CONTRACEPTIVE USE PATTERNS ACROSS TEENS’ SEXUAL RELATIONSHIPS: THE ROLE OF RELATIONSHIPS, PARTNERS, AND SEXUAL HISTORIES

Abstract

By using data from the National Longitudinal Study of Adolescent Health, we examine how adolescent relationship characteristics, partner attributes, and sexual relationship histories are associated with contraceptive use and consistency, incorporating random effects to control for respondent-level unobserved heterogeneity. Analyses show that teens’ contraceptive use patterns vary across relationships. Teens with more-homogamous partners, with more-intimate relationships, and who communicate about contraception before sex have greater odds of contraceptive use and/or consistency. Teens in romantic relationships, and who are older when engaging in sex for the first time, have greater odds of ever using contraceptives but reduced odds of always using contraceptives. Teens continue habits from previous relationships: teens with experience practicing contraceptive consistency and females who previously have used hormonal contraceptive methods are better able to maintain consistency in subsequent relationships. Also, relationship and partner characteristics are less important for females who previously used hormonal methods.

The initiation of heterosexual romantic relationships represents a key component of adolescent development (Carver, Joyner, and Udry 2003). The majority of high school teens have had a romantic relationship (Carver et al. 2003), and almost half of high school-age teens have had at least one sexual experience (Centers for Disease Control and Prevention 2006). Adolescent romantic relationships have several characteristics that distinguish them from friendship relationships (Giordano 2003), and several of these distinctive characteristics are potentially associated with teens’ reproductive health decisions. A better understanding of teen relationships and their association with contraceptive use and consistency will help parents, program providers, and teens themselves continue to reduce high rates of unintended teen pregnancy, childbearing, and sexually transmitted infections (STIs) in the United States (Abma et al. 2004; UNICEF 2001; Weinstock, Berman, and Cates 2004).

Distinct domains of adolescent romantic relationships (as opposed to nonromantic friendships) include a higher prevalence of relationship asymmetries or differences between partners, issues of exclusivity and intimacy with romantic partners, shorter relationship duration, and awkwardness in relationship communication (Giordano 2003; Giordano, Manning, and Longmore 2006). The research literature suggests that among sexually active adolescents, several of these important relationship dimensions may influence contraceptive use, including condom use. However, until recently, the majority of research on adolescent sexual activity and contraceptive use has focused on the influence of family background and individual characteristics, with less research attention paid to the importance of relationships and partner characteristics in reproductive health decisions (Giordano 2003; Kirby, Lepore, and Ryan 2005). Many of the studies that have focused on partner characteristics have used small, nonrepresentative samples; examined only one or two key dimensions of sexual relationships; and/or examined single-time measures of contraceptive use, such as use at the first or last sexual encounter, although consistency of contraceptive and condom use are more accurate predictors of unintended pregnancy and STIs (Ford and Lepkowski 2004; Glei 1999).

This study builds on previous research by using a detailed, sequential sexual relationship history among a nationally representative sample of high school males and females from two waves of the National Longitudinal Study of Adolescent Health. Because the data file does not report method-specific contraceptive consistency, our focus is on adolescent contraceptive use and consistency broadly defined rather than on the consistency of condom use or hormonal methods separately. Teens may report characteristics of several sexual relationships, so we first examine whether adolescent contraceptive use patterns differ across teen sexual relationships. Second, we examine whether several critical dimensions of adolescent sexual relationships and partners are associated with contraceptive use and consistency across relationships. Third, we assess whether adolescents’ previous experiences with sexual partners and with negotiating contraceptive use are associated with contraceptive use and consistency in their current relationship. Because we anticipate that contraceptive use and consistency in early and later relationships are endogenous, we incorporate random-effects models to control for respondent-level unobserved heterogeneity. Fourth, we examine whether the association between sexual relationships and contraceptive use differs by contraceptive method.

THEORETICAL FRAMEWORK

A life-course approach provides a framework for assessing factors associated with contraceptive consistency within adolescent sexual relationships. One primary life-course principle is that individual behavior can be understood only within the context of the relationships in which a person is involved (Bengston and Allen 1993; Elder 1998b). Therefore, we hypothesize that the changing dynamics of teens’ relationships and partners may influence continuity and change in contraceptive use patterns across relationships. In addition, a life-course perspective highlights the importance of the timing, sequencing, and duration of life events as well as their association with behavioral outcomes (Elder 1998a); we apply these concepts to adolescent sexual relationships. Specifically, we hypothesize that the timing of sexual experiences and the duration of sexual relationships are associated with reproductive health decisions, and that the sequencing of relationships are important predictors of contraceptive use and consistency. A life-course approach also emphasizes that individuals bring a history of experiences into their decision making. We propose that teens have underlying preferences for avoiding pregnancy or STIs that may, in turn, influence their contraceptive use and consistency across relationships. Over and above individual-level factors, we posit that characteristics of the relationship dyad, including characteristics of sexual partners, will be associated with contraceptive use within sexual relationships. We also hypothesize that teens may continue habits from their previous experiences, so that teens with experience practicing contraceptive consistency will be more likely to use contraceptives consistently in a current relationship.

PRIOR RESEARCH

Over the past decade, many studies have examined the association between relationship characteristics and contraceptive use or condom use, but often with nonrepresentative, small, or clinic-based samples and often with a focus on only a single dimension of sexual relationships, such as relationship type or length. We review this literature to identify six important relationship dimensions that may be associated with contraceptive use and consistency: (1) relationship homogamy (vs. relationship asymmetries), (2) self-defined relationship type, (3) exclusivity, (4) intimacy, (5) age of initiation, and (6) communication.

