Article Text
PDF
Abstract
Objectives The objectives of this study were to examine the reciprocity of adolescents' heterosexual relationships, the concordance in perceived partner types reported by partners among reciprocal dyads, and the association between dyad-level unprotected sex and relationship types.
Settings and Methods Data were obtained from the Bayview Network Study (San Francisco, California, USA), designed to examine the prevalence of STI risk behaviours and transmission patterns among adolescents between July 2000 and October 2001. For reciprocal dyads, Kappa statistics was used to determine the level of agreement between partner types reported by two sex partners. Multivariate logistic regressions were used to determine the odds of couple's unprotected sex.
Results A total of 782 unique heterosexual relationships were identified. Less than one-third were reciprocally nominated heterosexual dyads. A total of first observed 211 reciprocal dyads were reported by 198 females and 179 males. Agreement on partner type between adolescents and their sex partners among reciprocal dyads was poor, although main–main concordant relationships were the most frequent group (66.4%). Male partner's age and length of relationship significantly increased the odds of a couple's unprotected sex (adjusted OR=1.4, and 1.2, p<0.05), and male partner's frequency of sex significantly decreased the odds (adjusted OR=0.69, p<0.05). The concordance of partner type by two sex partners was not significantly related to couples' unprotected sex when other covariates were taken into account.
Conclusions This unique study advances knowledge about individuals' perception of their heterosexual partner types in reported relationships: the majority of adolescent couples were not reciprocally acknowledged, and whether or not two sex partners agreed on partner type did not change the odds of a couple's unprotected sex even among reciprocal dyads.
- Adolescent
- condoms
- heterosexual behaviours
- risk behaviours
- STD
Statistics from Altmetric.com
Introduction
A consistent predictor of condom use by adolescents is partner type. Adolescents use condoms more consistently with casual, less intimate partners and less consistently with main, stable, intimate partners.1–10 Research has relied on egocentric data in which consistency of condom use is reported by the same person who reported partner type. Studies in which information about the exposure and outcome are collected from a single respondent may result in misclassification, known as information bias. In previous studies of partner type and condom practice, the exposure was measured as individuals' perceptions about their sex partner while the outcome was a couple-level behaviour, introducing an additional source of bias from data from a single respondent.
Generally couples' contraceptive choices shift from male-controlled to female-controlled methods as the length of the relationship, commitment level, and trust between partners increase.11 12 Adolescents trade-off condom use for pill use as perceived STIs and pregnancy risks increase8; therefore, recall bias in reporting not using a condom is likely to be more common among individuals who perceive their sex partners as their main partners. Collecting data on the exposure and outcome in couple studies from one member of a couple may lead to biased results, if partners do not perceive the same level of commitment.
The study reported here utilises dyadic data collected from both members of a partnership to examine the association between condom practice and partner type. We combined two surveys of the longitudinal social network study, the Bayview Network Study, which employed two sampling methods to recruit participants: population-based random sampling and snowball sampling methods. By combining the population-based random sample of index individuals and the snowball sample of index individuals' social friends and sex partners, we were able to construct a population-based sample of adolescent sexual dyads.
Data were collected from predominantly African–American adolescents who referred their social friends and sex partners to the study. These referred friends and sex partners separately provided information about their sexual activities. The objectives of the study reported here were to describe the reciprocity of these heterosexual relationships (two partners agreed that they were sex partners), to estimate the level of the concordance in partner types reported by two partners among reciprocal dyads, and determine the odds of dyad-level unprotected sex by relationship types (main–concordant, discordant, and casual–concordant) among reciprocal dyads.
Methodology
Study description and participants
The Bayview Network Study was designed to examine the prevalence of STI risk behaviours and transmission patterns among adolescents in the Bayview-Hunters Point area in San Francisco between July 2000 and October 2001. The study reported here drew data from the Bayview Network Study, which was approved by IRBs at Johns Hopkins University and University of California, San Francisco. Adolescents of local social and sexual networks were identified and recruited through random digit telephone sampling and household enumeration. Adolescents were eligible for the Bayview Network Study if their age was between 14 and 19, and they were residing in the Bayview-Hunter's Point neighbourhood of San Francisco. A total of 580 adolescents were randomly selected to participate in the Bayview Network Study (Seed Adolescents: figure 1). Seed adolescents were contacted, and 348 self-identified African–American adolescents (females: 203, and males: 145) aged 14–19, residing in Bayview-Hunter's Point neighbourhood of San Francisco were interviewed, for a response rate of 60 percent. The interviewed adolescents formed the index cohort for the study reported here which included 216 sexually experienced adolescents (females: 122 (60% of female index cohort), males: 94 (65% of male index cohort)).
