|

Changes in Sexual Risk Behavior as Young Men Transition to Adulthood

Jacinda K. Dariotis Freya L. Sonenstein Gary J. Gates Randy Capps Nan M. Astone Joseph H. Pleck Frangiscos Sifakis Scott Zeger

First published online:

| DOI: https://doi.org/10.1363/4021808
Abstract / Summary
CONTEXT

Understanding how young men's sexual risk behaviors change during the transition from adolescence to early adulthood is important for the design and evaluation of effective strategies to reduce the transmission of HIV and other STDs.

METHODS

Data from three waves of the National Survey of Adolescent Males (1988, 1991 and 1995) were used to categorize 1,880 respondents into clusters according to sexual risk behaviors. Univariate and bivariate analyses were conducted to assess associations between clusters and rates of self-reported STD diagnoses and positive chlamydia tests.

RESULTS

Two dimensions of sexual risk-taking defined the clusters: partner characteristics and condom use. More than 50% of men remained in low-risk groups over time. In the first two waves, 24–32% of men reported engaging in high-risk behaviors (risky partners, condom nonuse); these behaviors were associated with elevated levels of STD outcomes. Nearly 40% of men who entered a high-risk group in the first two waves transitioned to a lower risk group by the third wave. Nine percent of men either engaged in increasingly risky behaviors or maintained membership in high-risk groups; elevated STD rates characterized both trajectories. Low condom use combined with having multiple partners during adolescence was associated with elevated STD rates in the year preceding the third wave; high condom use coupled with having risky partners was not.

CONCLUSIONS

The prominence of low-risk behaviors over time suggests that most young men avoid sexual risk-taking. Effective strategies to reduce HIV and STD risk in young men must simultaneously address multiple dimensions of sexual behavior.

Perspectives on Sexual and Reproductive Health, 2008, 40(4):218–225

Understanding how young men’s sexual risk behaviors change as they become adults is important to efforts to reduce transmission of STDs, including HIV. Behaviors that can increase STD risk in young men include having sex with strangers, sex workers, injection-drug users or same-sex partners; having concurrent relationships with two or more partners; and using condoms infrequently or not at all. Because each behavior constitutes an independent source of risk, young men who engage in combinations of these behaviors face even greater risks of STD transmission. In this article, we describe the sexual risk groups, as defined by clusters of risk behaviors, that emerge as young men transition from adolescence to early adulthood; the proportions of young men who belong to these groups at various points during this transition; and the relationship between membership in particular risk groups and young men’s STD status.

BACKGROUND

The burden of STDs disproportionately falls on young adults, and the prevalence of HIV is higher among young men than among young women.1 In 2000, nearly 55% of the 2.8 million new cases of chlamydia occurred among youth aged 15–24. Given the asymptomatic nature of chlamydia, incidence is most likely underestimated among men.2 The elevated STD rates among adolescents and young adults result, in part, from the high prevalence of sexual behaviors that put them at risk for infection.3 Choices regarding initiation of sex, partner selection and use of contraceptives (especially condoms) are important determinants of the probability of STD infection.4,5 Engaging in multiple sexual risk behaviors most likely increases, perhaps nonlinearly (e.g., exponentially), the odds that a person will become infected. Yet research on adolescent risk behaviors typically treats specific behaviors independently. For example, findings that levels of sexual experience among teenagers had declined and condom use in the same population had increased have been reported separately rather than empirically linked in analyses.6,7 Moreover, studies that have examined how risk behaviors vary together have been cross-sectional rather than longitudinal.8–10

Cross-sectional studies indicate that many HIV-related risk behaviors, including sexual and injection-drug use behaviors, begin in adolescence and peak in young adulthood.11–15 As young men reach adulthood, the prevalence of risk behaviors and the acquisition of new partners decline; relationships become more committed and long-lasting, and men develop more stable patterns (both professionally and personally) in their lives.13,16 Condom use, which is frequent among youth, tends to decline as young men age, often as a function of relationship type.17–20 These findings are consistent with the hypothesis that as sexual relationships become longer, more stable and more committed, men’s risk profiles change. Cross-sectional data, however, cannot determine whether this developmental hypothesis explains differences in risk better than an alternative explanation—that men in more recent cohorts simply behave differently from those in older cohorts. The current study, which uses longitudinal data, is well suited to examining developmental trends in sexual risk behavior and STD outcomes in the population at greatest risk for contracting STDs, men aged 15–26.

