School-Based Protective Factors for HIV Prevention in the United States: Secondary Analysis of the Youth Risk Behavior Survey 2015–2019

Abstract

This secondary analysis of the National Youth Risk Behavior Survey (years 2015–2019) examines associations between school-based protective factors (i.e., safe school environments and academic achievement) and HIV risk behaviors among sexually experienced adolescent gay and bisexual men (AGBM; n = 644), a population with the highest prevalence of undiagnosed HIV infections. Demographics included Hispanics/Latinos (25%, n = 158), Other race/ethnicity (14%, n = 88), and non-Hispanic Blacks/African Americans (13%, n = 81). Adjusted models showed protective factors reduced odds for early sexual debut, multiple sexual partners, sex under the influence of drugs/alcohol, and condomless sex, with an additive effect demonstrated when two protective factors were present. Hispanics/Latinos had greater odds of reporting multiple sexual partners and HIV testing, indicating opportunities for school-based HIV prevention and further research. Our findings provide support for school-based programs that aim to improve social and structural determinants of health and ultimately reduce adolescent HIV burdens.

Adolescent gay, bisexual, and other men who have sex with men (AGBM) aged 13–24 years exhibit the highest prevalence of undiagnosed HIV infections in the United States, with an estimated 56% unaware of their HIV seropositive status in 2021 (Centers for Disease Control and Prevention, 2023b). Additionally, HIV disparities persist among Black/African American and Hispanic/Latino AGBM, two key populations that are less frequently diagnosed, linked into care, and virally suppressed (Jeffries et al., 2020)—elements necessary for reducing morbidity, mortality, and onward HIV transmission. As such, there is a critical need for innovative approaches to adolescent HIV prevention for a reduction in new infections by 90% by the year 2030, an important benchmark in the United States’ initiative for Ending the HIV Epidemic (EHE; United States Department of Health and Human Services, 2022).

Identifying protective factors to strengthen resilience during adolescence and reduce HIV risk behaviors is one key strategy to improve HIV prevention, as much work has been done to identify risk (Garcia, Fletcher, Goldsamt, & Dunn Navarra, 2023). Protective factors are defined as characteristics, conditions, and behaviors that improve health outcomes or reduce the negative effects of risk factors, while resilience is described as the process in which individuals experience difficult situations and draw on protective factors to achieve success and overcome obstacles, such as stigma related to sexual orientation (Johns et al., 2019).

Generally, adolescence is a developmentally rich period of learning and growth that occurs between the ages of 10–24 years (National Academies of Sciences, Engineering, and Medicine, 2019; Twenge & Park, 2019). This is also a time when many adolescents, regardless of sexual orientation, make their sexual debut (i.e., engage in sexual activity with their first partner) (Vasilenko et al., 2021). Sexual expression is a natural occurrence during adolescence (Needham, 2021). However, for some AGBM, this period can be overshadowed by multiple co-occurring syndemic factors, such as substance use, violence, and mental health conditions that are known to increase the likelihood of HIV risk behaviors in what is known as an HIV syndemic (Garcia, Fletcher, Goldsamt, Bell, et al., 2023). Mounting research shows that protective factors can mitigate adverse experiences and support the development of positive trajectories later in life (National Academies of Sciences, Engineering, and Medicine, 2019). This is particularly true for school-based protective factors, as educational systems in the United States attend to the social, physical, and intellectual development of over 56 million students (Wilkins et al., 2022). These systems also provide important sexual health education to students, parents, and caregivers. Additionally, partnerships and collaboratives between community members and professional school staff facilitate the delivery of important health and social services. Such services foster youth development by strengthening connections between peers, adults, and communities.

The importance of school connectedness during adolescence has emerged as a key protective factor critical to health and wellbeing (McCabe et al., 2022). It is defined as the extent to which students feel accepted, included, and cared for at school (Marsh et al., 2019). Safe school environments and academic achievement are additional school-based protective factors that have been found to be interrelated with school connectedness (Duke, 2020; Wilkins et al., 2023). For AGBM, these protective factors are associated with reduced substance use, violence, mental health conditions, and suicidal behaviors (De Pedro et al., 2017; Ethier et al., 2018; Kidd et al., 2018; Marraccini, 2017; Wilson et al., 2018). However, current research is limited on the associations between these school-based protective factors and HIV risk behaviors among AGBM (Garcia, Fletcher, Goldsamt, & Dunn Navarra, 2023), such as early sexual debut, multiple sexual partners, sex under the influence of drugs or alcohol, condomless sex, and infrequent HIV testing (Jeffries et al., 2018).

With the availability of large secondary datasets, such as the Youth Risk Behavior Survey (YRBS) conducted by the Centers for Disease Control and Prevention, nationally representative data can be utilized to generate much-needed evidence regarding adolescent health-related behaviors. Since 1991, the YRBS has allowed for the biennial surveillance of important health-related behaviors and experiences among adolescents enrolled in public and private high schools, with the addition of survey items for sexual orientation beginning in 2015 (Underwood et al., 2020). More than five million students have participated in the YRBS, facilitating the surveillance of national progress toward many health objectives and goals, such as the EHE initiative. Therefore, the aim of this secondary data analysis was to examine associations between protective factors (safe school environments and academic achievement) and HIV risk behaviors among sexually experienced AGBM, using nationally representative YRBS data (years 2015–2019).