Relationship Homogamy Versus Heterogamy

Several studies have found a link between relationship asymmetries (with respect to age, race/ethnicity, and common social networks) and reduced contraceptive use and consistency, possibly because heterogamous relationships may be associated with a lower degree of comfort in communicating about sex and contraception. Among sexually active teens, having a much older partner is associated with reduced contraceptive use and consistency (Abma, Driscoll, and Moore 1998; Ford, Sohn, and Lepkowski 2001; Glei 1999; Manlove, Ryan, and Franzetta 2003), reduced condom use (DiClemente et al. 2002; Ford et al. 2001; Miller, Clark, and Moore 1997), and a greater likelihood of contracting sexually transmitted diseases (STDs) (Ford and Lepkowski 2004). Condom use is higher among teens whose partners were of a different race/ethnicity (Ford et al. 2001). However, among white teens, having a sexual partner who is a different race/ethnicity is associated with reduced contraceptive consistency (Manlove et al. 2006) and with increased STD infection (Ford and Lepkowski 2004). Contraceptive use is also lower, and STD incidence is higher, among high school teens who met their sexual partner outside of a common school setting (Ford and Lepkowski 2004; Ford et al. 2001).

Relationship Type

A large and consistent body of research indicates that condom use is more common with casual sexual partners than with more steady or serious partners (Cooper and Orcutt 2000; Ku et al. 1998; Noar, Zimmerman, and Atwood 2004; Santelli et al. 1996; Sheeran, Abraham, and Orbell 1999; Stark et al. 1998), most likely because casual partners are seen as presenting a greater risk for contracting STIs (Ellen et al. 2002; Rosengard, Adler et al. 2005). Much of this work, however, has focused on small, geographically restricted, nonrepresentative samples of adolescents and young adults (Corbin and Fromme 2002; Rosengard, Clarke, et al. 2005; Scholes et al. 2003; Sturdevant et al. 2001) or high-risk populations, such as STD clinic attendees (Fortenberry et al. 2002; Katz et al. 2000; Lansky, Thomas, and Earp 1998; Macaluso et al. 2000; Rosengard, Aler et al. 2005) or drug users (Williams et al. 2001).

Findings from studies linking relationship type to contraceptive use for pregnancy prevention, however, have been mixed. For example, teens who are “going steady” with their sexual partner at the time of first intercourse are more likely to use contraceptives when first having sex (Manning, Longmore, and Giordano 2000), but there is no association between “going steady” and contraceptive consistency in teens’ first relationships (Manlove and Terry-Humen 2007). One study showed a positive link between having a romantic sexual relationship and ever using contraception in that relationship (Ford et al. 2001), but others showed reduced consistency in romantic compared with “liked” last relationships¹ (Manlove, Ryan, and Franzetta 2004) or nonromantic (Manlove et al. 2006) last relationships. The differences across these studies may be attributable to sample differences or to other relationship measures used in models. Determining nuances in the meaning of self-defined relationship type may be particularly difficult, especially because the majority of girls report that their sexual relationships are romantic (Carver et al. 2003; Ford et al. 2001; Ryan, Manlove, and Franzetta 2003).

Exclusivity

Research linking nonmonogamous relationships and teen contraceptive use has shown mixed findings. Some research has found an association between concurrent or non monogamous relationships and reduced contraceptive use (Weisman et al. 1991) and condom use (Kelley et al. 2003). Other work has suggested that among those in concurrent relationships, the odds of condom use and condom use intentions are increased when having sex with a secondary, rather than the main, partner (Lansky et al. 1998; Santelli et al. 1996).

Intimacy

A higher level of intimacy between sexual partners may also be associated with greater contraceptive use and consistency, perhaps because of the greater predictability of having sex. For example, one study found that higher levels of intimate activities between sexual partners and connections to social networks are associated with greater contraceptive use and consistency among male teens (Manlove et al. 2004).

Age of Initiation

The younger teens are at the time of sexual initiation with their partners, the less likely they are to use contraceptives (Glei 1999; Manning et al. 2000). However, some research indicates that an older age is negatively associated with condom use (DiClemente et al. 1996; Ku, Sonenstein and Pleck 1994), perhaps because older couples are more likely to be in more committed relationships or relationships in which the woman is using hormonal contraceptives (Ku et al. 1994).

Communication

Romantic and sexual relationships, especially new relationships, may involve an amount of communicative awkwardness (Giordano et al. 2006). However, in order to maintain positive reproductive health, it is important for teens to communicate about issues related to sex and contraception with their partners (Kirby 2001). Recent research has shown a positive association between discussing contraception before first having sex with a partner and actual contraceptive use and consistency (Manlove et al. 2003, 2004; Stone and Ingham 2002). Discussions between sexual partners about condoms and sexual histories are consistently associated with improved condom use (Noar, Carlyle, and Cole 2006).

Relationship History

Net of current relationship and partner characteristics, we hypothesize that important characteristics of teens’ previous sexual partners and relationships will be associated with contraceptive consistency in their current relationship. For example, we hypothesize that their history of contraceptive use with previous partners will be related to their current use, as has been found in previous research (Ku et al. 1994; Manlove et al. 2006; Manlove et al. 2004; Sheeran et al. 1999).