METHODS

Data

We use data from the three waves of the National Survey of Adolescent Males (NSAM). The first wave, conducted in 1988, examined a nationally representative sample of 1,880 never-married men aged 15–19 who were living in households in the contiguous United States. The second wave was conducted in 1990–1991, among 1,676 men, who were aged 17–22. Respondents were interviewed again in 1995, at ages 21–26; a total of 1,377 men participated in this third wave, yielding a 75% follow-up rate across the waves (after omitting the 38 respondents who died between Waves 1 and 3). A total of 1,290 men participated in all three waves. Longitudinal weights were developed to adjust for nonresponse.21 After weighting, we could not identify any significant biases in age or behavioral characteristics.

Our analysis had two phases: an exploratory phase using cluster analysis and a validation phase using univariate and bivariate statistical tests. For the cluster analysis, we used all available data from participants in Wave 1. Each man contributed an observation—the unit of analysis—for each survey year of participation.* Thus, the 1,290 men who were interviewed three times contributed three observations, the 473 who were interviewed twice contributed two observations and the 117 who were interviewed only in 1988 contributed one observation. Analyses were adjusted for the nonindependence of the observations and for estimated robust standard errors using STATA 9.1.

For the validation phase of the analyses, men were the unit of analysis, and the sample size varied depending upon the outcome considered. We used appropriate cross-sectional weights for sampling effects, and longitudinal weights to adjust for nonresponse through the most recent wave.22

Measures

•Predictor variables

The measures available in NSAM that we believe best represent the dimensions of HIV and STD risk described in the literature are number of partners, partner quality, partner concurrency and condom use. We constructed these measures to represent behavior within the year prior to the survey and standardized them to z-scores; details on variable construction are available upon request.

The five measures we used in our cluster analysis were the number of female partners with whom the respondent had had vaginal intercourse in the past year; the number of female partners in the past year whom he had known for less than a day prior to first intercourse (stranger partners); the number of risky sex partners in the past year; the number of months in the past year during which the respondent had had two or more female sexual partners (concurrent partners); and the proportion of sexual acts in the past year during which the respondent had not used a male condom (unprotected sex). All of the measures are continuous.

•Outcome variables

In all three waves, respondents reported whether a doctor had ever told them that they had syphilis, gonorrhea or herpes. In Waves 1 and 2, respondents were also asked about genital warts; in Wave 3, they were also asked about chlamydia. For each wave of the survey, we constructed a measure indicating whether the respondent had ever had an STD diagnosed using his reports at that wave or previous waves. An additional self-reported STD measure was collected in Wave 3: STD diagnoses within the past year.

Finally, in Wave 3, respondents were asked to provide a urine sample to be used for chlamydia testing.23 Specimens were tested for the presence of C. trachomatis using a commercial polymerase chain reaction kit. Test results were unavailable for 28% of the 1,377 respondents in Wave 3. In half of the cases with missing test results, respondents were unwilling to take the test. In the remaining cases, either no urine sample was obtained because the interview was done over the phone; the test results were lost, damaged or mislabeled; or the urine sample was insufficient for testing. Analytic testing using multiple imputation determined that nonresponse bias had a negligible effect on the prevalence estimates. Although only a portion of the sample underwent chlamydia testing, the advantage of using this highly valid measure (as opposed to self-reports) offsets the disadvantage of reduced sample size for these analyses.

Analytic Strategy

Our first research question was, To what extent can young men be classified into meaningful groups based on how they select partners and use condoms? To answer it, we used cluster analysis, a multivariate procedure that organizes a set of observations into relatively homogeneous groups on the basis of a set of characteristics in a given time period.