Validations of respondents at the baseline survey.
Snowball sampling methods, used once at baseline, were employed to recruit social friends and sex partners of the index adolescents. Each index adolescent was allowed to nominate up to two closest social friends. In order to increase the size of eligible heterosexual networks for the study, we pooled index adolescents and their social friends together; they served as the partner recruitment cohort of snowball sampling of their sex partners. Adolescents in the partner recruitment cohort were asked to nominate their sex partners at both a baseline and 1-year follow-up survey. Snowball sampling of sex partners at each survey was repeated three times, each of which allowed respondents (adolescents in the partner recruitment baseline cohort, sex partners who were referred by the cohort, and sex partners referred by the sex partners of the cohort) to nominate at a time up to six sex partners in the 3 months prior to participation in both surveys.
Although homogeneity in sexual behaviours within friend clusters was assumed, the advantage of increasing the size of the partner cohort using index adolescents and their friends as independent observations in the partner recruitment cohort was assumed to exceed problems due to the clustering effect. The recruitment of social friends and sex partners was conducted independently from that of index adolescents; thus, the data from the referred participants were not influenced by the referring individuals. The recruitment procedures of the Bayview Network Study are discussed in more detail elsewhere.13
Definition and eligibility of sexual networks and sex partners
We identified all possible sequences of sexual relationships. Relationships were based on reports of the index cohort, their sex partners and sex partners of the sex partners interviewed at baseline and 1-year follow-up surveys. Figure 2 shows the myriad of relationships in the sample. We excluded relationships of those who initiated having sex between the baseline and 6 months surveys from the analysis as they were considered different individuals compared with those who were already sexually experienced at the beginning of Bayview Network Study. A sequence of sexual relationships was eligible for the analysis when each member of the sequence was nominated and interviewed. Individuals in the partner recruitment cohort or their sex partners could serve as roots of sequences of sexual relationships; sequences of sexual relationships starting from the partner recruitment cohort (an individual in the partner recruitment baseline cohort→first-generation sex partner→second-generation sex partner), and starting from first-generation sex partners (first-generation sex partner→second-generation sex partner→third-generation sex partner).
Examples of local sexual networks and partner nomination. †Individuals in the partner recruitment cohort or first-generation sex partner could serve as a root of a directed walk of length 2 (ie, a sequence of sexual relationships). Partner nomination could start from an individual in the partner recruitment cohort (individual in the partner recruitment cohort→first-generation sex partner→second-generation sex partner) or first-generation sex partners (first-generation sex partner→second-generation sex partner→third-generation sex partner).
Sequences of sexual relationships starting from individuals in the partner recruitment cohort could include reciprocal dyads and triads; however, sequences of sexual relationships starting from first-generation sex partners included only reciprocal dyads due to the eligibility of the sequences for this analysis (every member of a sequence of sexual relationships must be interviewed) (figure 3).
Partner nominations and a local social network.
Only reciprocal dyads were the focus of the objectives to estimate the level of the concordance in partner types reported by two partners, to determine the odds of dyad-level unprotected sex by relationship types (main–concordant, discordant, and casual–concordant), and to compare the odds of dyad-level unprotected sex by partner types reported by females and males. Adolescents and their sex partner(s) in reciprocal dyads not only acknowledged each other as sex partners but also identified partner types (ie, main or casual). We selected first observed reciprocal dyads when the same reciprocal dyads were observed at both the baseline and 1-year follow-up surveys to create a homogeneous sample in terms of the stage of relationships. Reciprocal dyads included both concordant and discordant dyads with respect to whether or not they identified each other as main or casual partners. An individual could form more than one reciprocal sexual relationship simultaneously.
Data analyses and variables
Description of the baseline characteristics and the local sexual networks
The distributions of female and male adolescents and their sex partners were compared by race, current age, age at first sex, number of life-time sex partners, number of sex partners in the 3 months prior to the baseline interview, prevalence of chlamydia and gonorrhoea, and pregnancy history for the 211 reciprocal dyads. Two-sample t-tests and χ2 tests were employed to test equal means for continuous variables and equal proportions for categorical variables, respectively.
Reciprocity of adolescents' heterosexual relationships
An identification number assigned for each respondent in the Bayview Network Study was used to verify if a sex partner reported by a respondent also nominated the respondent as their sex partner. We distinguished a reciprocal dyad through the process of matching sex partners' identification numbers. We also determined concurrent relationships in the sample.