Methods

Design

This secondary data analysis was conducted between January 2021–April 2022 and utilized cross-sectional data from the National YRBS years 2015–2019, as these survey time points included items for sexual orientation. Based on deidentified publicly available data, this secondary data analysis received an exemption from the Institutional Review Board.

Primary Data Source

Sample.

The National YRBS sampling frame consists of public and private high schools that educate students in grades 9–12 and are located in the United States and the District of Columbia (Underwood et al., 2020). Excluded from the sampling frame were (1) alternative instruction, special education, and vocational schools, (2) schools with fewer than 40 students, and (3) schools operated by the Department of Defense or Bureau of Indian Education. A nationally representative sample was developed using a three-stage cluster design. Student participation was voluntary and required signed parental consent. The anonymous 99-item survey was self-administered using a computer-scannable booklet in a 45-minute class period. All deidentified data are publicly available for download online (Centers for Disease Control and Prevention, 2023c).

Secondary Data Analysis

Sample.

Inclusion criteria for the analytic sample of sexually experienced AGBM men were based on the following survey items: (1) male sex assigned at birth (What is your sex? Female = 0 or Male = 1); (2) history of sexual intercourse (Have you ever had sexual intercourse? No = 0 or Yes = 1); and (3) self-identification as gay, bisexual, unsure/no answer, and/or ever having same-sex partners (Combined items: Which of the following best describes you? and During your life, with whom have you had sexual contact? Heterosexual with opposite sex partners = 0, Heterosexual with same sex partners = 1, Unsure or No Answer = 2, Bisexual identifying = 3, Gay identifying = 4). The final analytic sample included adolescents with a male sex assigned at birth, a history of sexual intercourse, and a sexual orientation other than heterosexual with opposite sex partners.

Variables of interest.

Variables that emerged from the literature were discussed by a multidisciplinary team of HIV researchers and clinicians representing nursing, medicine, psychology, and biostatistics. Selected variables were matched to survey items available in the primary data source. The dichotomization of answer choices from each item was based on the primary data source, extant literature, sufficient cell size, and/or to facilitate age-based comparisons. The final operationalization and coding are as follows.

Protective factors.

Two survey items from the primary data source measured school-based protective factors, and these were used as independent variables in this secondary data analysis. A safe school environment was operationalized using the survey item: During the past 30 days, on how many days did you not go to school because you felt you would be unsafe at school or on your way to or from school? 1 day or more (i.e., did not go to school and felt unsafe) = 0 or 0 days (i.e., went to school and felt safe) = 1. Academic achievement was operationalized using the survey item: During the past 12 months, how would you describe your grades in school? Mostly D’s, F’s, None of these, Not Sure = 0 or Mostly A’s, B’s, C’s = 1. A categorical composite score variable for protective factors was also developed (No protective factors = 0, One protective factor = 1, or Two protective factors = 2).

HIV risk behavior outcomes.

Six survey items from the primary data source measured known HIV risk behaviors. These were operationalized as outcomes (i.e., dependent variables) in this secondary data analysis while maintaining the original dichotomization of the answer choices from the primary data source. The six HIV risk behavior outcomes included: (1) early sexual debut (How old were you when you had sexual intercourse for the first time? After the age of 13 years = 0 or Before the age of 13 years = 1); (2) multiple sexual partners (During your life, with how many people have you had sexual intercourse? Three persons or less = 0 or Four persons or more = 1); (3) recently having sexual intercourse (During the past three months, with how many people did you have sexual intercourse? No intercourse during the past three months = 0 or One or more persons = 1); (4) sex under the influence of drugs or alcohol (Did you drink alcohol or use drugs before you had sexual intercourse the last time? No = 0 or Yes = 1); (5) condom use (The last time you had sexual intercourse, did you or your partner use a condom? No = 0 or Yes = 1); (6) HIV testing (Have you ever been tested for HIV, the virus that causes AIDS? (Do not count tests done if you donate blood.) No = 0 or Yes = 1).

Demographic covariates.

Demographic covariates used in this secondary data analysis included age, race, and ethnicity. Age (How old are you? ≤ 15 years old = 0 or ≥ 16 years old = 1) was dichotomized based on epidemiological data that indicate HIV infections increase after the age of 15 years when many adolescents begin sexual activity (US Preventive Services Task Force, 2019). Race and ethnicity were categorized using four categories that were available in the primary data source (non-Hispanic White = 0, non-Hispanic Black/African American = 1, Hispanic/Latino = 2, or Other race/ethnicity (Asian, Native Hawaiian, other Pacific Islander, American Indian, Alaska Native, and Multiple Races (non-Hispanic)) = 3).