A variety of clustering methods exist; we used k-means, a nonhierarchical method that allowed us to specify the number of clusters in the analysis.§ Because we wanted the clusters to have similar meaning over time, we generated them using all available observations from sexually experienced respondents across the three survey waves, and each respondent was assigned a cluster membership for each observation he contributed.†† We then computed z-scores across all time points for each risk variable. Hence, a respondent’s risk behavior score for a given wave was relative to that of all risk behaviors over time. On the basis of his risk behaviors across the five variables, each respondent was then classified into a cluster for each wave.

Data from men who reported at a particular wave that they had never had heterosexual intercourse were excluded from the cluster analysis. In total, 4,031 observations were used in the analysis (1,269 from the first wave, 1,452 from the second and 1,310 from the third).

Our second research question was, What patterns emerge over time regarding men’s adoption of safer sexual practices versus risky sexual practices? To address this issue, we examined the trajectories of men’s movement among risk groups. We were most concerned with men who moved from lower to higher risk groups and men in high-risk groups who increased their risk.

Our final research question was, How do these risk groups and transitions among groups relate to STDs? We determined the proportion of members in each risk group in Wave 3 who reported having ever had an STD, reported having had an STD in the past year or tested positive for chlamydia in Wave 3. In addition, we examined whether specific risk group trajectories were associated with having ever had an STD by Wave 3. These analyses were performed to validate our hypothesis that young men in the various clusters differed in STD risk.

RESULTS

Descriptive Findings

The proportion of respondents who had never had heterosexual sex declined from 40% in Wave 1 to 6% in Wave 3 (Table 1). Among men who had ever had heterosexual sex, the proportion who, during the past year, had had stranger, risky or concurrent partners increased between middle adolescence (ages 15–19) and late adolescence (ages 17–22), and then declined in early adulthood (ages 21–26), suggesting a "settling down" effect. For instance, the proportion who reported having had concurrent partners in the past year increased from 13% (Wave 1) to 20% (Wave 2), and then decreased to 13% (Wave 3).

By contrast, the proportion of sex acts that were unprotected increased steadily, from about one-fourth in the first wave to three-fifths in the final wave. This rise in unprotected sex may indicate increased risk for respondents who engaged in other sexual risk behaviors (e.g., having multiple partners), or it may reflect decreasing risk for respondents who formed stable, monogamous sexual relationships in early adulthood.

The proportion of sexually experienced young men who reported having ever had an STD increased from 2% in Wave 1 to 7% in Wave 2 and 12% in Wave 3. This trend can be explained in part by the increase over time in the proportion of respondents who were sexually experienced, and in part by the increase in unprotected sex across waves. In Wave 3, 1% of men reported that they had had an STD diagnosis in the past year, and 4% had a positive chlamydia test.

Cluster Composition

We began our cluster analysis by specifying a five-cluster solution, but found that a four-cluster solution worked better because it yielded two lower risk and two higher risk groups, which differed in partner characteristics and condom use—two important dimensions of STD risk described in the literature. We also attempted a three-cluster solution, but it did not capture as much variation along these dimensions of risk. The final four-cluster solution explained 24% of the variance in the five sexual risk measures. Men who had not had heterosexual sex constituted a fifth cluster.

Two of the clusters are considered low-risk because their members had had relatively few partners overall, and had rarely had stranger, risky or concurrent partners (Table 2). Condom use differentiates these clusters: One group had engaged in unprotected sex relatively infrequently—in 15% of their sexual acts, on average—whereas the other group had had unprotected sex 93% of the time. For this reason, we refer to these groups as "low-risk/high-protection" and "low-risk/low-protection," respectively.