Concordance in partner types among reciprocal dyads
A partner type (main, or casual) for each opposite-sex partner was labelled by female and male respondents. Two partner types reported in a reciprocal dyad were used to verify if partner types were concordant between two sex partners, as defined by females' report–males' report: main–main, main–casual, casual–main, and casual–casual.
Association between concordance in partner types and dyad-level unprotected sex
Outcome variables
The outcome of this analysis was condom use at last sex (used, did not use). There were 1.4% missing observations for condom use at last sex. Interviews of two partners in a dyad did not always happen at the same times; therefore we could reasonably assume that two partners referred to the different sexual event. In fact, more disagreement in partners' responses about condom items was observed the longer the time between the two interviews, although there was no significant association between the interval between two interviews and the interval between dates of last reported sex. The agreement between females' and males' responses to these two items were found to be poor (κ=0.43 (71.7% agreement)) for condom use at last sex.
Using condoms inconsistently and not using condoms are both behaviours which place adolescents at risk for STIs acquisition. We dichotomised unprotected sex by recording discordant responses to condom use at last sex as not using condoms: where at least one partner reported not using condoms at last sex versus both partners reported using a condom at last sex.
Exposure variables
Reported relationship types were defined as main–concordant: main–main, discordant: main–casual or casual–main, and casual–concordant: casual–casual for each reciprocal dyad. We used a relationship type for each reciprocal dyad as an exposure variable to estimate the odds of dyad-level unprotected sex.
Covariates
Covariates known to potentially confound the association between perceived partner types and condom use at last sex were self-reported age (years), reported length of relationship at the time of survey (<1 month, 1–3 months, 4–6 months, 7–12 months and >12 months), frequency of sex in the past 3 months (fewer than a few times per month, a few times per month, 1–2 times per week and everyday), and partner-specific perceived pregnancy wantedness. We repeated the analysis with the difference in age (years) between the two partners instead of self-reported ages; however, the results did not change, and we present the results of the analysis with the model including self-reported age. All covariates were partner-specific, except respondents' self-reported age.
Statistical analyses
Hierarchical modelling was used to estimate ORs, with multiple heterosexual relationships nested in an adolescent. Few reciprocal dyads were nested in the same adolescents; the average number of reciprocal dyads per female respondent was 1.07 (211/198), and per male respondent was 1.18 (211/179). ORs of dyad-level unprotected sex and robust standard errors were estimated, controlling for females clustered in male partners with the cluster option in logistic regression in Stata to produce weighted average estimates by cluster sizes. The ORs were compared in three logistic regression models: unadjusted, adjusted for all covariates and adjusted for a parsimonious set of covariates selected by backward stepwise reduction. Bias corrected standard errors and 95% CIs were produced using bootstrap estimation. Stata V.10.0 SE (Stata Corporation, College Station, Texas) statistical analysis software was used for the analyses.
Results
Description of the baseline characteristics and local sexual networks
A total of 782 unique sequences of sexual relationships were confirmed by 333 females and 268 males (table 1). When partner recruitment started from female adolescents, 151 females referred 202 male partners, 181 (89.6%) of the interviewed referred 454 female partners, branching out to 483 female–male–female sequences. When partner recruitment started from male adolescents, 134 males referred 236 female partners, 214 (90.7%) of the interviewed referred 266 male partners, branching out to 299 male–female–male sequences. Males referred on average more opposite-sex partners than females (1.8–2.5 female sex partners per male vs 1.2–1.3 male sex partners per female).
Characteristics of participants at the baseline survey
Reciprocity of adolescents' heterosexual relationships
Less than one-third (241) of all identified sequences of sexual relationships (782) were reciprocal dyads reported at the two surveys. Only 30 reciprocal dyads (12.4%) were observed twice at both surveys. A total of 181 reciprocal dyads were observed once, either at the baseline survey or at 1-year follow-up survey. A total of 211 first observed reciprocal dyads were confirmed by 198 females and 179 males. The characteristics of individuals who formed the 211 reciprocal dyads were relatively similar to the entire sample (table 1). The majority of reciprocal dyads were monogamously reciprocal; however, concurrent relationships were observed for close to a quarter of reciprocal dyad(s): 160 reciprocal dyads (75%) were monogamously reciprocal, 46 reciprocal dyads (22%) were reported by 23 partners (10 female, 13 males), each of whom had two reciprocal dyads; five reciprocal dyads (2.4%) were reported by one partner (male) who had five reciprocal dyads.