Family and Individual Controls

Family and individual characteristics associated with improved contraceptive use and consistency include higher family socioeconomic status, living with both biological parents, higher cognitive ability, and (in some cases) having received sex education (Afxentiou and Hawley 1997; Brindis, Pagliaro, and Davis 2000; Fisher 2004; Kirby 2001; Manlove et al. 2003; Manning et al. 2000; Miller 2002). Alternatively, racial and ethnic minorities, especially Hispanic teens, often show reduced condom and contraceptive use and consistency (Brindis et al. 2000; Ford et al. 2001; Kirby 2001; Ku et al. 1994), although black teens have been shown to report more frequent condom use than white or Hispanic teens (Everett et al. 2000; Ku et al. 1994).

HYPOTHESES

Based on the literature review, we specify four hypotheses. First, adolescent contraceptive use and consistency will vary across relationships for the same person. We expect that a substantial proportion of teens with two or more relationships will have differing levels of contraceptive consistency in those relationships. For example, they may be consistent users in one relationship, yet never use contraceptives—or use them inconsistently—in another.

Second, partner and relationship characteristics will be associated with contraceptive use and consistency. Teens who are older, in more homogamous relationships, in romantic relationships, in monogamous relationships, in relationships that involve higher levels of intimacy, and in relationships that involve communication about contraception will have higher levels of contraceptive use and consistency.

Third, adolescent relationship and contraceptive use histories will be associated with contraceptive use and consistency in the current relationship. Teens with a history of consistent contraceptive use, as well as those with a history of communicating with their sexual partners about contraceptive use, will have more-positive contraceptive use outcomes than other teens. Alternatively, teens with a history of risky sexual partners (including those with a history of short-term and/or nonromantic relationships) and teens with several previous sexual relationships will have more-negative contraceptive use outcomes.

Fourth, relationship and partner factors will influence teens using hormonal and coitus-dependent methods; however, we hypothesize that there will be less of an influence of relationship and partner characteristics on contraceptive use and consistency among teen females who have used hormonal methods because negotiation and communication skills may be less important for teens who are not using coitus-dependent methods.

DATA AND METHODS

Data and Sample

This study used data from the National Longitudinal Study of Adolescent Health (Add Health), a nationally representative school-based survey of U.S. students in grades 7–12 in 1995 (Harris et al. 2003). Add Health data collection included three waves of in-home interviews, in 1995 (Wave I), 1996 (Wave II), and 2002 (Wave III). For this study, we used data from teens who participated in both Wave I and Wave II in-home interviews (n = 14,736). Information on participants’ sexual relationships and contraceptive use was drawn from both survey waves, and family background and individual characteristics came from the Wave I survey. We created a rich relationship-level file that includes one observation for every sexual relationship that teens reported in the survey, ordered from the earliest to the most recent partner.

We drew our sample from 4,556 unmarried, sexually experienced adolescents who participated in both survey waves and had valid sample weights and partner-specific information about sexual relationships.² In order to determine the sequential order of relationships, we resolved missing and incomplete partner-specific dates of first intercourse. We eliminated 152 teens from our sample because they had missing or incomplete dates of first intercourse with a particular partner for two or more of the sexual partners reported in their partner history. For 321 respondents who had only one missing or incomplete date, we dropped the partner with the missing/incomplete date of the first sexual encounter with the teen but kept the rest of the sexual partners in our file. We also identified and dropped duplicate partners reported by a respondent³; although these deletions reduced the number of sexual partners in our data file, they did not reduce the number of teens in our sample. Finally, we eliminated relationships from our file that were missing on the dependent variables of contraceptive use and consistency. This caused 21 teens to drop out of our sample because they had missing information on these measures for all the relationships reported in their partner history. Our final sample consisted of 4,383 teens who reported a total of 9,668 sexual relationships.

Dependent Variables

We derived our dependent variables of contraceptive use and consistency from two questions about teens’ contraceptive use with their sexual partners: “Did you or [your partner] ever use any method of birth control?” and “Did one or the other of you use some method of birth control every time you and [your partner] had sexual intercourse?” From these questions, we constructed three dichotomous dependent variables: (1) ever used contraception within the relationship versus never used contraception at any point during the relationship; (2) always used contraception within the relationship versus only sometimes and/or never used a contraceptive method; and (3) among a sample of contraceptive users, always used contraceptives versus only sometimes used contraceptives. In descriptive analyses, we also included a measure of the number of different types of contraceptive consistency (never, sometimes, or always) reported in previous relationships.

Within-Individual Measures

The key measures of interest in our analyses are within-individual factors—that is, relationship-specific factors that vary across relationships for a given adolescent. These within-individual relationship-level factors are described in detail here. We also include an analytic control variable for the duration, in months, between the first and last sexual encounter with each partner.

Relationship homogamy.

For each relationship, teen-partner homogamy was measured with a three-item summative index assessing how similar teens and their partners were. Teens received one point each if (1) their partner was no more than two years older or younger than they; (2) they and their partner were the same race/ethnicity; and (3) they met in the teen’s typical social network (in school, church, or neighborhood) or they were friends at the start of the relationship (vs. a friend of a friend, a stranger, or other).

Self-defined relationship type.

Relationship type compares self-identified romantic relationships,⁴ with “liked” and nonromantic relationships. As we explained earlier, in the Add Health survey, a “liked” relationship was one that teens did not self-define as romantic but in which the respondents had held hands with and kissed their partner and had told their partner they liked or loved him or her. Nonromantic relationships were those categorized as neither romantic nor liked.

Exclusivity.

We operationalized exclusivity of the relationship with a measure of whether teens engaged in a nonmonogamous relationship. In other words, did the teen report at least one other concurrent sexual relationship? We used a conservative measure of nonmonogamous relationships: cases in which one relationship ended and the next one began in the same month were coded as monogamous, although we could not be sure whether such relationships were successive or overlapping. Because we cannot measure whether the teen’s partner had any concurrent sexual relationships, this measure is limited to the perceived exclusivity of the relationship based on only the fidelity of the focal teen.

