|Home | About | Journals | Submit | Contact Us | Français|
Sex partner concurrency is an important risk factor for STI transmission. Understanding how adolescents conceptualize the length of their relationships when they break up and get back together is essential to the assessment of concurrency.
A prospective cohort of 392 late adolescents and emerging adults, aged 14 – 19 years at baseline, were recruited from two clinics in Baltimore, MD and interviewed semi-annually for 3 years. At each interview, participants were asked to report on all of their sexual partners in the previous six months, the length of relationship and whether they thought their partner had other sex partners. For relationships that had broken up and gotten back together, reports of length of the relationship were compared before and after the break up. Random effects logistic regression was used to examine the association between length of relationship and both break up and partner concurrency.
For relationships that ended and subsequently got back together, participants considered the length of relationship to include the period when they were broken up. Longer relationships were at increased odds of both having broken up and gotten back together [OR = 1.04, 95% CI: 1.02, 1.05] and of partner concurrency [OR = 1.03, 95% CI: 1.02, 1.04]. The magnitude of the odds of concurrency was greater for relationships that had broken up and gotten back together [OR = 1.07, 95% CI: 1.02, 1.11].
Findings from this study emphasize the need for an improved understanding of the association between the temporal dynamics of late adolescent and emerging adult romantic relationships and concurrency.
Both empirical and simulation studies have found that concurrent sexual partnerships play a pivotal role in rates of STI transmission. Sex partnerships that overlap in time allow for more rapid spread of STIs than serially monogamous relationships.(1) Further, sex partner concurrency has been shown to be an important STI risk factor, independent of an individual’s number of sex partners, as it links the individual to a broader sexual network.(2) As late adolescents and emerging adults have the highest rates of STIs, the issue of concurrency is particularly salient for this population.
Data from a nationally representative sample, found that adolescents and emerging adults’ patterns of sexual partnerships in contrast to their number of sex partners was associated with risk of an STD.(3) While the majority of sexually active adolescents in this study had only a single partner in the 18 month study period, those who had sequential or concurrent partnerships were more likely to report having been diagnosed with an STD in the past year. The introduction of a new sex partner has been repeatedly shown to be a predictor of an incident STD infection.(4;5) While adolescence is frequently characterized as a period consisting of multiple sexual partners with frequent partner turn-over,(6;7)data suggest that many adolescents have longer term relationships that may last well over a year, similar to adults.(8;9) Studies have found relationship duration to be a correlate of concurrency.(10;11) There may be several explanations for this association. Longer relationships may provide more opportunity for a partner to take side partners, often referred to as contained concurrency (12;13)or longer relationships may be more likely to end with transitional concurrency where by a new relationship begins before the existing relationship ends.(14)
However, it may also be that these so-called longer length relationships are in fact just a combination of short serial relationships with one partner interrupted by a relationship with another partner. Length of relationship may be a marker for serial relationships, which may or may not serve as concurrent relationships with respect to STI transmission if there is a sufficiently long gap between when the relationship ends and resumes.(15;16) Understanding how adolescents conceptualize these relationships is essential to the assessment of concurrency. Using data from a prospective cohort of adolescents and young adults, the goal of this study was to examine the association between length of relationship and partner concurrency; specifically, we aim to describe how adolescents reported the length of relationship for those partners with whom they have broken-up and then got back together; second, we examine whether longer relationships are more likely to be relationships where there was a break up and then they got back together; and third, we examine whether partner concurrency is more likely in relationships where there was a break up.
A prospective cohort of late adolescents and emerging adults was recruited from two urban health clinics, a hospital-based adolescent medicine clinic and a public STD clinic, in Baltimore, MD. Both clinics serve similar populations, primarily African American youth, aged 12–21 yrs. Between July 2000 and September 2002, 392 participants were enrolled in the Perceived Risk for Sexually Transmitted Diseases (PRSTD) study. Eligibility criteria included age between 14 and 19 years, vaginal or anal intercourse with an opposite sex partner in the preceding 6 months, English-speaking, and residence in the Baltimore metropolitan area. Seventy-four percent of those screened to be eligible agreed to participate. All participants provided written informed consent; parental consent was not required as Maryland law allows for adolescents to consent for confidential health services. The Johns Hopkins University Institutional Review Board approved the study protocol.
