Racial and Ethnic Differences in the Relationship between Risk-Taking and the Effectiveness of Contraceptive Methods Used by Adolescents

Abstract

Context.

Little is known about whether risk-proneness, understood as the person-specific risk-taking beyond the realm of sex, is associated with the effectiveness of contraceptive methods used by adolescents. Furthermore, previous studies have not assessed whether such relationship varies by race-ethnicity.

Methods.

Data from the 2011, 2013 and 2015 National Youth Risk Behavior Surveys were used to examine non-sexual risky behaviors and contraceptive method choice of 5,971 sexually active females aged 13–18. Latent class analyses were used to identify risk-taking profiles for white, black, and Hispanic female adolescents. Multinomial logistic regressions were used to estimate the association between these risk profiles and use of less and more effective contraceptive methods in the last sexual intercourse.

Results.

There were three risk-taking profiles for white and Hispanic adolescents, and two for black adolescents. Higher risk-taking among white adolescents was associated with greater use of condom alone and condom paired with a prescription contraceptive, relative to not using contraception. But higher risk-taking among whites was also associated with greater use of prescription contraceptives relative to condoms. For black and Hispanic females, lower risk-taking was only associated with more condom use. Even those black and Hispanic girls who are low risk-takers in other dimensions relied mostly on less effective contraceptive methods.

Conclusions.

Interventions designed to reduce overall adolescent risk-taking may also reduce unintended pregnancy and sexually transmitted infection rates, particularly among white, female adolescents. Policy measures intended to increase use of more effective contraception among black and Hispanic adolescents should target barriers other than risk-proneness.

Introduction

Social scientists have long been interested in the determinants of unintended and adolescent pregnancy, and multiple hypotheses have been put forth to explain why sexually active adolescents use less effective methods, or do not use contraception at all. Previous studies have assessed the role of external barriers, such as access to contraceptive services,¹⁶ and internal barriers, such as motivations to avoid pregnancy,²² dissatisfaction with contraceptive methods,³¹ knowledge and attitudes about contraception,⁴⁵ and self-efficacy.²¹ However, personal inclinations for risk-taking have received surprisingly little attention as a possible internal barrier to contraceptive use. In particular, the relationship between risk profiles and the probability of using more effective contraceptive methods is not well understood. By risk profile, we mean the person-specific propensity to engage in multiple risk-taking behaviors that are not limited to the realms of sexual activity and contraceptive use. To the extent that using less effective contraceptive methods, or no contraception at all, involves greater risk-taking, heterogeneity in risk-profiles may contribute to explaining the different contraceptive method choices adolescents make.

In this study, we test the hypotheses that risk-averse adolescents will tend to use more effective contraceptive methods, whereas risk-prone adolescents will be more likely to use less effective methods, or none. We bring together literature about adolescent risky behaviors and the study of inequalities in the use of more effective contraceptive methods, and assess two research questions. First, whether disparities in contraceptive effectiveness reflect patterns in risk-taking propensities or risk-profiles, and second, whether this relationship varies by race-ethnicity. To our knowledge, these specific research questions have not been thoroughly assessed in the past, despite having important policy implications.

As a first step in our analysis, we use nationally representative, pooled cross-sectional data from the National Youth Risk Behavior Survey (YRBS) to identify risk profiles separately for white, Hispanic, and black female high school students, based on their probabilities of substance use, activities conducive to motor vehicle accidents or injury while riding or driving in a vehicle, physical fights, and weapon carrying.^a We adopt a person-centered approach and use Latent Class Analysis (LCA) to identify groups of respondents that are similar in terms of non-sexual risk patterns, separately for white, black, and Hispanic adolescents. Based on rates of unintended pregnancy after one year of typical use,²⁶ we classified contraceptive strategies according to their effectiveness. Then we estimated the association between risk profiles and the relative effectiveness of the contraceptive method used in the last intercourse, separately by race-ethnicity.

Our paper presents several innovations and key contributions relative to the few existing studies assessing the relationship between non-sexual risk taking and contraceptive method choice.^9,24,30,41 First, we assess a broader spectrum of non-sexual risky behaviors, including not only substance use, but also risk-taking when riding or driving in a vehicle, physical fights, and weapon carrying. Second, as mentioned above, we use separate latent class models to identify the different risk profiles among black, Hispanic, and white adolescents, which were neither explored nor accounted for in these previous studies. Third, our outcome categories encompass a variety of contraceptive strategies that have rarely been analyzed together in the aforementioned studies: no method, withdrawal, condom alone, prescription contraception, and prescription contraception paired with condom. Fourth, we use nationally representative data of female adolescents for multiple years and multivariate analysis to assess the relationship between risk profiles and the probability of using each of those contraceptive strategies. And fifth, our study focuses on the relationship between risk profiles and the effectiveness of the contraceptive method used relative to all possible options, rather than on the use of a particular subset of methods. No previous study assessing similar or related research questions has combined all these features. After reviewing the existing literature and presenting our methods and results, we discuss the research and policy implications of our findings.