Intimacy.

Intimacy was assessed in romantic or liked relationships with an eight-item index of the number of couple-like activities that preceded sex in the relationship. The couple-like activities were thinking of themselves as a couple, telling others they were a couple, going out together (alone or in a group), exchanging “I love you’s,” meeting each other’s parents, exchanging presents, and spending less time with friends in order to spend more time together.

Age of initiation.

We measured the teen’s age at the beginning of each relationship.

Communication.

We measured communication for romantic and liked relationships by using a dichotomous variable that indicated whether the teen and his/her partner had discussed contraception before having sex for the first time.

Relationship history.

We measured the number of previously reported sexual partners and the length of time between the end of one sexual relationship and the beginning of the next one. Because of our interest in whether contraceptive consistency in early relationships influences current contraceptive use, we also measured the proportion of previous relationships in which the teen reported always using contraception. In addition, we measured whether the teen reported using a hormonal contraceptive method in any previous relationship, the proportion of previous relationships that were risky (characterized as either nonromantic or a one-time sexual encounter), and the proportion of previous romantic or liked relationships in which the teen and partner discussed contraception before having first sex.

Between-Individual Measures

Our analyses also include several measures of between-individual factors. These are control factors that remain constant for each adolescent across multiple relationships. We controlled for family structure, parental education (1 = less than a high school graduate, to 7 = at least some graduate or professional school), race/ethnicity, cognitive ability (measured by the teen’s score on a modified Peabody Picture Vocabulary Test [PVT], on which the national average is 100 [Dunn and Dunn 1981]), whether the respondent had received pregnancy and AIDS prevention education in school, and age at the first sexual encounter with the first reported partner. For analytic reasons, we also include an indicator of whether the relationship was the teen’s first relationship in our ordered relationship-level file.⁵

Methods

Our relationship-level analysis file, consisting of a separate record for each sexual relationship teens reported, allowed us to examine multiple sexual relationships, in sequential order, for adolescents who had more than one relationship (up to 10 partners). In the first stage of analysis, we used descriptive statistics to assess patterns of contraceptive use and consistency across teens’ sexual relationships (within-individual factors). Characteristics for comparison also included individual and family background factors (between-individual factors). In the second stage, we conducted bivariate analyses, using t tests for continuous variables and chi-squares for categorical variables, to examine associations between (1) current and previous relationship and partner characteristics and (2) contraceptive use and consistency.

In the third stage, we ran multivariate analyses to examine the effects of current partner and relationship characteristics and previous sexual history characteristics (within-individual factors) on contraceptive use and consistency in the current relationship, controlling for individual and family background factors (between-individual factors). For the contraceptive use outcome, we compared teens who ever used a contraceptive method with those who never did. For the contraceptive consistency outcome, we modeled two contrasts: (1) always versus sometimes or never used contraception; and (2) always versus sometimes used contraception, for a subsample of respondents who ever used contraception. This approach allows us to examine consistency among all teens as well as among only those teens who ever used contraception.

Past and current contraceptive use may be correlated because both behaviors are determined by respondent-specific unobserved characteristics, so we controlled for unobserved respondent-specific heterogeneity. Without heterogeneity, we might have overstated the effects of past contraceptive use on current contraceptive use. Our models are longitudinal random-effects logistic regression models:

$$\frac{p}{1-p}=e^{{\beta }_0}{e}^{{\beta }_1{x}_1}{e}^{{\beta }_2{x}_2}\cdots e^{{\beta }_k{x}_k}{e}^{\delta },$$

where δ represents the heterogeneity residual. The multivariate tables present the multiplicative effect of each covariate on the odds ratio (i.e., e^β0, e^β1, and so on). These models are appropriate for data with repeated outcomes and control for respondent-specific unobserved heterogeneity by linking sexual relationships to each teen. The respondent-specific residual term, δ, affects the odds of contraceptive use and consistency in all sexual relationships and is assumed to be normally distributed. The models produce a rho value indicating whether teen-specific unobserved heterogeneity affects contraceptive consistency. We ran these models by using the xtlogit command in Stata, which allows us to incorporate weights and to control for having multiple observations per respondent.⁶

Finally, in the fourth stage of analysis, we examined interactions between partner/relationship characteristics and whether female teens used any hormonal contraceptive method in a previous relationship. These interactions were conducted to test our hypothesis that the influence of partner and relationship characteristics would be weaker for females who had a history of using hormonal contraception because they do not need to negotiate contraceptive use each time they engage in sexual intercourse with a partner.

RESULTS

Descriptive Analyses

Characteristics of relationships and respondents.

Table 1 presents data on contraceptive use in sexual relationships, for all relationships aggregated together and by the total number of relationships each teen had. Considering all relationships (column 1), teens used contraceptives at least once in 76% of their sexual relationships, leaving one-quarter of relationships in which teens never used any contraceptives. Teens always used contraceptives when they had sex in over half (58%) of their relationships, and they used contraceptives inconsistently in 17% of relationships. Looking across relationships, teens with a greater number of sexual relationships were more likely to never use contraceptives and less likely to always use contraceptives. For example, among teens with one relationship, 22% never used contraception and 62% always used contraception, compared with 29% and 53%, respectively, for teens with five or more relationships.

Table 1.