Participants were interviewed at baseline and at six-month intervals for 3 years (a total of 7 interviews). Trained research assistants conducted the baseline and semi-annual interviews face-to-face. Participants received $25 as compensation for completing the baseline, 6-, and 12-month interviews; $35 for 18-month; $45 for 24-month; $55 for 30-month, and $65 for 36-month interview.
At each interview, participants were asked to list, using first name only, all of their sexual partners in the three months preceding the interview. Interviewers came to the follow-up visits with a list of partners named in previous interviews. Interviewers compared the list of partners named at the current visit to the list of partners named at any previous visit with the participant in order to de-duplicate partners and to ensure each partner was assigned a unique ID. For this analysis, our primary variables of interest were length of relationship and partner concurrency, which were assessed at each semi-annual visit.
Length of the sexual relationship was measured in two ways. Participants were asked, “How long have you been (or were you) having sex together?” and subsequently reported on the number of days, weeks, months or years. Length of relationship was also measured by calculating the difference in the dates of first sex and last sex reported by the participant. Relationships were categorized as having been broken up and gotten back together if the participant completed at least one follow-up interview in which the partner was not named. To assess partner concurrency participants were asked, “To the best of your knowledge, did he/she ever have other sex partners while you two were having a sexual relationship?” with answer choices yes, no or don’t know. Partner concurrency was dichotomized into yes vs. no or don’t know.
We used t-tests to examine the length of relationship reported by participants before and after the break-up. We then used t-tests and tests of proportions to examine differences in length of relationship, concurrency and other relationship characteristics between relationships that broke up and got back together and those that were uninterrupted. To test our hypotheses, we first examined whether length of relationship was associated with partner concurrency and second, whether this association differed for those relationships that broke up and got back together compared to those that were uninterrupted. We then examined whether length of relationship was associated with having been broken up and gotten back together. Each relationship was in the analysis only once. For partners who were mentioned at multiple interviews, data from the most recent interview was used. As both outcomes were dichotomous, we used multilevel mixed effects logistic regression to account for the correlation of multiple relationships nested within an individual.(17) Participant’s age in years, which was treated as continuous, gender with male as the reference group, and whether the participant had concurrent sex partners (yes/no) during the relationship were included as covariates in the models as well as mother’s education, with having completed high school or less education as the reference group, in order to avoid confounding by socioeconomic status. Sensitivity analyses were conducted to determine the impact of dropping the “don’t know” response for the partner concurrency measure (less than 2% of responses). Comparing yes responses to only no responses had virtually no impact on our estimates. All analyses were conducted using Stata 11.(18)
Sixty-three percent (247/392) of the cohort was retained over three years. The majority of attrition occurred between the baseline and six-month interview. Among participants who completed the six-month interview, the retention rate at 36 months was 97%. There were no differences in age or risk behavior at baseline between those completing the study and those lost to follow-up.
Participants were 17.2 years old on average at enrollment, 97% African American race and 76% female. Sixty-two percent of participants reported that their mother had less than or equal to a high school education. The mean age at sexual debut was 14.8 yrs. The median number of lifetime sexual partners was 6,59% of the sample reported ever having an STD at baseline and 47% of the cohort reported using a condom use at last sex. Participants reported a mean (standard deviation (SD)) of 4.0(2.8) sex partners over the three-year follow-up period, 60% of which were classified as main partners. Over the follow-up period, 15% of participants had a relationship that broke up and got back together (59/392). Table 1 presents baseline demographic and behavioral characteristics of participants who had a relationship that broke up and got back together. Using a two-tailed alpha of 0.05, there was no age difference between adolescents who got back together with a partner compared to adolescents who did not get back together with a partner. As has been published previously, participants from the two study sites were found to be similar with regard to demographic and behavioral characteristics, and differences were not statistically significant.(19)
Over the follow-up period, five percent of the relationships (77/1,489) had a break up and got back together. Of those, 70 relationships that broke up and got back together and 976 that were uninterrupted had sufficient data to examine length of relationship. After dropping impossible values, the mean (SD) length of relationship for all relationships was 16.0 (19.8) months when asking how long together and 15.6 (19.4) months when subtracting dates of first and last sex. The median using both measures was 7 months and the range was less than one month to 96 months.