Two higher risk clusters also emerged. One resembled the low-risk/high-protection group on all measures except number of risky partners in the past year. We refer to this group as "risky-partners/high-protection." Young men in this group engaged in safer-sex practices more often than their counterparts in the other high-risk group, which we term "many-partners/some-protection." Men in the risky-partners/high-protection group had had sexual partners who were at elevated risk for HIV and other STDs, but they had had sex with fewer partners (both concurrently and overall) and had a relatively low rate of unprotected sex (on average, 17% of their sex acts were unprotected). Men in the other high-risk group, many-partners/some-protection, had markedly higher levels of three of the four partner risk variables in the past year: number of partners, number of stranger partners and number of months with concurrent partners. They reported a moderate number of risky sexual partners relative to the other clusters, and despite the risky nature of their partnering practices, 61% of their sex acts were unprotected. To put the behavior of this risk group in context, men in the many-partners/some-protection group reported having, on average, more than seven female partners in the past year—nearly four times the number reported by the other high-risk group. Furthermore, they had had two or more concurrent partners for nearly two-thirds of the past year, whereas men in the other groups reported having had concurrent partners for an average of less than one month.

Stability of Risk Groups

In all three waves, most respondents were in the no-heterosexual-sex category or the two lower risk clusters (Table 3). The relative sizes of both low-risk groups increased over time: The proportion of respondents in the low-risk/high-protection group doubled, from 19% in Wave 1 to 37% in Wave 3, and the proportion in the low-risk/low-protection group nearly tripled (from 17% to 48%) during that time. These trends may provide further evidence of a settling down effect—especially among men in the low-risk/low-protection group.

Combined membership in the two high-risk groups increased from 24% to 32% between Waves 1 and 2, and then decreased to 9% in Wave 3. This peaking of risk has been observed in the literature and recognized as reflecting sexual experimentation during late adolescence.5 The changes in the high-risk groups differed in degree, however. In Wave 1, the risky-partners/high-protection group encompassed 20% of respondents (making it the largest, by a small margin, of the sexually experienced risk groups); the proportion increased to 22% in Wave 2 and then declined dramatically, to 2%, in Wave 3. The proportion of respondents in the many-partners/some-protection group was 5% in Wave 1, doubled to 10% in Wave 2 and then declined to 7% in Wave 3. Thus, even as respondents reached adulthood (ages 21–26), approximately one in 15 still had numerous partners and did not use condoms regularly.

In addition, the distribution of men between these two high-risk groups changed over time. During the first two waves, the vast majority of men in high-risk clusters (81% and 68%, respectively) were members of the risky-partners/high-protection group. By the third wave, the distribution had reversed, and 75% of high-risk men were members of the many-partners/some-protection cluster.

Between the first and second waves, four of the five risk groups showed marked stability (Table 4). The low-risk/low-protection and many-partners/some-protection groups were particularly stable, as 61% of the men who had been in the former group at Wave 1 and 48% of those who had been in the latter group were in the same group in Wave 2. Alarmingly, 23% of men who had been in the no-heterosexual-sex group in Wave 1 had moved into a high-risk group by Wave 2. Meanwhile, members of the two high-risk groups at Wave 1 who transitioned into a new category by Wave 2 were most likely to move to the low-risk/low-protection group. Thirty-five percent of young men in the risky-partners/high-protection cluster and 28% of those in the many-partners/some-protection group showed this pattern.

The transition from high- to low-risk groups was even more striking between Waves 2 and 3. By the latter wave, 81% of men who had been in the many-partners/some-protection group in Wave 2 and 87% of those who had been in the risky-partners/high-protection group in Wave 2 had moved into a low-risk group. Overall, nearly 40% of men who entered a high-risk group in the first two waves had transitioned to a lower risk group by the third wave (not shown). In contrast to the high proportion of young men in the no-heterosexual-sex group who moved to one of the high-risk groups between the first two waves, fewer than 1% of those who had been in the no-heterosexual-sex group in Wave 2 had moved to a high-risk group by Wave 3.

Risk Groups and STDs

Across survey waves, lifetime STD rates increased steadily for each of the risk groups (not shown). This finding is consistent with what would be expected of the natural developmental progression. Given the availability in Wave 3 of expanded measures of STD diagnoses and of urine samples for chlamydia testing, we focus on Wave 3 STD findings.