Contraceptive Use in Teens’ Sexual Relationships, by the Number of Sexual Relationships They Had

	All Relationships	Number of Relationships
		One	Two	Three	Four	Five or More
Contraceptive Consistency (%)
Never used contraceptives	24.4	21.8	21.2	24.8	25.5	29.2
Sometimes used contraceptives	17.3	16.0	18.8	17.8	15.4	17.6
Always used contraceptives	58.3	62.2	60.0	57.4	59.1	53.2
Ever used contraceptives	75.6	78.2	78.8	75.2	74.5	70.8
Number of Relationships	9,668	1,749	2,572	1,914	1,496	1,937
Number of Respondents	4,383	1,749	1,286	638	374	336

Note: A full table showing all predictor variables and tables of bivariate analyses are available from the authors upon request.

Teens reported up to three different contraceptive use outcomes across relationships (never, sometimes, or always using contraception), and our analyses indicated that teens varied their contraceptive practices across their sexual relationships. For example, more than one-half (56%) of the 2,634 teens with two or more sexual relationships reported two or more types of contraceptive consistency (never, sometimes, or always; analyses not shown here). Among the 1,348 teens with three or more sexual relationships, 21% reported all three contraceptive use outcomes (analyses not shown here). Teens who had a greater number of relationships were more likely to have reported all three levels of contraceptive consistency.

Contraceptive use and consistency by relationship parity.

Table 2 shows information on contraceptive use and consistency by parity of the relationship. For teens who had more than one sexual relationship, we found an overall, although nonsignificant, pattern of improvement in contraceptive use and consistency across relationships. For example, for teens who had engaged in a total of four relationships, 72% were contraceptive users in their first relationship. By the fourth relationship, 79% had ever used contraception within the relationship. The pattern of change across relationships in contraceptive consistency was similar. This pattern may be due to the facts that the latter relationships tend to be shorter relationships and that being consistent contraceptive users in shorter relationships is easier. Note, also, that consistent contraceptive use in the first reported relationship declined as the total number of partners increased, suggesting that the number of sexual partners and contraceptive consistency are endogenous.

Table 2.

Comparison of Teens’ Contraceptive Use and Consistency, by Parity of the Relationship

	Parity of the Relationship					N
	First	Second	Third	Fourth	Fifth
Ever Used Contraceptives (%)
Had one sexual relationship	78.2	—	—	—	—	1,749
Had rwo sexual relationships	78.4	79.2	—	—	—	1,286
Had three sexual relationships	73.7	74.7	77.3	—	—	638
Had four sexual relationships	72.2	72.9	73.8	79.0	—	374
Had five or more sexual relationships	72.3	71.2	72.6	67.6	72.2	336
Always Used Contraceptives (%)
Had one sexual relationship	62.2	—	—	—	—	1,749
Had two sexual relationships	59.8	60.1	—	—	—	1,286
Had three sexual relationships	56.3	56.8	59.2	—	—	638
Had four sexual relationships	53.3	56.5	60.5	65.6	—	374
Had five or more sexual relationships	51.8	52.2	54.9	54.3	54.7	336
Sometimes Used Contraceptives (%)
Had one sexual relationship	16.0	—	—	—	—	1,749
Had two sexual relationships	18.6	19.0	—	—	—	1,286
Had three sexual relationships	17.4	17.9	18.1	—	—	638
Had four sexual relationships	18.3	16.3	13.3	13.4	—	374
Had five or more sexual relationships	20.5	19.0	17.7	13.3	17.5	336

Multivariate Analyses

Table 3 presents odds ratios from the random-effects logit models for the full sample, and Table 4 shows findings for a subsample of teens in romantic or liked relationships. All models in Tables 3 and 4 control for unobserved heterogeneity, and we present the parameters for the model fit at the bottom of each table. The respondent-specific residual (measured by the likelihood ratio test of rho) is significant in all models (although marginal for the final model in Table 4), indicating that our estimates would be biased without accounting for the unobservable variables.

Table 3.

Odds Ratios From Random-Effects Logit Models Predicting Whether Teens Ever or Always Used Contraceptives in Their Sexual Relationships: Full Sample

Variable	Ever Used (vs. never used)		Always Used (vs. never or sometimes used)		Always Used (vs. sometimes used)
	Females	Males	Females	Males	Females	Males
Within-Individual Factors
Teen-partner homogamy
Teen-partner homogamy scale	1.09	1.06	1.10*	1.05	1.11*	1.04
Relationship type (ref. = romantic)
Liked	0.66***	0.88	1.25*	1.09	2.37***	1.24
Nonromantic	0.42***	0.48***	1.23*	1.14	11.82***	6.68***
Exclusivity
Nonmonogamous relationship	1.16	1.22	1.05	1.06	0.90	0.88
Age at initiation
Age at first intercourse with partner	1.12	1.22**	1.11*	1.07	1.09	0.86*
Sexual history
Number of previous sexual partners	1.00	0.96	0.95	0.94	0.90	0.90
Length of time since last relationship ended	1.02	1.02	1.01	1.01	1.00	1.00
% of previous relationships in which respondent always used contraceptives	2.55***	3.05***	2.71***	2.37**	2.69***	2.39***
Any previous relationship in which respondent used hormonal contraceptives	1.33*	0.92	1.74***	1.05	2.08***	1.18
% of previous relationships that were “risky”	0.77	1.38	0.97	1.11	1.29	0.86
Length of sexual relationship	1.05***	1.02**	1.00	0.99**	0.97***	0.96***
Between-Individual Factors
Two biological/adoptive parents	1.01	1.21	1.02	1.35**	1.03	1.40**
Parent education	1.07*	1.02	1.04	1.01	0.99	0.99
Race/ethnicity (ref. = white)
Black	1.25	1.39*	1.19	1.53**	1.06	1.46*
Hispanic	0.83	0.72*	1.07	0.83	1.50*	1.21
Other race/ethnicity	0.91	0.56**	0.92	0.74	0.89	1.26
PVT score	1.03***	1.02***	1.01***	1.01***	1.00	1.00
Had pregnancy and AIDS education	0.86	0.94	0.89	1.18	0.92	1.54**
Age at first intercourse with first reported partner	0.93	0.91	0.94	0.94	0.98	1.02
rho	0.17	0.19	0.11	0.18	0.09	0.12
Likelihood Ratio Test of rho	26.82***	28.14***	4.89*	10.87***	3.43*	4.56*
Wald Chi-Square	386.8***	229.8***	176.91***	119.84***	329.31***	199.02***
Likelihood Ratio	−2,707.85	−2,148.99	−3,469.60	−2,685.71	−2,073.43	−1,403.74
df	25	25	25	25	25	25
N	5,344	4,324	5,344	4,324	3,980	3,335