For relationships categorized as having been broken up and gotten back together, we compared participants’ reports of the length of relationship at the last visit in which the partner was named before the break up to the first visit when they were back together. Table 2 presents results for relationship duration using two measures of length of relationship: asking how long the two had been together and subtracting dates of first and last sex. When asking how long together, the mean (SD)length of relationship reported at the visit before the break-up was 18.0 (2.2)months compared to an average of 27.6 (2.7)months reported at the visit when the couples were just back together[p<0.001]. Overall, the date of first sex was found to be the same date reported before the break-up and when they were back together. Subtracting dates of first and last sex, the average length of relationship calculated at the visit before the breakup was 16.4 (2.1)months compared to an average of 25.2 (2.5)months at the visit when the couples were just back together[p<0.001]. When comparing the two relationship duration measures to one another, the asking how long together measure produced longer relationship duration results than the subtracting dates measure both when assessed before the break up (difference: 1.6 months; p < 0.001) and when back together (difference: 2.4 months; p < 0.001).
On average, the length of the gap in which the partner was not mentioned or the duration of the break-up was 14.1(SD= 3.7) months and ranged from 12–24 months. Among the 74% of the relationships that had a 12 month gap in the relationship, 25% of participants did not have another sex partner during the break-up, 48% had one sex partner during the break up, and the remaining 27% had two or more sex partners during the break up [mean (sd) 1.3 (1.4) partners, range: 0–8]. Among the 19% of the sample that had an 18 month gap in the relationship, 8% of participants did not have another sex partner during the break-up, 46% had one sex partner during the break up, and the remaining 46% had two or more sex partners during the break up [mean (sd) 1.7 (1.3) partners, range: 0–5].
Using both duration measures, relationships that broke up and got back together were significantly longer than those that were uninterrupted [question: 33.9 vs. 15.5 months, p<0.001; dates: 34.6 vs. 14.4 months, p<0.001]. Relationships with partner concurrency were significantly longer than those without partner concurrency [question: 27.6 vs. 15.6 months, p<0.001; date: 25.5 vs. 14.8 months, p<0.001].
As both length of relationship measures produced the same inferences, results of regression analyses are reported using only the question measure (i.e. how long together) for simplicity. Table 3 presents associations between length of relationship and partner concurrency. For every additional month in the relationship, there was a 3% greater odds of the partner having additional sex partner(s)[odds ratio (OR)= 1.03, 95% confidence interval (CI): 1.02, 1.03]. This estimate was unchanged after controlling for index’s age, gender, concurrency and mother’s education [OR = 1.03, 95% CI: 1.02, 1.04]. We examined the interaction of gender and length of relationship in the multivariable model and found no evidence of effect modification[data not shown].
Length of relationship was associated with a relationship having broken up and gotten back together. As presented in Table 3, the unadjusted regression indicates that for every additional month in the relationship, there was a 4% greater odds that the relationship was one that had broken up and gotten back together [OR = 1.04, 95% CI: 1.03, 1.05]. The association was unchanged after adjusting for covariates [OR = 1.04, 95% CI: 1.02, 1.05]. We found no evidence of effect modification by gender [data not shown].
We examined the association between length of relationship and partner concurrency stratified by relationship category and found that for those relationships that broke up and got back together, for every additional month in the relationship, there was a 5% greater odds of the partner having additional sex partner(s)[OR = 1.05, 95% CI: 1.01, 1.08] with an adjusted odds ratio indicating a 7% greater odds of the partner having additional sex partner(s)[OR = 1.07, 95% CI: 1.02, 1.11]. As there were only two individuals who had multiple relationships that broke up and got back together, logistic regression was used to estimate the association of length of relationship and partner concurrency in this strata. For the uninterrupted relationship strata, the unadjusted and adjusted associations resembled that of the overall sample [Adjusted OR (AOR) = 1.03, 95% CI: 1.02, 1.04].
Table 4 presents relationship and partner characteristics for all the relationships and then stratified by relationship category. Significant differences were found between relationships in which there was a break up and then subsequently got back together compared to uninterrupted relationships. In relationships that got back together, participants reported a higher percentage had children together, the partner was known by all or almost all of their friends, the partner was older on average, and more of these partners had a history of incarceration. No statistically significant differences at the p <0.05 level of significance were found for index concurrency, partner giving the index money, index giving the partner money, totally loving the partner, trusting the partner completely, partner ever selling or using drugs, or partner ever having an STD.
In a post-hoc analysis, we examined reports of partner concurrency at the last visit before the break-up and then at the first visit in which the partner re-appears. A higher prevalence reported that their sex partners had other sex partners at the last visit before the break-up compared to the first visit in which the partner re-appears (34% vs. 22%, p <0.10). In addition, we examined partner type and reports of condom use at last sex at the last visit before the break-up compared to the first visit in which the partner re-appears. No difference was found in condom use, 50% reported using a condom at last sex before the break up compared to 49% percent when back together. Participants were asked to classify their partner as main or casual. More partners were classified as main when they got back together compared to the visit before the break up (72% vs. 53%, p < 0.01).