Background

Risk-taking in Adolescence

Research on adolescent risk-taking has traditionally focused on studying activities of non-sexual character, such as getting involved in fights or carrying a weapon, substance use and abuse, and unsafe practices while driving or riding a motor vehicle,^36,37 as well as behaviors of sexual nature, such as having multiple sexual partners.^18,44 Because risky behaviors during adolescence tend to be correlated, recent research has adopted a person-centered approach that aims to identify profiles of risk-taking among adolescents.^7,20,49,54 A frequent finding in the aforementioned studies is that there is evidence for a common etiology of risk-taking, with individuals who engage in a particular risky behavior, such as physical fights, also being more likely to present other risky behaviors, such as lack of condom use.⁴⁹ This tendency towards risk-taking has been interpreted as a consequence of person-level characteristics, such as higher risk-proneness,⁴² low self-control,¹⁷ or a “problem behavior syndrome,”²⁷ all of which are plausibly shaped by socialization, and socioeconomic and environmental conditions.¹⁷

Although there is evidence that adolescent risk-taking varies across racial-ethnic groups,^14,35 the few studies that have assessed the relationship between risk-taking and contraceptive effectiveness have not accounted for this heterogeneity.^9,24,30,41 There are multiple reasons why we would expect risk profiles to differ by race-ethnicity. White, black, and Hispanic adolescents grow up in different family and sociodemographic contexts, with contrasting cultural expectations, and are exposed to different levels of disadvantage and discrimination, all of which predict divergent patterns of risk-taking.^14,35 Furthermore, the relationship between risk-taking and protective characteristics, such as family and school bonds, varies by race-ethnicity.¹⁴ Because black, white, and Hispanic adolescents also face heterogeneous barriers to access contraception, are treated differently by healthcare providers, hold divergent preferences for contraceptive-method characteristics, and have contrasting attitudes towards adolescent childbearing,¹⁵ it is also possible that the relationship between risk-taking and contraceptive use may vary across racial-ethnic groups.

Assessing racial-ethnic heterogeneity in risk-taking, and in the relationship between risk-taking and contraceptive effectiveness is essential to understand how this potential internal barrier to contraceptive use may affect the rates of unintended pregnancy for black, Hispanic, and white adolescents. Furthermore, it is key to tailor specific policy measures that reduce such rates for each racial-ethnic group.

Risk-taking and Contraceptive Effectiveness

The relationship between risk-taking patterns and disparities in contraceptive effectiveness has been scantly studied. Most studies analyzing the co-occurrence of non-sexual and sexual risky behaviors have primarily focused on binary measures of contraceptive risk, such as lack of any contraceptive use and lack of condom use.^{5,6,12,20,37,44,51} These studies have not assessed whether proneness to non-sexual risk-taking is associated with the effectiveness of the contraceptive methods used by adolescents. The few studies that have specifically analyzed the association between non-sexual risky behaviors and the effectiveness of contraceptive strategies have not assessed racial-ethnic differences in such relationships, and have either used small convenience samples,^24,41 or focused exclusively on associations with substance use, such as alcohol, cigarettes, marijuana, and cocaine,^9,30 while excluding other non-sexual risk-taking from their analyses.

These previous studies have found mixed evidence about the relationship between non-sexual risk-taking and contraceptive method choice. Using bivariate analysis and a convenience sample of 244 sexually active female college students, one study found that an index of non-sexual risky behaviors (drug use, criminal activities, risky automobile experiences, and being in unsafe situations) was unrelated to the use of either condom, hormonal pill, hormonal pill combined with condom, or no contraceptive method.⁴¹ Another study evaluated whether “rule-breaking” behaviors, such as cheating or stealing were associated with using different prescription contraceptive methods in the past two weeks using a sample of 262 teenage girls recruited from a community clinic and multivariate analysis.²⁴ They found evidence of a positive association between use of injectable contraception and rule-breaking behavior. The authors attributed this finding to forward-looking decisions on the part of adolescents and their parents, given that rule-breaking may be correlated with low self-efficacy, and thus with a lower probability of following up with a more user-dependent method, such as the pill.²⁴ However, this study did not evaluate risk profiles based on a broad spectrum of risky behaviors, and did not analyze non-hormonal contraceptive use.

We know of only two studies that have used nationally representative data and multivariate analysis to assess the association between non-sexual risk-taking and contraceptive method choice. The first used data on female and male adolescents interviewed by the 1999–2007 YRBS to evaluate the relationship between adolescent “risk scales,” defined based on a sum of substances used by respondents (cigarettes, marijuana, alcohol, and cocaine), and the contraceptive method used in the last intercourse.⁹ These authors did not find a significant relationship between their risk scale and contraceptive method choice among females.⁹ The second study used data from the 2011 YRBS to assess the relationship between contraceptive method choice, with a focus on withdrawal, and indicators of adolescent substance use.³⁰ They found that female adolescents with substance use are more likely to use withdrawal relative to condoms and highly effective methods, but less likely to use withdrawal relative to no method.³⁰ Therefore, the two studies^9,30 that used nationally representative data to assess the relationship between risk-taking and contraceptive effectiveness are based only on substance use, and did not evaluate the existence of heterogeneous associations across racial-ethnic groups. We argue that underlying risk-taking propensities are better measured using a broader set of non-sexual risky behaviors.

Methods

Data

We use data from the 2011, 2013 and 2015 waves of the National YRBS¹⁰. This is a cross-sectional, biannual survey representative of students enrolled in grades 9 through 12 in both public and private high schools in the 50 states and the District of Columbia.¹¹ The data collection protocol is described elsewhere.²⁸ De-identified National YRBS microdata for all waves are publicly available,¹¹ and no additional IRB approval is required for their use. Beginning in 2011, the survey started collecting detailed data about use of prescription contraception methods other than the pill.¹¹ Even though the 2017 YRBS data is now available to the public, some of our key variables, such as drinking five or more drinks in a row, and use of prescription drugs, are not comparable for this latest round, so our analysis is restricted to waves 2011, 2013, and 2015. A total of 15,425 female and male students were interviewed in 2011; 13,583 in 2013; and 15,624 in 2015.¹¹ All of our analyses use the appropriate survey weights to produce nationally representative estimations. We focused on 6,740 female respondents who had been sexually active in the three months previous to the interview, understood as having had sexual intercourse.