STDs in the past year and positive chlamydia tests were both more prevalent among young men in the many-partners/some-protection cluster than among those in the low-risk groups (Table 5). Kruskal-Wallis nonparametric tests of significance revealed differences among the groups for both STD measures (not shown). Because Kruskal-Wallis tests serve as an omnibus test revealing whether any difference among groups is significant, we conducted pairwise comparisons and discovered that for each STD measure, the many-partners/some-protection cluster differed from both low-risk groups but not from the other high-risk group. Men in the low-risk groups reported STD diagnoses and tested positive for chlamydia at rates comparable to men in the risky-partners/high-protection cluster; although the lack of statistically significantly differences may in part be an artifact of the small size of the latter group and the limited number of respondents tested for chlamydia, the findings underscore the point that low-risk does not equal no risk.

To further validate the power of our high-risk clusters to predict self-reported and actual STD infection, we examined cluster membership over time. For the 1,290 young men who participated in all three waves, we determined respondents’ cumulative risk group—whether they had always been in either the no-heterosexual-sex group or a lower risk group, had ever been in one of the higher risk groups but not in both, or had been in both of the high-risk groups. Then, we assessed the STD measures by cumulative risk group membership.

Nine percent of men either engaged in increasingly risky behaviors or maintained membership in high-risk groups. Specifically, 5% of men who had engaged in low-risk behaviors in their late teenage years were engaging in high-risk behaviors fairly persistently by their early 20s, as they belonged to high-risk groups in Wave 2 and Wave 3. Moreover, an additional 4% of respondents engaged in high-risk behaviors in both Wave 1 and Wave 3 (not shown).

More than half of young men who participated in all three waves had always belonged to one of the three lower risk clusters (including no-heterosexual-sex), and this group reported the lowest rates of cumulative and recent STDs in Wave 3 (Table 6). Nearly 5% of men who had always been in a lower risk group tested positive for chlamydia.

Thirty-nine percent of men participating in all three waves had been members of the many-partners/some-protection group, but not the other high-risk cluster, at some point between Waves 1 and 3; in 1995, some 32% of men in this cumulative risk group category reported ever having had an STD, 4% had had an STD within the past year and 3% tested positive for chlamydia. Of the 6% of men who had been members of both high-risk groups, 21% reported ever having received an STD diagnosis, 5% had had an STD diagnosis within the past year and 10% tested positive for chlamydia.

Men who had belonged to both high-risk groups had significantly higher rates of all three STD outcomes than did men in the other groups, except for those in the many-partners/some-protection group. Men with membership in the latter group had higher rates on two of the three STD measures than men who were always in a low-risk group and those in the risky-partners/high-protection group.

Risk Group Trajectories

To further explore the relationship between group membership and STD risk, we explored lifetime STD status (ever had an STD by Wave 3) by risk-group transitions over time. By tracing membership stability in and shifts among risk groups as young men move from adolescence into early adulthood, we take full advantage of the longitudinal nature of NSAM.

We highlight three findings (not shown). First, there are many possible trajectories—85 of them across the three waves. Just 39 of these trajectories characterized 90% of the men. Research on other types of trajectories has often observed that a large plurality (20–30%) of individuals fall in one or two trajectories, which might therefore be described as normative.24 In contrast, the most common trajectory we observed (not engaging in sex in Waves 1 and 2 and then being in the low-risk/high-protection cluster in Wave 3) characterized only 7% of the young men, and the proportions were much smaller for the next most frequent trajectories. Despite the small proportions of men in the individual trajectories, general patterns did emerge. Eight of the 10 most common trajectories involved only the low-risk groups, and men in these trajectories constituted 41% of the sample. Seventeen percent of the sample delayed sex for at least two waves. However, high-risk sexual activity did occur: Nearly 10% of the young men—who accounted for two of the 10 most common trajectories—were members of the risky-partners/high-protection group during the first or second wave.

Second, membership in a high-risk group was associated with an elevated cumulative STD risk. Of the 12 trajectories associated with the highest lifetime STD levels, nine included membership in at least one high-risk group at some point.

Third, some evidence suggests a dose-response association between the behavioral risk groups and STD status. Men who were characterized as high-risk in all three survey waves or in the last two reported the highest levels of lifetime STD diagnoses at Wave 3.