Concordance in partner types among reciprocal dyads
Among the reciprocal dyads in which females identified their sex partners as their main partners, most (75.7%) males identified the female as a main partner (table 2). Among the reciprocal dyads in which females identified their sex partners as casual, about half of (46.2%) males also reported the female as a casual partner, but the other half (53.8%) reported the female as a main partner. Similarly, among the reciprocal dyads in which males identified their sex partners as casual, most (79.0%) females identified the male as a main partner. Among the reciprocal dyads in which males identified their sex partners as main partners, most (91.0%) females also reported the male as a main partner. κ Statistics testing the null hypothesis that there was no more agreement between partner types reported by females and males than might occur by chance was 0.14 (72% agreement), indicating that agreement was poor, yet higher than expected by chance alone.
Partner types identified by respondents' gender among 211 reciprocal dyads
Association between concordance of partner types and dyad-level unprotected sex
As shown in table 3, there was a weak positive association between relationship type characterised by commitment level and dyad-level unprotected sex (p=0.20). The ORs for dyad-level unprotected sex by relationship types are shown in table 4. Main–concordant was the majority group and served as a reference group in the analysis. Model 1 adjusted for female and male partners' self reported age, perceived length of the relationship, frequency of sex, and desire for pregnancy with the partner. Partner type concordance (main–main, discordant, casual–casual) was not associated with the odds of inconsistent condom use even after further adjustment for potential confounders (Model 2).
Dyad-level unprotected sex* by relationship types (n=211)
OR (95% CI*) for dyad-level unprotected sex† by relationship types
Using univariate logistic regressions, dyad-level unprotected sex was compared between perceived main partner type and casual type; the OR was 1.3 (95% CI 0.6 to 3.2) for females, and 2.0 (95% CI 1.0 to 3.7, p=0.041) for males (table 5). When other partners' perceived partner type was controlled in the model (unadjusted multivariate model), the OR of dyad-level unprotected sex was 1.2 (95% CI 0.5 to 2.8) for females' perception of main-partner and 1.9 (95% CI 1.0 to 3.7 p=0.049) for males. The two ORs did not differ significantly (p=0.39).
OR (95% CI*) of dyad-level unprotected sex† by perceived partner types
Discussion
This study used network data, and focused on reciprocally nominated heterosexual dyads. Reports from both partners of a single sexual relationship were available in the Bayview Network Study so that it was possible to reveal more accurately the ambivalent nature of sexual relationships among adolescents and young adults. Less than one-third of identified sexual relationships were reciprocally acknowledged by both partners. It was very common that an adolescent and their recent sex partner(s) did not agree on who the other partner was in their sexual relationship. When a sex partner was separately interviewed and given the opportunity of listing up to six sex partners in the past 3 months, two-thirds of these sex partners did not nominate the individual who initially referred them to the study. This finding was more frequently observed among male sex partners who were referred by a female in the partner recruitment cohort. This finding suggests that the studies using the traditional survey approach of relying on egocentric data may include up to two-thirds of unconfirmed sexual relationships, or relationships of such insignificance that they are not recalled.
Adolescents' sexual relationships overlap in time, and concurrent relationships among adolescents are commonly observed in past research using egocentric data.14–19 Our study confirmed this observation; even within reciprocal dyads; the concurrency rate was estimated at nearly 25% of reciprocal relationships. About 30% of partner types identified by female and male partners of reciprocal dyads were discordant in which the two partners did not agree on whether their partner was main or casual. This corresponded to a κ statistics of 0.14 (72.0% agreement), suggesting poor agreement between two persons' perceived partner type. Females, however, were significantly more likely to report their male partner as main compared with males when they disagreed on reports of partner types (76.3% vs 23.7%, p<0.001, n=59 discordant relationships). Females were also more likely than males to report not using condoms when they disagreed on reports of condom use at last sex (40.0% vs 60.0%, p<0.001, n=55). Our findings suggest that females may misperceive their relationship quality and their partners' risk behaviours20 21 We cannot rule out, however, the possibility that the responses of males to the questions about condom use may reflect socially desirable answers, while females may have reported condom use more accurately.
Partners agreed on their partner type as main in 73.0–87.7% of reciprocal dyads, among females and males respectively. This result encourages couple interventions of STI counselling or treatment for adolescent and young couples; both partners must participate in these interventions, and they may also require both to perceive the other as their main partner.
There are several limitations of this study. The dyad-sample size for the multivariate logistic regression was 196 instead of 211 due to missing data among covariates; however, the results were robust in various models after controlling for potential confounding factors. Our analysis used only first observed relationships when they were reported at both baseline and 1-year follow-up surveys (n=30). This could potentially attenuate the effect of perceived partner type in our analysis, as it could include a relationship at early phase and later phase. We believe that the model including reported length of relationship in the survey took into account this potential bias.
