We first examined data completeness. Missingness comprised <1% for sex (460/61,298 observations) and age (317/61,298), and 13.4% for SoSP (8,204/61,298). Significant predictors of greater SoSP missingness (Table S1) included male sex, lesbian/gay identity, being “not sure” of SI, declining to provide SI, and each survey year after 2015. No ages were statistically significant.

Figure 1 and Table S2 show the proportion of respondents with a given reported SoSP lifetime history, by sex and year. The proportion of all female respondents reporting sex with only females increased significantly (1.7% to 3.0%, in 2015 and 2021, respectively; p<0.001 for the full temporal trend), and with both males and females (5.7% to 9.2%, p<0.001), and declined in those reporting sex with only males (33.7% to 32.2%, p=0.023) or no prior sex (58.9% to 55.6%, p=0.002). There was a significant increase in males with no prior sex (52.7% to 57.5%, p<0 .001) and decrease in males having only female partners (44.2% to 38.7%, p<0 .001). The point estimate for the proportion of males having only male partners also increased (1.3% to 1.7%), although not significantly (p=0.053). In post-hoc analyses, we considered the proportion of males having sex with males (whether also with females or not), which increased significantly (3.1% to 3.8%, p<0.02); similarly, the proportion of females having sex with females (whether also with males or not), increased significantly (7.4% to 12.2%, p<0.001).

Conditioning on having ever had sex, there were again significant increases in females having female (4.0% to 6.7%, p<0.001) and both female and male partners (13.9% to 20.8%, p<0.001), and a significant decrease in only male partners (82.1% to 72.5%, p<0.001). The proportion of males having only male partners increased significantly (2.8% to 4.0% p=0.002), while the point estimate for the proportion having both male and female partners also increased (3.8% to 4.9%), although not significantly (p=0.052). Thus, in 2021 more than one-quarter of sexually experienced female respondents (27.5%) reported a history of same-sex sexual contact, while a smaller but still substantial proportion of sexually experienced male respondents (8.9%) did.

We next considered trends in respondents’ self-reported SI (Figure 2, Table S3). Female respondents showed significant declines in reporting heterosexual identity over time (82.6% in 2015 to 62.2% in 2021, p<0.001 across the entire time trend). This change mostly reflected increases in reporting bisexual (9.9% to 18.6%, p<0.001) or unsure identity (4.0% to 14.3%, p<0.001), although lesbian identity also increased significantly (2.2% to 3.7%, p<0.001). Males reported a much smaller but also significant decline in straight identity (92.3% to 87.0%, p<0.001), similarly mimicked by increases in bisexual (2.2% to 4.2%, p<0.001) and “not sure” responses (2.4% to 5.9%, p<0.001). The increase for gay identity was not significant (1.9% to 2.1%, p=0.21). We repeated the trend tests, removing respondents who answered “I describe my sexual identity some other way” while retaining the other two options in the “not sure” comparison (Table S4); the increase in “not sure” retained significance for both female and male respondents.

Figure 3 and Table S5 show the proportion of female respondents reporting each SoSP option for each SI response, including no sexual history. Female respondents identifying as straight or lesbian showed no significant shifts in SoSP. Bisexual females, in contrast, displayed a significant decrease in the proportion having both female and male partners (36.2% in 2015 to 30.6% in 2021, p=0.04), although these figures always reflected the majority of those with sexual history. The proportion of bisexual females with only male partners also increased significantly (17.4% to 22.9%, p=0.006). Females responding “not sure” and declining to respond were both more likely than their peers to have never had sex, although this declined significantly with time, while the proportion having sex with males only increased significantly.

Turning to male respondents (Figure 4, Table S5), straight-identified males showed a decline in the proportion having only female partners (46.5% to 41.8%, p<0.001), although this was due not to increases in male partners, but to those having no prior sex (52.7% to 57.5%, p<0.001). Gay males reported a significant decrease in the proportion with only female partners (9.8% to 2.0%, p=0.04), with a comparably sized, but not significant, change in those having no prior sex (37.6% to 47.3%, p=0.20).