DISCUSSION

Effective strategies to reduce young men’s risk of HIV and other STDs must take into account several aspects of sexual behavior, including condom use; partner concurrency; and the number, frequency and types of partners. Men’s level of risk in one of these areas does not necessarily reflect risk in another, and thus changing one set of risky behaviors need not result in behavioral changes in other areas.

Our findings suggest that some men have sex with risky partners but use condoms quite frequently with these partners, whereas other men engage in multiple partnerships and tend to use condoms with substantially less frequency. In NSAM, these two high-risk groups were distinct; although both had higher rates of STDs than did low-risk groups, men who belonged to the many-partners/some-protection cluster for one or more waves reported markedly higher levels of past and recent STDs than did men in the other high-risk group.

Even though we classified two groups as low-risk, men in these groups contracted STDs. Perhaps most alarming, a nonnegligible proportion of them tested positive for chlamydia and reported not having received an STD diagnosis from a physician, either in the past year or in their lifetime (not shown). These men were carriers of STDs but presumably did not consider themselves at risk or present symptoms, and their behavior and choice of partners may have been riskier than they perceived and reported. Efforts to reduce HIV and STD risk should focus on targeting a different mix of strategies—e.g., increasing condom use, changing partner selection and reducing the number of sexual partners—to different groups of men.

Refinement of Safer-Sex Messages

The results of this study suggest that safer-sex messages should not only include recommendations for consistent condom use, but also highlight the need for individuals to be aware of how the number and types of their partners influence their STD risk. If condom-use recommendations are the centerpiece of efforts to promote safer sex, this study uncovers both good and bad news regarding men’s sexual behaviors during adolescence and the transition into adulthood.

First, the message to use condoms seems to have reached most of the people who needed it, with the unfortunate caveat that men in the risky-partner/high-protection group were not applying this message to a significant minority (17%) of their sex acts. This group included about one-fifth of the sample through age 21. In light of these findings, the following refinement of safer-sex messages may be necessary for men with this risk pattern: Although they have done fairly well in using condoms, they must be encouraged to do even better. Even if they use condoms 80% of the time, the 20% of the time they do not is putting them at high risk, given the nature of their partners.

Second, condom-use messages may have fallen short with men in the many-partners/some-protection group. More than half of these men’s sexual acts did not involve condoms. Thus, two aspects of these individuals’ behavior—having numerous partners and not using condoms—put them at risk. The reasons for the lack of condom use among young men with this risk profile need to be explored and addressed.

Third, across adolescence and into adulthood, men moved among risk groups. The overarching trend was a downward shift in risk. A small proportion of men, however, transitioned from low-risk to high-risk groups. While it is important to recognize that young men vary in their need for safer-sex messages, many individuals move among risk groups and thus would benefit most from safer-sex messages that address numerous dimensions of risk.

Future Directions

Effective STD program design and evaluation can be guided by the notion that sexual behavior is dynamic; some men engage in higher risk behaviors during adolescence, and others during adulthood. Addressing risk in teenagers does not guarantee that risk behaviors will not increase in early adulthood. Our findings suggest that young men who practice at least one of the two safer-sex behaviors, having fewer partners or using condoms consistently, contract STDs less often than men who adopt neither behavior. Young men who adopt both approaches over time appear to have the best STD outcomes. Future research should seek to identify barriers to young men’s adopting both behaviors, ways in which these barriers differ among groups of men and strategies to successfully address these barriers.

Limitations and Contributions

The longitudinal nature of NSAM allowed us to assess stability and change over the developmental course from adolescence to adulthood. A substantial proportion of the original sample, however, was lost to follow-up in Waves 2 and 3. Although we used weights to adjust not only for sampling and design effects but also for selective attrition, some of our cell sizes for high-risk groups were quite small by Wave 3. The resulting reduction in statistical power was exacerbated by the tendency for young men to "age out" of risk. However, we found meaningful and significant patterns in young men’s sexual practices that have implications for prevention and intervention efforts, and that we believe will contribute to moving the field forward and reducing STD risk for young men. The next logical steps, which were outside the scope of this study, would be to use risk clusters at earlier waves to predict STD status at later waves, and to examine whether and how age, race and ethnicity influence the patterns we observed.