Respondents were asked to identify their race from the following options: white, black or African American, Asian, American Indian or Alaska Native, and Native Hawaiian or other Pacific Islander. In a separate question, they were asked if they were Hispanic or Latino. We classified respondents into three categories: non-Hispanic white, non-Hispanic black, and Hispanic. We excluded 615 (9% of sexually active females) non-Hispanic respondents who identified their race as Asian, American Indian or Alaska Native, or Native Hawaiian or other Pacific Islander, because their sample sizes were very small. We further excluded 12 respondents younger than 13 at interview time (less than 1% of remaining sample). Finally, we excluded 63 respondents (1% of remaining sample) due to missing values in key variables such as contraceptive method used in the last intercourse, age, number of sexual partners, and age at first sexual intercourse. Our main working sample comprised 2,871 white, 1,886 Hispanic, and 1,214 black sexually active female high school students aged 13 to 18.

All of our analyses were based on the sample described above, with the exception of the LCA, which used all female respondents aged 13 to 18, regardless of sexual activity (9,110 white, 6,579 Hispanic, and 3,760 black female adolescents). We assess each group’s risk profiles and each respondent’s membership to those profiles based on the complete sample because adolescent sexual activity can be considered a risky behavior itself and is likely correlated with non-sexual risk-taking. If we based LCA models only on sexually active adolescents, we would be underestimating the risk-proneness of some individuals in our sample, since we would be excluding those at the lower end of the risk-taking distribution. In our LCA models, we used pairwise present deletion³⁹ to preserve the 14% of observations with missing values in any of the non-sexual risk variables in our sample. As a robustness check, we reproduced our entire analysis using listwise deletion at all steps, and found closely similar results (not shown, available upon request).

Contraceptive Use

To assess contraceptive use, all respondents were asked two questions: “The last time you had sexual intercourse, did you or your partner use a condom?” The response options were “A. I have never had sexual intercourse,” “B. Yes,” and “C. No.” The next question asked to all respondents: “The last time you had sexual intercourse, what one method did you or your partner use to prevent pregnancy? (Select only one response.)” The response options were “A. I have never had sexual intercourse,” “B. No method was used to prevent pregnancy,” “C. Birth control pills,” “D. Condoms,” “E. An IUD or implant,” “F. A Shot, patch, or birth control ring,” “G. Withdrawal or some other method,” and “H. Not sure.” Only in 2011, options E and F were combined. Based on their responses to these two questions, we classified adolescents into the following mutually exclusive categories of contraceptive use in the last intercourse: condom only, condom and prescription contraceptive, prescription contraceptive, withdrawal, and no contraceptive method. Our “prescription contraceptive” category includes respondents who selected the birth control pill, IUD, implant, patch, injectable contraception, or the ring. Because the most popular contraceptive methods were represented in response options, we assume that the adolescents who marked the “G. Withdrawal or some other method” response and used a method other than withdrawal were using a method of similar effectiveness, such as fertility awareness.²⁶ For simplicity, we will refer to this group as “withdrawal” throughout the paper. Because only 1% of respondents reported they paired condom use with “withdrawal or some other method”, we grouped them into our “condom only” category. Finally, we included respondents who marked “H. Not sure” into our “no contraceptive method” category.

Based on rates of unintended pregnancy after one year of typical use,⁵⁰ we classified contraceptive strategies in the following order of effectiveness, from lowest to highest: no method, withdrawal, condom alone, prescription contraceptives alone, and prescription contraceptive combined with condom. We will refer to prescription contraceptives alone or combined with condom as “more effective” methods. We will refer to withdrawal and condom alone as relatively “less effective”, or riskier methods, while recognizing that using condom alone is more effective than withdrawal.⁵⁰

Our effectiveness classification is exclusively based on pregnancy prevention. Notably, although condoms are less effective than prescription contraceptives to prevent pregnancy, they protect against sexually transmitted diseases (STDs). If STDs are a concern, it is not clear whether low risk-taking girls would be more likely to use a prescription contraceptive alone instead of condom alone. In contrast, combining prescription contraceptive use and condoms is unambiguously the most effective strategy to prevent both unintended pregnancy and STDs.

Non-Sexual Risky Behaviors

We selected nine binary indicators covering risk-taking in a variety of non-sexual domains, indicating whether the respondent a) rarely or never wore a seatbelt when riding a car driven by someone else; b) in the past month, rode in a car driven by someone who had been drinking alcohol; c) smoked cigarettes in the past month; d) had five or more drinks in a row in the past month; e) smoked marijuana in the past month; f) ever used heroin, inhalants, methamphetamines, or cocaine; g) ever used prescription drugs without a doctor’s prescription; h) carried a weapon such as a knife, gun, or club in the past month; and i) engaged in a physical fight in the past year. These variables combine the different types of non-sexual risky behaviors that are typically used in studies focused on youth risk,^6,20,54 while not suffering from excessive missing values.