When conditioning on having ever had sex, overall patterns of significant change in SoSP by SI remained largely similar for female respondents (Figure S1, Table S6); in contrast, the previously significant trends for straight and gay male respondents disappeared (Figure S2, Table S6). However, the point estimate for the proportion of gay males reporting only male partners increased by over 10 percentage points (66.3% in 2015 to 76.9% in 2021), and the proportion reporting only female partners decreased similarly (15.7% to 3.9%), despite neither achieving significance (p=0.10, and 0.08, respectively). We note that over 98% of sexually experienced straight males reported only female partners across all surveys, while >20% of gay males reported a prior female partner, either exclusively or along with male partners.

Finally, we disaggregated trends by respondent age (Figure 5; Table S7). For females, the proportion with no prior sex declined significantly over time for each age 14-17, with significant age-specific increases for having only female partners (ages 14, 15, 17), only male partners (ages 14, 15), and both (ages 14, 15, 16, 18). In contrast, males of specific ages showed almost no significant time trends, with one exception (increase in having both female and male partners for 16-year-old males). When conditioning on prior sex (Figure S3, Table S8), males showed the same paucity of significant effects as in unconditional tests. However, female respondents showed strong and significant shifts at even the youngest ages away from sex with males only and towards sex with either females only or both males and females. Indeed, the significant shift away from only male partners was replicated at every age, with effect sizes between 2015 and 2021 ranging from 5-10 percentage point reductions.

We examined recent (2015-2021) trends in sex of sex partners (SoSP) and sexual identity (SI), alone and in combination, among US adolescents. The proportion of females and males reporting prior same-sex sexual activity increased. The increase was even larger for males when measured as the proportion of those with any sexual experience. Previous research using 2005-2015 YRBS data found an increase for females but not males (10); our work suggests that the increase among females has continued, with males increasing as well. We also saw a significant decline in the proportion of female respondents with no sexual experience, reversing the prior analysis’s trends. These findings appear to contradict YRBS summary reports showing declines through 2021 in adolescents having sex (17); however, the summary report measure asks about sexual intercourse, whereas the one in our analyses asks about sexual contact. These constructs may not perfectly match, as intercourse does not encompass all forms of sexual contact and adolescents’ interpretations of each concept may vary. Similarly, the decrease in female heterosexual identification may reflect differences in survey timing, as 2021 data were collected in fall due to COVID-19, unlike previous years’ spring surveys. Data from the 2023 YRBS were not available when this research was conducted, but have recently become so; analyses of this and future waves will enable examination of continuation or change in the trends identified here.

SM identities increased over time, while straight identity decreased, matching findings from earlier YRBS waves (10) and other sources (12). Bisexual female identity saw the largest growth, followed by bisexual males and by lesbians, with no significant increase for gay males. This could portend a coming increase in STIs among young women, as multiple studies have found that bisexually identified women have the highest rates among women of both bacterial and viral STIs (27, 28). Young bisexual males similarly have higher rates of STIs than their heterosexual peers (28), and also exhibit unique risk profiles such as more cumulative and concurrent partners and younger sexual debut (8). The proportion responding some form of “not sure” also increased, although the exact effect is challenging to isolate given the expansion of relevant options in 2021. Reducing the comparison to two options that most clearly reflected being “not sure”—not understanding the question or questioning their sexual identity—still showed significant increases. Future studies would need to determine how young people who reported being unsure of their sexual identity vary from their peers in terms of STI burden.

Identity is of course only a proxy for STI risk, with the network of sexual relationships and behaviors in those relationships forming one determinant of the link between identity and risk. Our analysis of SoSP conditional on SI showed that same-sex sexual contact also increased, and was due primarily to the increase in youth with SM identities, i.e., heterosexual adolescents showed no significant increases in same-sex contact. However, as the proportion of the female population identifying as bisexual grew, the proportion of bisexual females reporting a history of sex with females and males declined. The former effect was larger than the latter, such that the proportion of all female respondents reporting bisexual identity and both male and female partners increased from 3.6% to 5.7%. Young women with both bisexual identity and behavior appear to have the highest rates of STIs among all young women (28), making this increase especially notable.

Both straight and gay males saw a decline in the proportion having sex with females only, although this was due solely to an increase in those having no prior sex. Nonetheless, the trend among gay males represents an absolute decline in those reporting a sexual history apparently incongruent with their stated identity. Although such a pattern may have many possible explanations—including sexual exploration or shifting identities—such incongruence has repeatedly been found to predict multiple forms of negative psychosocial health experiences (29-32). Thus, while increased male-male sex may have HIV/STI implications, the decrease in incongruence may portend a parallel increase in positive health outcomes. The magnitude of this effect will depend on the nature of the relevant causal web, however, which is only partly understood. Similarly, the decrease in incongruence may have beneficial implications for interpreting HIV/STI risk among adolescent males, as fewer would be misclassified by SI-based assessments.