The most recent round of NSAM interviews was conducted in 1995. Thus, our findings may not apply to more recent cohorts. However, we were able to compare our 21–26-year-olds with men of the same age in 2002 who participated in the National Survey of Family Growth. We calculated that in 2002, men reported having had an average of 1.9 female sex partners in the past 12 months. After weighting for attrition and oversampling, men in the 1995 wave of the NSAM reported having had 2.0 female partners in the past 12 months. This similarity suggests that our trajectory results may apply to later cohorts. The forthcoming fourth wave of NSAM, currently in the field, will allow us to continue examining these trajectories into middle adulthood.

These analyses are intended to help health and other professionals design and evaluate effective strategies to reduce the risk of HIV and other STDs in young men. Our findings demonstrate the need for effective programs to consider a range of risk behaviors—rather than a single indicator—in both their design and their evaluation. The findings also show the need to conceptualize and measure risk behavior as a dynamic process that changes over a young man’s life course.

REFERENCES

Footnotes

*In the cluster analysis, we excluded 1991 data from two respondents, who reported extraordinarily high numbers of partners, as the inclusion of their responses would have unduly influenced the analyses.

For Wave 1, risky sex partners were onetime partners (regardless of how long the respondent had known the partner before the two had sex), sex workers, injection-drug users and male sex partners. For Waves 2 and 3, risky partners were all of these plus solicitors (if the respondent himself was a sex worker) and persons with HIV or AIDS.

§The k-means procedure, part of the STATA software package, is an iterative partitioning cluster analysis method that arbitrarily assigns cases into a specified number (k) of clusters (source: Aldenderfer MS and Blashfield RK, Cluster Analysis, Beverly Hills, CA: Sage Publications, 1984). The centroids of the clusters are computed, as are the Euclidean distances between the cases and the centroids. Cases are then moved from the cluster in which they were initially placed to the cluster with the nearest centroid. After reassignment, new centroids are computed and cases are again moved if there is a nearer centroid. This process continues until no more cases are reassigned.

††Initially, we created clusters separately for each wave. After examining descriptive statistics, however, we determined that the mean values for all five variables defining the clusters—and hence the meaning of risk for each cluster—changed over time. For example, youth classified as high-risk at Wave 1 had lower mean values on risk measures than did members of the high-risk cluster at Wave 2. Thus, what was considered high-risk in Wave 1 was considered less risky in Wave 2. To ensure that the mean values on the variables of interest were the same for each cluster over time, we created the clusters using observations across all three waves. This allowed us to avoid referring to men who were low-risk at Wave 2 as being high-risk at Wave 1.

In this article, “ever had an STD” and “had an STD in the past year” refer to STDs diagnosed by a physician, and do not include the results of the chlamydia tests conducted at Wave 3.

References

1. Centers for Disease Control and Prevention (CDC), Cases of HIV infection and AIDS in the United States and dependent areas, 2005, HIV/AIDS Surveillance Report, Vol. 17, rev. ed., Atlanta: U.S. Department of Health and Human Services and CDC, 2007.

2. Weinstock H, Berman S and Cates W, Jr, Sexually transmitted diseases among American youth: incidence and prevalence estimates, 2000, Perspectives on Sexual and Reproductive Health, 2004, 36(1):6–10.

3. National Institute of Allergy and Infectious Diseases (NIAID), Workshop Summary: Scientific Evidence on Condom Effectiveness for Sexually Transmitted Disease (STD) Prevention, Bethesda, MD: NIAID, 2001.

4. DiClemente RJ et al., Prevention and control of sexually transmitted infections among adolescents: the importance of a socio-ecological perspective—a commentary, Public Health, 2005, 119(9):825–836.

5. Kotchick BA et al., Adolescent sexual risk behavior: a multi-system perspective, Clinical Psychology Review, 2001, 21(4):493–519.

6. Abma JC and Sonenstein FL, Sexual activity and contraceptive practices among teenagers in the United States, 1988 and 1995, Vital and Health Statistics, 2001, Series 23, No. 21.