Risk Profiles and Latent Class Analysis

In the first part of our analysis, we used Latent Class Analysis (LCA) to identify underlying latent classes or typologies of risk-taking. LCA is a commonly used strategy in person-centered analyses of risky behaviors.^12,13,20,32 In our specific application, LCA groups individuals who have similar response patterns to the aforementioned questions about non-sexual risky behaviors, and who are highly different from other adolescents in terms of such responses.⁴³ And advantage of LCA is that it reveals the risk-taking typologies of individuals that actually exist in the data, as opposed to constructing them according to predefined hypotheses of what risk profiles may look like.³⁴ After cleaning and recoding the non-sexual risk indicators in Stata 15, we exported our data to MPlus 6.1, a software designed to estimate LCA and similar models. An exploratory LCA based on a sample that pooled white, Hispanic and black revealed statistically significant racial-ethnic differences in the probabilities of belonging to each risk-profile (results not shown). Moreover, previous studies have found substantial differences in how youth of different racial and ethnic groups engage in risk-taking.^13,28,35 For these reasons, we expected white, black, and Hispanic girls to have distinct latent behavioral risk typologies, and estimated separate LCA models by race-ethnicity based on our nine non-sexual risk variables, while accounting for survey weights. This represents an innovation relative to many existing studies aimed to identify risk profiles or indices based on multiple risky behaviors, which pool racial groups for their analyses, and introduce race only as a control variable.^6,20,48,54

For each subgroup, we estimated models with one to five classes and used a series of goodness of fit indicators to evaluate them. We assessed the Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), and sample-size adjusted BIC. Better fitting models have lower AIC and BIC values, although BIC is a preferable fit indicator for LCA.⁴⁰ We calculated the Lo-Mendell-Rubin Likelihood Ratio Test (LMR LR) and the Vuong-Lo-Mendell-Rubin Likelihood Ratio Test (V-LMR LR), which determine if there is a statistically significant improvement after adding one more class.⁴⁰ Finally, we calculated the entropy, for which larger values denote greater class delineation.³⁸ Given that the AIC, BIC, and entropy do not necessarily point to a single best model (Table 1), we followed the conventional advice of not adding more classes once the V-LMR LR test is non-significant for a particular model.⁴⁰ With these criteria in mind, and after assessing the estimated latent classes for theoretical interpretability, the three-group models were the best fitting for whites and Hispanics, whereas the two-class model emerged as the best fit for blacks. MPlus reported the probability that each respondent would belong to each of the classes in the corresponding model. Following conventional practice, we assigned each respondent to the class for which this probability was the highest,^33,34 and exported the data back to Stata 15 for further analyses.

Table 1.

Goodness of fit indicators for latent class models of risk profiles, by race-ethnicity and number of classes

	Goodness of Fit Indicators
	Akaike Information Criterion (AIC)	Bayesian Information Criterion (BIC)	Sample-size adjusted BIC	Vuong-Lo-Mendell-Rubin Likelihood Ratio Test	Lo-Mendell-Rubin Adjusted Likelihood Ratio Test	Entropy
White sample
2-class	57165.4	57300.6	57240.2	***	***	0.827
3-class	56547.5	56753.8	56661.7	**	**	0.734
4-class	56230.5	56508.0	56384.1	N.S.	N.S.	0.730
5-class	56150.4	56499.0	56343.3	N.S.	N.S.	0.745
Hispanic sample
2-class	46419.2	46548.2	46487.8	***	***	0.803
3-class	45947.2	46144.2	46052.1	**	**	0.707
4-class	45837.9	46102.8	45978.9	N.S.	N.S.	0.771
5-class	45753.9	46086.7	45931.0	N.S.	N.S.	0.709
Black sample
2-class	23169.2	23287.5	23227.2	***	***	0.747
3-class	23080.6	23261.3	23169.1	N.S.	N.S.	0.712
4-class	23023.4	23266.3	23142.4	N.S.	N.S.	0.624
5-class	22980.2	23285.5	23129.8	N.S.	N.S.	0.659

Notes:

^*p<0.05,

^**p<0.01,

^***p<0.001. N.S. = Non-significant.

Multinomial Logistic Analysis

In the second part of our analysis, we evaluated whether risk profiles were associated with the contraceptive method used in the last intercourse using multinomial logistic regressions. These models controlled for age, survey year, and two sexual risk-taking measures: having had four or more sexual partners and having had vaginal intercourse before age 13. We included these controls in order to account for sexual history characteristics that may influence contraceptive use, as well as for contraceptive learning that might be acquired with sexual experience. We estimated this model separately by race, first using “no method” as the outcome reference category, in order to test whether riskier profiles are associated with using any method relative to no contraception; and then using “condom only” as the outcome reference category, to test whether risk profiles are associated with using more and less effective contraceptive strategies, relative to a method with medium-level effectiveness. Our tables present relative risk ratios from these models. Based on these results, we also estimate the average predicted probabilities of using each contraceptive method for each risk profile and race-ethnicity, while using the observed values for sexual behavior measures, age, and year.

Results

Sample Characteristics

White adolescents are less likely to report they did not use any contraceptive method in their last intercourse (12%), compared to Hispanics (24%) and blacks (23%) (Table 2). Similarly, white adolescents are more likely to have used more effective methods, such as prescription contraception, with or without condom (38%), whereas Hispanics and blacks are half as likely to have used these strategies (18 and 20%, correspondingly). Black and white adolescents are also more likely to have had multiple sex partners (32 and 36%, correspondingly), compared to Hispanics (24%). In contrast to their white and Hispanic counterparts, black adolescents are more likely to have had sexual intercourse before age 13. The distribution of age at interview was closely similar across racial-ethnic groups.

Table 2.