Another limitation of this study is that data were collected in 2000 and 2001 when the source or interactions of social networks differ from many today. Social networking among youth today is much more dominated by the use of technology, particularly the internet, text messaging and other modes of communication.22 It is unclear how the sources may affect the relations reported here, or alter the nature of heterosexual relationships among adolescents and young adults, even from disadvantaged backgrounds.
Finally, the timing of the interview of the sex partners differed from that of the index adolescents, increasing the likelihood of discordant reports about condom use. We measured dyad-level condom use to reflect different responses on condom use over time. Two respondents may have referred to different sexual events in time. Nevertheless, when two partners' responses on condom use were discordant, we assume that at some point in the past, this couple used condoms (as reported by one of partners) but at another point in time the couple did not use condoms (as reported by the other partner). Therefore, we assumed that the couple did not consistently use condoms over a period of time. Future studies of this nature should attempt to synchronise these interviews.
Our investigation was innovative and had several strengths. It used sociometric network data and examined partners in reciprocally acknowledged sexual relationships. The outcome variable was partner-specific and indicated couples' unprotected sex when at least one partner reported not using condoms. The exposure variables were perceptions not only about sex partner types but also about relationship types, informing concordance of perceived partner types by the two partners in a dyad, and adjusting for both females' and males' self-reported individual and relationship characteristics. As expected, very few casual–casual concordant relationships were identified (12/211) among reciprocal dyads. The small cell size could be responsible for the counterintuitive result that did not suggest any trend in couple's inconsistent condom use (table 4). Individuals in casual–casual concordant relationships were by definition hard to reach subpopulation of adolescents. Our study had the ability to include this sample to construct three distinct relationship types in the analysis.
A small number of males reported sexual relationships with more than one of the females in the reciprocal dyads sample. Self-selection in our sample of reciprocal dyads may have occurred at entry into the Bayview Network Study (about 65% response rate of the index cohort), at partner nomination by the index cohort, at sex partners' entry and at sex partner's nomination of their sex partners.
There was no strong evidence of differences in the odds of dyad-level unprotected sex for main-partner status as perceived between females and males. Multivariate regression analyses suggested that couple-level inconsistent condom use was determined predominantly by males' characteristics: male's age (AOR: 1.2 (p<0.05)) and length of relationship (AOR: 1.4 (p<0.05)), and male's frequency of sex (AOR: 0.7 (p<0.05)).
Conclusions
The sexual relationships reported by adolescents in our study were often not reciprocally acknowledged. Adolescents' perceptions of their sexual relationships were largely concordant in main type with their sexual partners' once both partners acknowledged the others as their sex partners. Female adolescents, however, were more likely to consider their partner as main partners than male adolescents. Neither perceived partner types of each partner nor concordance in partner types showed an independent effect on couple level inconsistent condom use when other related factors were taken into account. Adolescents were at risk of unprotected sex and STIs regardless of their perceived partner type or concordance in partner type.
This study supports couple interventions for STI treatment for adolescents and young couples when relationships are reciprocal dyads. Reciprocal partner nomination may be more informative than perceived partner type, and help to achieve successful STIs interventions among young couples. The study results also suggest that there is considerable discordance in perceived partner type and that this may be a problem in developing effective interventions focused on couples. Moving to more couple-focused interventions may not improve the effectiveness of STI preventive interventions. The findings of this study have important implications for counselling, testing and treatment among adolescents and young couples.
Key messages
Sexual relationships among adolescents and young adults are often not reciprocally acknowledged by two partners, and perceived partner type (main, or casual) is often discordant even among reciprocally acknowledged couples.
Adolescents and their sex partners use condoms inconsistently and they are at risk of STIs regardless of perceived partner type, and even when both partners consider the other as a casual partner.
Couple-focused interventions for STI treatment may be successful for adolescents and young couples only when both partners acknowledge the other as their sex partner.
Acknowledgments
We thank K Bandeen-Roche for her instruction on data management and statistical analysis. We also thank C Fichtenberg for sharing her experiences in network data management and analysis.
References
Footnotes
Funding The study reported here drew data from the Bayview Network Study, which was funded by National Institute of Allergy and Infectious Diseases, National Institutes of Health (5 UO1 AI47639).
Competing interests None.
Ethics approval Ethics approval was provided by the IRBs of The Johns Hopkins University and University of California, San Francisco.
Patient consent Obtained.
Provenance and peer review Not commissioned; externally peer reviewed.
Linked Articles
- Whistlestop tour
- Correction