Note: These models also controlled for whether the respondent had partners not captured in the sexual history, sequentially ordering the respondent’s relationships required making assumptions about duplicate dates, the number of nonromantic partners was potentially overestimated, the relationship was reported at Wave 1, and the relationship was the teen’s first one in the sequential file.

^*p < .05;

^**p < .01;

^***p < .001

Table 4.

Odds Ratios From Random-Effects Logit Models Predicting Whether Teens Ever or Always Used Contraceptives in Their Sexual Relationships: Romantic Sample

Variable	Ever Used (vs. never used)		Always Used (vs. never or sometimes used)		Always Used (vs. sometimes used)
	Females	Males	Females	Males	Females	Males
Within-Individual Factors
Intimacy: Couple-like activities	1.06**	1.12***	1.02	1.05**	0.99	0.97
Communication: Discussed contraception before sex	2.29***	1.24	1.59***	1.17	1.16	1.06
Sexual history: % of previous relationships in which contraception was discussed before sex	1.15	1.05	1.21	1.25	1.30	1.33
rho	0.21	0.15	0.09	0.13	0.19	0.03
Likelihood Ratio Test of rho	27.36***	11.48***	3.03*	7.27**	12.20***	2.19†
Waid Chi-Square	352.57***	216.21***	302.11***	161.55***	217.44***	150.30***
Likelihood Ratio	−2,118.41	−1,582.07	−2,910.86	−2,111.11	−1,993.58	−1,304.68
df	26	26	26	26	26	26
N	4,553	3,387	4,553	3,387	3,521	2,686

Note: These models also controlled for all measures included in Table 3.

^†p < .10;

^*p < .05;

^**p < .01;

^***p < .001

Full sample, ever use versus never use.

For females, teens in liked or nonromantic relationships had reduced odds of ever using contraceptives (odds ratios, 0.66 and 0.42, respectively) compared with those in romantic relationships (see Table 3). For males, only those in nonromantic relationships had reduced odds of contraceptive use. The older teens were when they initiated sex with their partner (males only) or the longer a sexual relationship lasted (both males and females), the greater the odds were that they used contraceptives.

One aspect of teens’ sexual history also was related to contraceptive use for both females and males: previous contraceptive consistency. Even after we controlled for unobserved heterogeneity, prior contraceptive use had a positive effect on current contraceptive use. Specifically, the greater the proportion of previous relationships in which teens had been consistent contraceptive users, the higher the odds were of using contraception at least once in their current relationship. Gender differences existed in the effects of previous hormonal contraceptive use. Among females only, having used a hormonal contraceptive method in any previous sexual relationship was associated with 33% greater odds of contraceptive use in the current relationship. The longer a sexual relationship lasted, the greater the odds were that males and females used contraceptives. Controls operated in the expected direction.

Model 1 had a significant rho value of 0.17 for females and 0.19 for males, indicating that 17% and 19% of the total variance in the models for females and males, respectively, was contributed by between-individual factors that are constant for each teen across their relationships (e.g., race/ethnicity). The balance of the variance (83% for females and 81% for males) was explained by within-individual relationship-level factors and sexual history factors that vary across a particular teen’s relationships.

Full sample, consistent use versus inconsistent or no use.

Females who selected partners who were more similar to themselves had higher odds of always using contraceptives than those who selected partners scoring lower on the homogamy index. This association was driven primarily by the partner age difference component of the index. Being in a liked or nonromantic relationship, as compared with being in a romantic relationship, was related to approximately 20% higher odds of female teens always using contraceptives. A one-year increase in the age at first intercourse with the partner was related to 11% higher odds of always using contraceptives for females. The findings for relationship type among females and duration among males are opposite to the ever-use models in which liked and nonromantic relationships were linked to reduced odds, and relationship length was associated with greater odds, of contraceptive use.

For teens of both genders, the greater the proportion of previous relationships in which teens always used contraceptives, the greater were the odds of being a consistent contraceptive user in their current relationship. For females, using hormonal contraceptive methods in any previous relationship was associated with 74% higher odds of always using contraceptives in their current relationship. For males, relationship duration was a significant predictor of contraceptive consistency. Significant controls operated in the expected direction. Rho values indicate that the majority of the model variance (89% for females and 82% for males) was explained by within-individual relationship-level factors.

Full sample, consistent use versus inconsistent use.