Characteristics of respondents by race-ethnicity and latent risk profiles. Sexually active female high school students aged 13–18, National YRBS 2011–15

	White				Hispanic				Black
	Total	Abstainers, no marijuana use	Experimenters, low hard-drug use	High substance use and violence	Total	Abstainers	Experimenters, moderate hard-drug use	High substance use and violence	Total	Abstainers, moderate violence	Experimenters, high marijuana use and violence
Observations	2,871	1,148	1,186	537	1,886	795	822	269	1,214	804	410
Weighted percent of each profile	100	39.8	40.7	19.5	100	41.6	42.2	16.3	100	64.0	35.9
CONTRACEPTIVE METHOD CHOICE IN LAST INTERCOURSE
None	11.8	8.9	11.4	18.7	24.2	20.3	24.5	33.6	22.7	19.2	29.0
Withdrawal	11.7	10.7	9.3	18.6	12.0	11.2	12.1	13.6	11.9	11.0	13.7
Condom	38.7	45.3	37.7	27.3	45.6	50.0	45.8	34.2	44.9	49.3	37.1
Prescription contraception	22.9	16.5	27.4	26.6	13.6	14.0	12.5	15.4	13.0	12.6	13.6
Condom and prescription contraception	14.9	18.6	14.2	8.8	4.5	4.4	5.2	3.2	7.4	7.9	6.6
AGE
15 or younger	18.4	19.0	17.5	19.2	23.7	20.9	24.7	28.4	21.4	21.3	21.7
16	26.7	26.0	26.4	28.5	23.4	20.8	26.3	22.6	25.7	23.9	28.9
17	33.7	32.9	35.8	31.0	31.2	35.1	28.1	28.8	30.4	32.8	26.1
18 or older	21.2	22.0	20.3	21.3	21.7	23.1	20.9	20.1	22.5	22.1	23.3
SEXUAL BEHAVIOR
Had intercourse before age 13	4.2	1.7	3.1	11.9	6.8	2.7	7.1	16.2	10.2	7.2	15.7
Has had 4 or more sex partners	31.7	13.2	35.5	61.6	24.3	11.5	28.3	47.0	35.7	29.1	47.4
SURVEY YEAR
2011	35.9	31.9	33.2	49.7	32.1	28.9	33.4	36.9	38.4	41.1	33.4
2013	33.3	32.6	34.8	31.3	34.7	30.4	38.1	37	37.8	37.2	39
2015	30.8	35.5	32	19	33.1	40.7	28.5	26.1	23.8	21.7	27.6

Notes: All descriptive statistics in the table are percentages. Sexually active women are defined as those who had sexual intercourse in the last 3 months. Contraceptive use corresponds to last sexual intercourse in the past 3 months.

Latent Class Analysis of Non-Sexual Behaviors

We identified three risk profiles among white adolescents: “abstainers with no marijuana use” (40%) who had a low probability of engaging in any risky behavior, and notably did not use marijuana; “experimenters with low hard-drug use” (41%) who show a moderate probability of substance use, violence, and riding in a car with a drunk driver, but relatively low hard-drug use; and “high substance use and violence” (20%), who have high probability of using alcohol, cigarettes, marijuana, hard drugs, prescription drugs, and a high risk of being involved in fights (Figure 1). Similarly, we found three risk profiles among Hispanics: “abstainers” (42%) who had a low probability of engaging in any risky behavior, with some marijuana use; “experimenters with moderate hard-drug use” (42%) who show a moderate probability of substance use, including hard-drugs, and violent behavior; and “high substance use and violence” (16%), who have high probability of using alcohol, cigarettes, marijuana, and hard-drugs, and a high risk of being involved in fights.

Figure 1.

Non-sexual risky behaviors by race-ethnicity and latent risk profiles, among sexually active female high-school students aged 13–18, National YRBS 2011–15

Note: N= 2,871 white, 1,214 black, and 1,886 Hispanic sexually active female high-school students aged 13–18.

Finally, we identified two risk profiles among blacks: “abstainers with moderate violence” (64%), who had low risk-taking across most dimensions, except for moderate engagement in marijuana use, riding in a car with a drunk driver, and violence; and “experimenters with high marijuana use and violence” (36%) who had moderate levels of risk-taking in most dimensions, similar to those of Hispanic and white experimenters, except for high probabilities of using marijuana and being in a physical fight. Although these comparisons are relative and do not necessarily represent statistically significant differences, they are consistent with heterogeneous patterns of risk-taking across racial-ethnic groups.

Relationship between Contraceptive Method Used and Risk-Taking

If contraceptive behavior is associated with underlying risk-proneness, as risk-taking increases we would expect to find a decreasing gradient in use of more effective methods, and an increasing gradient in non-use and use of less effective strategies. According to our multinomial logistic models, relative to not using contraception, white adolescents in the “high substance use and violence” group have a 58% lower risk of using condoms alone, and a 68% lower risk of using a prescription contraceptive paired with condom, compared to “abstainers” (Table 3). Furthermore, we find that white adolescents in the “high substance use and violence” profile are more than twice as likely to use withdrawal relative to condom alone, compared to “abstainers” (Table 4). But white girls in the “high substance use and violence” and “experimenters” profiles also have a 70% and 93% higher risk of using a prescription contraceptive alone, relative to condom only (Table 4).

Table 3.