Female increases in same-sex sex extended significantly across all observed ages from 14 to 17; the male pattern did not. Adolescent sexual health education may be more impactful if content delivery incorporates more information on female-female sexual contact, which appears to be increasing across the adolescent age range.

Our consideration of the absolute proportions of youth with each SoSP history was driven by an interest in the prevalence of partnerships associated with different HIV/STI risks. We also considered proportions after removing those who had never had sex, isolating the SoSP distribution among those with sexual experience. The two trends sometimes diverged in magnitude or significance, especially for males. For instance, while the proportion of all males having sex with males grew, its growth as a proportion of all those having had sex was even larger. If those who have not yet had sex end up shifting their SoSP distribution in the same direction as their peers, this pattern could portend even more rapid increases in male-male sex in the coming years, with implications for HIV/STI transmission, prevention, testing, and treatment. The accuracy of this assumption, for male-male sex or any other combination, is difficult to predict. If the shifts in SoSP occur in part from an overall change in cultural norms, we might expect to see similar patterns among those with both early and late sexual initiation. However, it is possible that known differences between those with early and late initiation (33), could lead to different magnitudes or even directions of behavioral shifts. The patterns here likely differ by sex. Examination of post-adolescent cohorts could test whether the relationship between early and late initiators for SoSP has changed over time.

Some measures exhibited large changes from 2015 to 2021, without significance. We see two key reasons for this. First, we conducted trend tests on the full set of surveys but only reported the 2015 and 2021 values in the text for simplicity; the figures and tables reveal cases in which the intervening years did not follow a consistent trend. Secondly, sample sizes varied considerably in the disaggregated analyses, being much smaller for SM groups than for heterosexual respondents (Tables S5-S6).

Our analysis has multiple limitations, many imposed by the available source data. Respondents were only given an “other” SI option in 2021, making direct comparison with earlier years difficult. The conceptualization of different identities might be shifting among youth (34, 35) in ways that are only partially captured by comparing prevalence of identities over time; additional qualitative information about respondents’ understanding of each choice would enhance interpretation of trends. The data preclude consideration of temporal changes in individuals’ SI and the ordering of identity formation and sexual history. We cannot distinguish motivations for different sexual partnerings, e.g., identity exploration, peer pressure, or coercive or forced sex, nor can we determine sex frequency within partnerships, which is especially relevant for HIV/STI risk. We did not analyze trends by race or ethnicity, despite known differences in SI and SoSP metrics (36) and HIV/STI burden (3), as the analysis was already complex; future work may re-examine our most salient findings more deeply by disaggregating on this important dimension.

Finally, sexual contact is a general term, which is purposefully left undefined in YRBS. It may thus be interpreted differently by different respondents, making the direct links to HIV/STI risks difficult to know confidently. A recent study among Australian adolescents found that someone’s definition of what constitutes the term “sex” is influenced by many factors, including sexual orientation, previous sexual experience, and cultural/religious context (37). “Sexual contact” is likely similar. It is thus possible that YRBS’s unspecified definition of sexual contact could introduce bias based on participant identity. However, this approach has the strength of allowing participants to define their own sexual experiences in terms of what they consider most salient.

Notably, SoSP missingness was high. Male respondents and gay/bisexual respondents were more likely than others not to report their SoSP, likely underestimating the prevalence of male-male sex. As the under-reporting increased with time, the actual increase in male-male sex may be greater than observed in these data.

STIs prevalences vary by people’s sex and that of their sex partners, including among adolescents (1-7). Meanwhile, sexual identity can shape one’s perception of STI risk, exposure to prevention messaging, and engagement with testing and treatment (7-9). The nature of sexual identity and sexual partnerships among US adolescents is becoming more diverse. This work finds that earlier-identified shifts have continued, and in some cases expanded. Public health and education in the US may anticipate potential shifts in HIV/STI incidence, while also meeting adolescents where they are at in order to provide the quality education, prevention and treatment services needed to maximize adolescent sexual health.