Both concurrency and partnership duration have been beset with measurement issues.(20;21) Findings from this study further illustrate the complexities of examining the length of late adolescent and emerging adult romantic relationships and the implications on concurrency. Late adolescent and emerging adult relationships of longer duration were more likely to be relationships where there was a break up and then they got back together. Overall, for relationships that ended and subsequently got back together, participants considered the length of relationship to include the period when they were broken up. We found that late adolescents and emerging adults conceptualize relationships that are temporarily interrupted in their entirety rather than in parts. Thus, in asking late adolescents and emerging adults how long they have been with their partner as well as subtracting dates of first and last sex, the gap in the relationship was not detected. Results from this study have important implications on estimates of relationship stability, timing of behaviors within relationships and assessments of concurrency.
Partner concurrency was more likely in adolescent relationships of longer duration. While partner concurrency was associated with length of relationship for all relationships, the magnitude of the association was greater for those relationships that had a break-up. The qualitative work by Gorbach and colleagues identified that partner concurrency takes place largely during relationship transitions.(14) Breaking up and getting back together would introduce multiple transitions within the relationship, thereby increasing the opportunities for partner concurrency. Our post-hoc analysis found some empirical support for transitional concurrency. Participants reported a higher prevalence of partner concurrency before the break up compared to when the couple was back together. It is possible that concurrency was the cause of the break-up. The current data; however, could not distinguish the partner having a side partner during the relationship, from an intermittent partner, where the additional sex partner emerged during the gap in the relationship. As a result, the impact of a partner having additional sex partners on STI transmission is ambiguous. These findings may add additional insight to the work of Nelson and colleagues who found only fair agreement when examining measures of concurrency by overlapping dates and direct question.(11) Relationships that break up and get back together may appear to be concurrent when using overlapping dates, but may not be considered concurrency as assessed by the direct question.
Previous authors have shown that being in a new partnership is an important predictor of biologically confirmed STD infection.(4) An underlying assumption with respect to late adolescent and emerging adult relationships is that partner turn-over is largely to introduce a new partner. The current findings provide evidence that serial relationships can include the reintroduction of a previous partner. Conceptualizing relationships in their entirety rather than in starts and stops has implications on behaviors within the relationship. In our post-hoc analysis we found that upon re-introduction of the partner, levels of condom use were similar to that before the break-up, despite the potential for additional sex partners during the gap.
Prior research has found that relationship quality predicts partner change.(5) The current study found qualitative differences between partners who reappeared and those that do not, including knowing most of the participants’ friends, strong feelings of love, older age of the partner and the partner having been incarcerated. It is possible that some of the relationship gaps were a result of the partner’s incarceration. Research in adult samples have found that a substantial number of primary relationships dissolve during periods of incarceration, which leads to increased HIV/STI risk.(22) Future work is needed to examine late adolescent and emerging adults’ commitment to partners during and following periods of incarceration.
There are limits to what can be inferred from these analyses. While this was a prospective cohort study, our data are cross-sectional with respect to the relationship. Relationships that we have categorized as uninterrupted are a mix of relationships that are ongoing and terminated, some of those relationships that have ended may get back together. Our six month intervals between follow-up interviews may have missed shorter gaps in relationships. Thus we have likely underestimated the frequency of relationships getting back together. However, a major strength of this study was the prospective design in that we could detect a partner coming and going over time. Also of note is that perceived partner concurrency is an imperfect proxy for actual partner concurrency;(23;24)however, there is no evidence to suggest that misclassification of actual partner concurrency is differential by length of relationship, thus we likely present an attenuated estimate of the odds ratio.
Findings from this study emphasize the need for an improved understanding of the association between the temporal dynamics of adolescent and emerging adult romantic relationships and concurrency. The next steps in this line of research will be to follow adolescent and emerging adult relationships from beginning to end in order to examine relationship trajectories.
The research was supported by the National Institute of Allergy and Infectious Diseases (Grant # R01AI36986), the Eunice Kennedy Shriver National Institute of Child Health & Human Development (Grant# R01HD058309) and the National Institute on Drug Abuse (Grant# F31DA019822).
The authors have no conflicts of interest to report.