Results of models predicting consistent use versus inconsistent use among teens who ever used contraceptives with their partner are similar to findings from the models predicting consistent use versus sometimes or never use; however, a few variables operate differently in the consistent versus inconsistent use models. Nonromantic relationship type and having had pregnancy/AIDS education (for males only) and being Hispanic (for females only) were associated with greater odds of consistent use versus inconsistent use, but these factors were not significant predictors of always use versus sometimes or never use. Age at the time of the first sexual encounter with partner (for males) and length of sexual relationship (for females) were associated with reduced odds of contraceptive consistency versus inconsistency; these measures were not significant in the consistency versus inconsistency/non-use models. Two factors lost significance in the always-use versus sometimes-use models (compared with the always-use versus sometimes-/never-use models): age at first intercourse with partner (females only), and PVT score. The rho values indicate that approximately 90% of the variance in each model is explained by within-individual relationship-level factors.

Romantic sample, ever use versus never use.

Table 4 displays the subset of variables that were measured for the romantic/liked sample only. For each additional couple-like activity that teens engaged in with their partners before having sex for the first time, the odds of ever using contraceptives were 6% greater for females and 12% greater for males. Female teens who discussed contraception with their partners before first having sex had more than twice the odds of using contraceptives. The parameters of model fit across columns at the bottom of Table 4 indicate that the majority of model variance (79%–97% across models) was explained by within-individual relationship-level factors and relationship history factors.

Romantic sample, consistent use versus inconsistent or no use.

For the models predicting consistent contraceptive use (versus sometimes or never use), discussing contraception was significant for females, and couple-like activities was significant for males. Both measures operated in the same direction as they did in the ever-use models.

Romantic sample, consistent use versus inconsistent use.

Among the sample of teens who ever used contraception with their partner, none of the variables that were measured only for the romantic/liked sample were significantly associated with consistent contraceptive use.

Interactions between relationship characteristics and previous hormonal contraceptive use.

Table 5 displays significant interactions between current relationship characteristics and previous hormonal method use, for females only. For teen females who used a hormonal contraceptive method in any previous relationship, the age when having sex for the first time with one’s partner was not a significant predictor in either model. However, for those who had never previously used a hormonal method, being older at the first sexual encounter with one’s partner was associated with greater odds of ever using contraceptives and using them consistently (odds ratios were 1.16 and 1.15, respectively).

Table 5.

Odds Ratios From Random-Effects Logit Models Showing Significant Interactions by Whether Female Teens Previously Used Any Hormonal Method

Variable	Ever Used vs. never used)	Always Used (vs. never or sometimes used)
Age at First Intercourse With Partner
Previous hormonal use × Age at first intercourse with partner	0.83*	0.87*
Age at first intercourse for teens who ever used hormonal method	0.96	1.00
Age at first intercourse for teens who never used hormonal method	1.16**	1.15*
Couple-like Activities Before First
Having Sex (romantic sample only)
Previous hormonal use × Couple-like activities	—	0.92*
Couple-like activities for teens who ever used hormonal method	—	0.95
Couple-like activities for teens who never used hormonal method	—	1.04*
Discussed Contraception With Partner Before
First Having Sex (romantic sample only)
Previous hormonal use × Discussions about contraception	0.51*	—
Discussion about contraception for teens who ever used hormonal method	1.27	—
Discussion about contraception for teens who never used hormonal method	2.49***	—

Note: The models with the “age at first intercourse” interaction also controlled for all measures included in Table 3; the models with the other two interactions controlled for all measures included in Table 4.

^*p < .05;

^**p < .01;

^***p < .001

Engaging in a greater number of couple-like activities with one’s partner before first having sex was associated with greater odds of consistent contraceptive use for females without prior hormonal use but was not significant for those who had ever used hormonal contraception. Similarly, discussing contraception with one’s partner before first having sex was associated with more than twice the odds of ever using contraceptives for females who had never previously used hormonal contraceptives but was not significant for those with a history of hormonal contraceptive use. In sum, these interactions suggest that the effects of relationship characteristics are weaker for females who have a history of hormonal contraceptive use. There were no significant interactions in the analyses of consistent use versus inconsistent use.

DISCUSSION

The analyses presented in this paper provide some support for all of our hypotheses. We found that teens vary their contraceptive use patterns across their relationships, and the hypothesized relationship dimensions (with the exception of exclusivity) were associated with contraceptive use and/or consistency in at least some of the models. Within- individual relationship-level factors accounted for the vast majority of the variance in our models. Relationship-level factors associated with greater odds of contraceptive use and/or consistency include having a more homogamous partner, having a more intimate relationship, and communicating about contraception before sexual intercourse with a specific partner.

Two factors—age at first sex with a partner and the romantic relationship type—showed an opposite association with ever using contraception compared with always using contraception. The positive association between age (and having a romantic partner) and ever using contraceptives matches other research findings (Ford et al. 2001; Manning et al. 2000), suggesting that older teens and those in a romantic relationship at the time of sexual initiation with a partner may be more comfortable negotiating contraception with that partner. The negative association between an older age (and a romantic partner) and contraceptive consistency matches recent research on this topic (Manlove et al. 2006; Man-love et al. 2004), and is also in line with more extensive research on condom use (Corbin and Fromme 2002; DiClemente et al. 1996; Fortenberry et al. 2002; Ku et al. 1994; Noar et al. 2004; Sheeran et al. 1999), suggesting that relationships to older teens and romantic relationships may represent more-committed relationships in which teens may be more willing to occasionally skip using contraceptives or may even want to get pregnant.