Relative risk ratios from multinomial logistic regressions using latent risk profiles to predict contraceptive method used in last intercourse, with ‘no contraceptive use’ as the reference category. Sexually active female high school students, ages 13–18, National YRBS 2011–15

	White				Hispanic				Black
	Withdraw.	Condom	Presc. Contrac.	Condom & Presc. Contrac.	Withdraw.	Condom	Presc. Contrac.	Condom & Presc. Contrac.	Withdraw.	Condom	Presc. Contrac.	Condom & Presc. Contrac.
NON-SEXUAL RISK PROFILE (Ref: Abstainers, no marijuana use)
Experimenters, low hard-drug use	0.74 [0.17]	0.76 [0.15]	1.29 [0.25]	0.70 [0.16]
High substance use and violence	1.03 [0.30]	0.42*** [0.10]	0.82 [0.20]	0.32*** [0.08]
NON-SEXUAL RISK PROFILE (Ref: Abstainers)
Experimenters, moderate hard-drug use					0.89 [0.19]	0.81 [0.14]	0.79 [0.15]	1.04 [0.30]
High substance use and violence					0.73 [0.23]	0.49** [0.13]	0.69 [0.22]	0.49 [0.25]
NON-SEXUAL RISK PROFILE (Ref: Abstainers, moderate violence)
Experimenters, high marijuana use and violence									0.97 [0.25]	0.55** [0.12]	0.70 [0.20]	0.62 [0.22]
AGE (Ref: 15 or younger)
16	1.28 [0.35]	1.15 [0.26]	1.87** [0.44]	1.70* [0.42]	1.11 [0.36]	0.83 [0.19]	0.90 [0.26]	1.92 [0.80]	0.83 [0.40]	0.70 [0.18]	2.20 [1.08]	1.69 [0.88]
17	1.08 [0.32]	1.04 [0.24]	2.22** [0.57]	1.59 [0.42]	1.46 [0.43]	0.93 [0.20]	1.54 [0.40]	1.80 [0.78]	1.63 [0.81]	0.86 [0.29]	2.89* [1.38]	2.47 [1.13]
18 or older	0.93 [0.28]	0.99 [0.24]	2.88*** [0.79]	1.80* [0.52]	1.06 [0.35]	0.84 [0.18]	1.05 [0.34]	1.53 [0.81]	0.89 [0.42]	0.59 [0.18]	2.36 [1.02]	1.92 [1.08]
SEXUAL BEHAVIOR
Had intercourse before age 13	0.58 [0.22]	0.40** [0.13]	0.51* [0.14]	0.76 [0.28]	0.80 [0.32]	0.43* [0.16]	0.82 [0.31]	1.28 [0.55]	0.99 [0.46]	1.27 [0.45]	2.35* [0.88]	2.21 [1.15]
Has had 4 or more sex partners	0.67 [0.14]	0.52*** [0.08]	1.09 [0.19]	0.49*** [0.09]	1.13 [0.28]	0.76 [0.13]	1.35 [0.33]	0.72 [0.24]	0.7 [0.24]	0.64* [0.12]	0.94 [0.25]	0.54 [0.17]

Notes:

^*p<0.05,

^**p<0.01,

^***p<0.001.

Standard errors not shown. Models are estimated separately by race and ethnicity. Sexually active women are defined as those who had sexual intercourse in the past 3 months. Contraceptive use corresponds to last sexual intercourse in the past 3 months. Relative risk ratios for survey year are not shown to preserve space.

Table 4.

Relative risk ratios from multinomial logistic regressions using latent risk profiles to predict contraceptive method used in last intercourse, with ‘condom only’ as the reference category. Sexually active female high school students, ages 13–18, National YRBS 2011–15

	White				Hispanic				Black
	None	Withdraw.	Presc. Contrac.	Condom & Presc. Contrac.	None	Withdraw.	Presc. Contrac.	Condom & Presc. Contrac.	None	Withdraw.	Presc. Contrac.	Condom & Presc. Contrac.
NON-SEXUAL RISK PROFILE (Ref: Abstainers, no marijuana use)
Experimenters, low hard-drug use	1.32 [0.26]	0.98 [0.20]	1.70*** [0.23]	0.93 [0.14]
High substance use and violence	2.36*** [0.55]	2.43*** [0.57]	1.93*** [0.34]	0.75 [0.13]
NON-SEXUAL RISK PROFILE (Ref: Abstainers)
Experimenters, moderate hard- drug use					1.23 [0.21]	1.09 [0.25]	0.96 [0.19]	1.28 [0.34]
High substance use and violence					2.05** [0.53]	1.50 [0.46]	1.41 [0.45]	1.00 [0.53]
NON-SEXUAL RISK PROFILE (Ref: Abstainers, moderate violence)
Experimenters, high marijuana use and violence									1.82** [0.40]	1.76 [0.50]	1.27 [0.28]	1.13 [0.32]
AGE (Ref: 15 or younger)
16	0.87 [0.19]	1.11 [0.30]	1.62* [0.33]	1.47 [0.30]	1.2 [0.27]	1.33 [0.43]	1.08 [0.34]	2.31* [0.97]	1.43 [0.37]	1.2 [0.58]	3.15** [1.27]	2.42 [1.16]
17	0.96 [0.22]	1.04 [0.24]	2.15*** [0.43]	1.53* [0.31]	1.08 [0.24]	1.57 [0.44]	1.66* [0.41]	1.93 [0.80]	1.17 [0.40]	1.9 [0.86]	3.37** [1.39]	2.89** [1.13]
18 or older	1.01 [0.24]	0.93 [0.25]	2.90*** [0.58]	1.81** [0.41]	1.19 [0.26]	1.26 [0.42]	1.25 [0.34]	1.82 [0.85]	1.69 [0.52]	1.51 [0.71]	3.98*** [1.56]	3.24* [1.57]
SEXUAL BEHAVIOR
Had intercourse before age 13	2.52** [0.83]	1.45 [0.63]	1.28 [0.47]	1.92 [0.82]	2.30* [0.84]	1.83 [0.80]	1.89 [0.69]	2.94** [1.20]	0.79 [0.28]	0.78 [0.32]	1.85 [0.70]	1.74 [0.84]
Has had 4 or more sex partners	1.92*** [0.30]	1.29 [0.24]	2.08*** [0.29]	0.94 [0.15]	1.32 [0.22]	1.49 [0.35]	1.78** [0.38]	0.95 [0.36]	1.56* [0.30]	1.09 [0.31]	1.47 [0.33]	0.85 [0.25]

Notes:

^*p<0.05,

^**p<0.01,

^***p<0.001.

