach for health study. Fam Plann Perspect 2001; 33:268–275.
6. Centers for Disease Control and Prevention. HIV/AIDS Surveillance Report 2003, Vol. 15. Atlanta, GA: US Department of Health and Human Services, Center for Disease Control and Prevention; 2004.
7. Rothenberg RB. Commentary: Sampling in social networks. Connections 1995; 18:105–111.
8. Watters JK, Biernacki P. Targeted sampling: Options for the study of hidden populations. Soc Probl 1989; 36:416–430.
9. Stoner BP, Whittington WL, Hughes JP, et al. Comparative epidemiology of heterosexual gonococcal and chlamydia networks: Implications for transmission patterns. Sex Transm Dis 2000; 27:215–223.
10. Centers for Disease Control and Prevention. Sexually Transmitted Diseases Treatment Guidelines 2002. Recommendations and Reports 2002. MMWR Morb Mortal Wkly Rep 2002; 51:1–78.
11. Deceuninck G, Asamoah-Adu C, Khonde N, et al. Improvement of clinical algorithms for the diagnosis of Neisseria gonorrhoeae
and Chlamydia trachomatis
by the use of Gram-stained smears among female sex workers in Accra, Ghana. Sex Transm Dis 2000; 27:401–410.
12. Dicker LW, Mosure DJ, Berman SM, et al. Gonorrhea prevalence and coinfection with chlamydia in women in the United States, 2000. Sex Transm Dis 2003; 30:472–476.
13. Hart G. Risk profiles and epidemiologic interrelationships of sexually transmitted diseases. Sex Transm Dis 1993; 20:126–136.
14. Pabst KM, Reichart CA, Knud-Hansen CR, et al. Disease prevalence among women attending a sexually transmitted disease clinic varies with reason for visit. Sex Transm Dis 1992; 19:88–91.
15. Centers for Disease Control and Prevention. Sexually Transmitted Disease Surveillance, 2003. Atlanta, GA: U.S. Department of Health and Human Services; 2004.
16. Greenland S, Gago-Dominguez M, Castelao JE. The value of risk-factor (“black-box”) epidemiology. Epidemiology 2004; 15:529–535.
17. Skrabanek P. Has risk-factor epidemiology outlived its usefulness? Am J Epidemiol 1993; 138:1016–1017.
18. Susser M. Does risk factor epidemiology put epidemiology at risk? Peering into the future. J Epidemiol Community Health 1998; 52:608–611.
19. Centers for Disease Control and Prevention. Program Operations Guidelines for STD control. Atlanta, GA: Centers for Disease Control and Prevention; 2004.
20. Rothenberg R, Kimbrough L, Lewis-Hardy R, et al. Social network methods for endemic foci of syphilis: A pilot project. Sex Transm Dis 2000; 27:12–18.
21. DiClemente RJ, Wingood GM, Crosby RA, et al. Sexual risk behaviors associated with having older sex partners: A study of African American female adolescents. Sex Transm Dis 2002; 29:20–24.
© Copyright 2007 American Sexually Transmitted Diseases Association
22. Miller KS, Clark KF, Moore JS. Sexual initiation with older male partners and subsequent HIV risk behavior among female adolescents. Fam Plann Perspect 1997; 29:212–214.