The positive association between partner homogamy and contraceptive consistency suggests that teen females who don’t know their partner through their usual social networks, who have a partner of a different race/ethnicity, or who have an older sexual partner may be less able to negotiate consistent contraceptive use. This finding was driven by partner age difference, reinforcing other research showing that sexual relationships between females and older partners can result in contraceptive inconsistency (Abma et al. 1998; Ford et al. 2001; Glei 1999; Manlove et al. 2003) and reduced condom use (DiClemente et al. 1996). Teens who engaged in more intimate, couple-like activities with their partners had greater odds of ever and always using contraceptives with that partner. Both greater intimacy and greater partner homogamy indicate a higher level of familiarity or potential comfort with a partner at the time of first sex, which may be linked to a better ability to effectively communicate and negotiate contraceptive use needs.

We found a strong positive link among females between discussing contraception with a sexual partner before first having sex and ever and always using contraceptives within that relationship. This finding confirms other research on contraception (Manlove et al. 2003, 2004; Stone and Ingham 2002) and condom use (Bowleg, Lucas, and Tschann 2004), and extends research suggesting that females who are more comfortable talking with males, in general, are more likely to use contraceptives (Stone and Ingham 2002). Discussions between sexual partners may be especially important for HIV/AIDS prevention, especially those conversations that help assess partner sexual history and focus on condom use (Noar et al. 2006; Whitaker et al. 1999). Discussions about contraception may reflect individual and/or partner motivations to prevent pregnancy. In analyses of rigorously evaluated pregnancy prevention programs, Kirby (2001) found that effective programs provide teens with opportunities to practice communication and negotiation skills as a way to improve contraceptive use. Thus, teens who have experience practicing how to say “no” to unprotected sex and negotiating contraceptive use with their partners may be better prepared to avoid unintended pregnancy. This study’s findings suggest that it may also be helpful for programs to have teens role-play negotiating contraception with multiple types of partners, such as with a partner whom a teen doesn’t know well, an older sexual partner, or a partner who doesn’t find contraceptive use to be “romantic.”

We hypothesized that consistency of contraceptive use in previous relationships would be associated with consistency in the current relationship. We found that models that controlled for unobserved heterogeneity improved the data fit. However, even after controlling for respondent-specific unobservable variables, we found a remaining effect of previous consistency on contraceptive use outcomes for both females and males. This finding suggests that teens with experience using contraceptives consistently (on their own initiation, in combination with a partner, or on the partner’s initiation) were better able to maintain consistency in subsequent relationships. Alternatively, teens with inconsistent patterns of contraceptive use may face a greater risk of unintended pregnancy in future relationships.

Even after we controlled for unobserved heterogeneity, the models showed a positive association between hormonal method history and current contraceptive use and consistency for female teens. These findings suggest that hormonal method users are more consistent users of contraceptives, possibly because female hormonal method users do not need to negotiate a condom or other coitus-dependent method every time they engage in intercourse with partners (some of whom may be unwilling to use a condom) (Pleck, Sonenstein, and Ku 1991). Alternatively, some of the association between hormonal method history and current contraceptive practices may be endogenous. Specifically, female teens who choose hormonal methods may also be risk averse and more likely to use contraceptives consistently.

As we hypothesized, several within-individual relationship factors were not associated with contraceptive use among females who used hormonal methods. For example, we found that communication for teen females was not significantly associated with ever using contraceptives for hormonal method users. Interactions between hormonal method history and couple-like activities also suggest that an overall nonsignificant association between couple-like activities and always using contraceptives among females in the main effects model masked a significant association for those who had not used hormonal methods in previous relationships. Finally, the reduced odds of contraceptive consistency for younger teens seen in the main effects model was not significant for teen females who had used hormonal methods in a previous relationship, suggesting that hormonal methods may be especially protective for younger teens.

We did not find interactions with previous hormonal use for three measures that were significant in our main effects models: partner homogamy, self-defined relationship type, or relationship duration. The absence of interactions indicates that regardless of whether teens used or did not use hormonal methods with previous partners, they are still influenced by relationship and partner factors in their contraceptive decisions. In addition, the vast majority of sexual relationships involving a contraceptive method, including the majority that listed a hormonal method, also involved nonconcurrent coitus-specific methods. Unless teens use long-acting hormonal methods (such as injectables or a contraceptive patch), they will continue to need to negotiate contraceptive use with partners or to remember to use contraceptives regularly.

Our analyses have some limitations, including the fact that teens provided retrospective information on partner characteristics. However, the maximum amount of recall required was 18 months, which limits recall bias. Also, Add Health incorporated audio computer-assisted self-interviews to improve the validity of reports of risky or sensitive behaviors (Turner et al. 1998). In addition, the measure of nonmonogamous relationship is based on the respondent’s report only; ideally, we would also have information from the respondent’s sexual partner. We also recognize that despite controls for unobserved heterogeneity, some of the associations between relationship and partner characteristics and contraceptive use and consistency may be endogenous. Also, we anticipated that males would be less-accurate reporters of relationship-level contraceptive use, especially of hormonal method use, than females. We ran additional analyses excluding relationships in which hormonal methods were the only reported method and found similar results (analyses not shown here). Thus, we are confident that the associations between relationship and partner factors and contraceptive use outcomes are robust for males. Finally, because of the Add Health data structure, we were unable to differentiate consistent hormonal method use from consistent condom use. However, our findings match those reported in much of the literature on condom use, even the work that used small, non-representative samples.

Despite these limitations, we were able to extend previous research by showing that many teens change their contraceptive use patterns across sexual relationships, and that several within-individual relationship and partner characteristics are of critical importance to adolescent contraceptive use decisions, even among females who have used hormonal methods. These findings suggest that models of contraceptive use that focus on between- individual factors (i.e., family and individual characteristics) ignore critical within- individual dimensions of teen relationships that influence their reproductive health decisions and risk of unintended pregnancy.