Standard errors not shown. Models are estimated separately by race and ethnicity. Sexually active women are defined as those who had sexual intercourse in the past 3 months. Contraceptive use corresponds to last sexual intercourse in the past 3 months. Relative risk ratios for survey year are not shown to preserve space.

Among Hispanic and black adolescents, using condom alone is the only contraceptive strategy associated with risk-taking, and the relationship is as hypothesized. Compared to “abstainers,” Hispanic females in the “high substance use and violence” profile have a 51% lower risk of using condom alone, relative to no contraception (Table 3). And compared to black “abstainers”, black “experimenters” had a 45% lower risk of reporting using condom alone relative to not using contraception. These findings are mirrored in the models that use “condom only” as the reference outcome (Table 4).

Consistent with our hypotheses, the average predicted probabilities of using each contraceptive method (Figure 2) show that there is a negative relationship between risk-taking and condom-only use for all racial and ethnic groups, and that the probability of dual contraceptive use decreases monotonically as risk-taking increases among whites. Contrary to our hypothesis, it also shows the higher probability of prescription contraception use among higher-risk white adolescents. And among black and Hispanic girls, these predicted probabilities show the lack of association between use of more effective contraceptive methods and risk-taking, which is also contrary to our hypotheses.

Figure 2.

Average predicted probabilities of using different contraceptive methods in last intercourse, by latent risk profiles and race-ethnicity, among sexually active female high-school students aged 13–18, National YRBS 2011–15

Note: N= 2,871 white, 1,214 black, and 1,886 Hispanic sexually active female high-school students aged 13–18. For white teenagers, Abs. = “Abstainers, no marijuana use”; Exp. = “Experimenters, low hard-drug use”; High = “High substance use and violence.” For Hispanic teenagers, Abs. = “Abstainers”; Exp. = “Experimenters, moderate hard-drug use”; High = “High substance use and violence.” For black teenagers, Abs. = “Abstainers, moderate violence”; Exp. = “Experimenters, high marijuana use and violence.” Average predicted probabilities are based on race-specific multinomial logistic models of contraceptive use, where predictors are risk profile, multiple sexual partners, age at first intercourse, age, and survey year.

We found statistically significant associations between sexual behavior measures and contraceptive use. Relative to not using contraception, having multiple sex partners is associated with a lower risk of using condom alone for white and black adolescents, and with a lower risk of using prescription contraception paired with condom for white adolescents (Table 3). For whites and Hispanics, having had sexual intercourse before age 13 is associated with lower risk of condom use, relative to no contraceptive method. And for whites, early sexual onset is also associated with a lower risk of using prescription contraception, relative to no method.

In other cases, sexual risky behaviors were associated with using more effective contraceptive strategies. Black adolescents who had sexual intercourse before age 13 had a higher risk of using prescription contraception alone, relative to no method (Table 3). And Hispanic adolescents who had an early sexual onset had a higher risk of using condom paired with prescription contraception, relative to condom alone (Table 4). For white and Hispanic adolescents, having multiple sex partners was also associated with a higher risk of using prescription contraception alone, relative to condom.

For white and black adolescents, older age is associated with greater risk of using prescription contraception alone, relative to condom (Table 4). For these groups, older age is also associated with a greater risk of using prescription contraception paired with condom, relative to only condom. We did not find monotonic relationships between age and other contraceptive methods.

Discussion

The present study is one of the first to analyze risk profiles separately by race and ethnicity using LCA. We know of only one other study that has identified risk profiles separately by race and ethnicity using this method.¹⁴ Although not directly comparable to our analysis, since it pooled female and male adolescents, our findings are consistent with that study’s main conclusion, that there are substantial racial-ethnic differences in adolescent risk profiles that cannot be adequately captured by pooling racial and ethnic groups to identify latent classes of risk-taking. After analyzing the risky behaviors of white, Hispanic, and black female adolescents using latent class analysis, we found three risk profiles for white adolescents: “abstainers with no marijuana use,” “experimenters with low hard-drug use,” and “high substance use and violence.” Among Hispanics, we found three profiles that nonetheless had some differences with those identified for whites: “abstainers” “experimenters with moderate hard-drug use,” and “high substance use and violence”. For black female adolescents, we found two risk profiles: “abstainers with moderate violence,” and “experimenters with high marijuana use and violence.” Some of the non-sexual risk-taking patterns that we identified are consistent with those documented in previous research, showing that compared to white adolescents, black female adolescents are less likely to engage in substance use, with the exception of marijuana, and are also more likely to be involved in physical fights.⁸ Our results also show that white adolescent females are more likely to have misused prescription drugs, followed by Hispanic and black adolescents, which is consistent with findings in previous studies.⁵³