7. Division of STD Prevention, CDC, Sexually Transmitted Disease Surveillance 1999, Atlanta: CDC, 2000.

8. Hallfors DD et al., Sexual and drug behavior patterns and HIV and STD racial disparities: the need for new directions, American Journal of Public Health, 2007, 97(1):125–132.

9. Hussey JM et al., Sexual behavior and drug use among Asian and Latino adolescents: association with immigrant status, Journal of Immigrant and Minority Health, 2007, 9(2):85–94.

10. Halpern CT et al., Implications of racial and gender differences in patterns of adolescent risk behavior for HIV and other sexually transmitted diseases, Perspectives on Sexual and Reproductive Health, 2004, 36(6):239–247.

11. Johnston LD, O’Malley PM and Bachman J, Illicit Drug Use, Smoking, and Drinking by America’s High School Students, College Students, and Young Adults: 1975–1987, Rockville, MD: National Institute on Drug Abuse, 1988.

12. Substance Abuse and Mental Health Services Administration (SAMHSA), National Household Survey on Drug Abuse: Population Estimates 1992, Rockville, MD: SAMHSA, 1993.

13. Laumann EO et al., The Social Organization of Sexuality: Sexual Practices in the United States, Chicago: University of Chicago Press, 1994.

14. Ku L, Sonenstein FL and Pleck JH, Young men’s risk behaviors for HIV infection and sexually transmitted diseases, 1988 through 1991, American Journal of Public Health, 1993, 83(11):1609–1615.

15. Gottfredson MR and Hirschi T, A general theory of adolescent problem behavior: problems and prospects, in: Ketterlinus RD and Lamb ME, eds., Adolescent Problem Behavior: Issues and Research, Hillsdale, NJ: Lawrence Erlbaum Associates, 1994, pp. 41–56.

16. Billy JO et al., The sexual behavior of men in the United States, Family Planning Perspectives, 1993, 25(2):52–60.

17. Dolcini MM et al., Demographic characteristics of heterosexuals with multiple partners: the National AIDS Behavioral Surveys, Family Planning Perspectives, 1993, 25(5):208–214.

18. Barone C et al., High-risk sexual behavior among young urban students, Family Planning Perspectives, 1996, 28(2):69–74.

19. Pleck JH, Sonenstein FL and Ku L, Changes in adolescent males’ use of and attitudes towards condoms, 1988–1991, Family Planning Perspectives, 1993, 25(3):106–110 & 117.

20. Ku L, Sonenstein FL and Pleck JH, The dynamics of young men’s condom use during and across relationships, Family Planning Perspectives, 1994, 26(6):246–251.

21. Ku L and Kershaw J, Notes on attrition bias and weighting in Wave 2 of the National Survey of Adolescent Males, working paper, Washington, DC: Urban Institute, 1991.

22. Ku L, Martinez G and Williams S, Attrition and weighting for the third wave (1995) of the National Survey of Adolescent Males, working paper, Washington, DC: Urban Institute, 1996.

23. Ku L et al., Risk behaviors, medical care, and chlamydial infection among young men in the United States, American Journal of Public Health, 2002, 92(7):1140–1143.

24. Rindfuss RR, The young adult years: diversity, structural change, and fertility, Demography, 1991, 28(4):493–512.

Author's Affiliations

Jacinda K. Dariotis is assistant scientist, Department of Population, Family, and Reproductive Health; Freya L. Sonenstein is professor and director, Center for Adolescent Health, Department of Population, Family, and Reproductive Health; Nan M. Astone is associate professor, Department of Population, Family, and Reproductive Health; Frangiscos Sifakis is assistant scientist, Department of Epidemiology; and Scott Zeger is professor, Department of Biostatistics, and vice provost for research— all at The Johns Hopkins Bloomberg School of Public Health, Baltimore. Gary J. Gates is senior research fellow, Williams Institute, University of California, Los Angeles, School of Law. Randy Capps is senior research associate, Urban Institute, Washington, DC. Joseph H. Pleck is professor, Department of Human and Community Development, University of Illinois at Urbana-Champaign.

Disclaimer

The views expressed in this publication do not necessarily reflect those of the Guttmacher Institute.