Once the race-specific risk profiles were identified, our primary goal was to assess their association with contraceptive effectiveness among white, black, and Hispanic female adolescents. We hypothesized that greater risk-taking would be associated with using less effective contraception. Our findings show that risk profiles and contraceptive behavior have a more consistent, albeit imperfect, relationship among white females, compared to their black and Hispanic peers. Consistent with our initial hypothesis, white females in the “high substance use and violence” profile were less likely to use condom alone or paired with a prescription contraceptive, relative to not using contraception. We also found that high-risk white females were more likely to use withdrawal relative to condom alone, which is aligned with our hypothesis, and with findings from a previous study.³⁰

Nonetheless, we also found that white adolescents in the “high substance use and violence” profile were more likely to use prescription contraceptives rather than condoms. Although inconsistent with our initial hypothesis, this is aligned with the findings in another previous study,²⁴ which found that young women who used injectable contraception reported more “rule-breaking” behaviors, compared to those using no method. White adolescents in the “high substance use and violence” profile may plausibly have more frequent sexual encounters and more sexual partners, and may choose to use prescription contraception because they anticipate that they will be at greater risk of unintended pregnancy. High-risk white adolescents or their parents may also make the forward-looking decision of choosing a contraceptive method that requires no further efforts or partner negotiations during intercourse, to reduce the probability of contraceptive inconsistency. Taken together, these findings suggest that risk-taking is an important but previously overlooked correlate of contraceptive effectiveness among white adolescents, although the direction of this association depends on method type.

In contrast to white females, for black and Hispanic females there was not a significant association between risk-taking and use of any method other than condom alone. This means that risk-taking as an internal barrier to effective contraception is mostly relevant for white adolescents, and that the lower use of more effective contraception among black and Hispanic adolescents must be explained by alternative types of—internal and external—barriers. Previous studies have shown that women of color face disadvantages in reproductive health care access, such as lower insurance coverage,¹⁵ and greater discrimination experiences when interacting with healthcare providers.²⁹ In addition, women of color are more likely to report that their own preferences about fertility and contraception are not respected when interacting with providers,²³ and are more likely to believe the government wants to limit the population of people of color,⁴⁵ which undermines their trust in the healthcare system. Non-citizenship or undocumented status may also limit reproductive healthcare access.⁵² The combination of these circumstances may lead Hispanic and black women to prefer contraceptive methods that they can access and control without the intervention of providers, even if these methods are less effective at preventing pregnancy.²⁵

Previous studies have also identified racial-ethnic disparities in internal barriers to contraception such as attitudes and motivations towards motherhood and pregnancy. For instance, although they may not actively desire a pregnancy, Hispanic women are more likely to have a positive attitude towards unintended pregnancies compared to white women.^22,45 Women of color are also more likely to hold ambivalent pregnancy preferences, which are associated with use of less effective contraception.⁴⁷ Self-efficacy, understood as the perception that one’s actions can lead to the achievement of desired outcomes,⁴ is another personal characteristic that has been associated with use of more effective contraceptives.²¹ However, black and Hispanic women tend to have lower confidence in their own agency, and are more likely to hold fatalistic attitudes towards pregnancy,¹⁹ which may contribute to explain the lower use of more effective contraception among adolescents in these racial-ethnic groups. Finally, Hispanic women have also been found to have lower knowledge of effective contraceptive methods, which is associated with lower use of such contraceptive strategies.⁴⁵ Together, these internal and external barriers to contraception may mitigate the association between risk-taking and contraceptive effectiveness among black and Hispanic adolescents.

The pattern that we consistently found across racial-ethnic groups was a negative association between risk-taking and condom use. It is possible that risk-taking female adolescents tend to have risk-taking male adolescents as sexual partners.^1,3 To the extent that condom use is the product of a negotiation with a male sex partner,⁴⁶ the association between risk-taking and lower condom use may at least partially reflect greater difficulty for high-risk females to enforce condom use among their high-risk partners. Similarly, the fact that high-risk white females were more likely to use withdrawal relative to condom alone may at least be partially explained by risk-taking male partners preferring withdrawal over condoms.²

Limitations

Our results should be interpreted in light of several limitations. First, latent class membership was defined based on the class with the highest probability of assignment for each observation. Inevitably, this introduces some degree of uncertainty to our analysis and may attenuate the relationship between risk profiles and contraceptive effectiveness. Second, the YRBS has no information about socioeconomic status, romantic relationship status and context, whether respondents have ever been pregnant, or the reasons why they decide not to use contraception. In particular, we were unable to control for competing barriers to contraceptive use, because these were not observed in our data. Future research should further investigate the relationship between risk-taking and contraceptive effectiveness while accounting for such variables, as well as for alternative barriers to contraceptive use. Lastly, follow-up studies should explore additional risk measures beyond those available in the YRBS to assess whether contraceptive use among Hispanic and black adolescents may be tied to different types of risk than those that differentiate use among whites.

Conclusions

The relationship between risk-taking and contraceptive effectiveness varies by race-ethnicity and by contraceptive method choice. Risk-taking is negatively associated with use of prescription contraception paired with condom only among white adolescents, but it is negatively associated with condom use for all racial-ethnic groups. Policies that aim to reduce both sexually transmitted infections and unintended pregnancy can maximize their impact by targeting adolescents who engage in a broad spectrum of non-sexual risk behaviors to help address underlying characteristics that contribute to overall risk-taking. Nonetheless, in order to increase the use of the most effective contraceptive strategies among black and Hispanic adolescents, interventions must address internal and external barriers other than risk-proneness that may disproportionately affect young women of color. Future research should continue analyzing the differential relationships between risk-taking and contraceptive behavior across racial-ethnic groups.