Statistical Analysis

Stata version 17 BE (StataCorp, 2021) was used for data cleaning, recoding, and statistical analysis in consultation with a biostatistician with expertise in public health and large datasets. To account for the complex sampling design, sampling weights for sex, race, ethnicity, and grade were applied to adjust for the nonresponse and oversampling of non-Hispanic Black/African American and Hispanic/Latino participants (Underwood et al., 2020). Using statistical weights also ensured that weighted counts were equal to the total sample size and that weighted proportions for each grade level matched national population proportions. As such, a nationally representative sample of all students attending public and private high schools in the United States in grades 9–12 was developed. For comparison of the results of this study with age-based HIV epidemiological data, multivariate logistic regression models were adjusted for age rather than grade level. To prevent additional forms of bias, missing data were not statistically imputed, consistent with the primary data source (Underwood et al., 2020) and previous YRBS secondary data analyses (Clayton et al., 2019).

Univariate analysis.

Individual indicators were analyzed at the categorical level. The distributions within the categorical levels were examined to ensure they were balanced, and categories with insufficient cell sizes were combined. Independent variables (i.e., protective factors) were examined for multicollinearity, which was not present. Descriptive statistics (i.e., weighted frequencies, percentages, means (M), standard deviations (SD), and ranges) were used to describe demographic covariates, protective factors, and HIV risk behavior outcomes.

Bivariate and multivariate logistic regression analyses.

Bivariate logistic regression analysis was used to analyze independent associations between the six HIV risk behavior outcomes and (1) demographic covariates, and (2) the two protective factors. Multivariate logistic regression analysis was then used to examine associations, at a level of significance of p < 0.05, between the six HIV risk behaviors and the two protective factors while adjusting for age, race, ethnicity, and sexual orientation. Finally, to examine for an additive effect associated with the incremental addition of a protective factor, the categorical composite score variable for protective factors was substituted into the adjusted multivariate logistic regression models (i.e., no protective factors, one protective factor, or two protective factors).

Results

Univariate Analysis

The demographic characteristics of the analytic sample of sexually experienced AGBM (n = 644) are described in Table 1. The majority of AGBM were aged 16 years and older (n = 474, 74%), with 35% (n = 226) identifying as bisexual, 28% (n = 183) identifying as unsure or having no answer, and 23% (n = 147) identifying as gay. The racial and ethnic diversity of the analytic sample included those categorized as Hispanics/Latinos (25%, n = 158), Other race/ethnicity (14%, n = 88), and non-Hispanic Blacks/African Americans (13%, n = 81).

Table 1.

Demographic Covariates

	Analytic Samplea n = 644
	n b	%
Age
≤ 15 years	167	26%
≥ 16 years	474	74%
Race/Ethnicity
White, Non-Hispanic	317	49%
Black/African American, Non-Hispanic	81	13%
Hispanic/Latino	158	25%
Other Race/Ethnicityc	88	14%
Sex d
Male	644	100%
Sexual Orientation
Heterosexual, with Same Sex Partners	88	14%
Unsure or No Answere	183	28%
Bisexual Identifying	226	35%
Gay Identifying	147	23%

^aThe analytic sample was restricted to those with a male sex assigned at birth, history of sexual intercourse, sexual orientation other than heterosexual with opposite sex partners, and participation in YRBS (years 2015–2019).

^bWeighted frequencies.

^cThe Other race or ethnicity category combined Asian, Native Hawaiian, other Pacific Islander, American Indian, Alaska Native, and Multiple Races (non-Hispanic).

^dSurvey item for sex assigned at birth = female or male. Gender identity or intersex is not available in the dataset.

^eThose who answered the survey item for sexual orientation as unsure or did not answer were combined into one category.

Individual indicators measuring protective factors and HIV risk behavior outcomes are described in Table 2. The following paragraphs summarize the high and low prevalence ranges that were observed.

Table 2.

Univariate Analysis – Protective Factors and HIV Risk Behavior Outcomes

	M	SD	Range
Protective Factors Composite	1.62	0.53	0–2
	% Yes	n a	Totalb
Protective Factors, Composite
No protective factors	4%	23	604
1 protective factor	30%	181	604
2 protective factors	66%	400	604
Protective Factors, Individual
Safe school environment	82%	518	628
Academic achievement	80%	491	614
HIV Risk Behavior Outcomes
Recent sexual intercourse	65%	406	629
Condom use	52%	204	392
Multiple sexual partners	33%	208	636
Sex under the influence of drugs or alcohol	32%	124	391
Early sexual debut	24%	151	632
HIV testing	21%	130	622

Note. SD = Standard Deviation.

^aWeighted frequencies.

^bTotals for each variable vary due to incomplete data across survey items.

Protective factors.

Those in the analytic sample exhibited a mean of 1.62 protective factors (SD = 0.53, range = 0–2). Overall, 66% (n = 400) reported two protective factors, 30% (n = 181) reported one protective factor, and 4% (n = 23) reported no protective factors. Individual indicators showed 82% (n = 518) reported a safe school environment, and 80% (n = 491) reported academic achievement.

HIV risk behavior outcomes.

Individual indicators for HIV risk behavior outcomes had prevalence ranges between 65% (n = 406) for those who reported recent sexual intercourse and 21% (n = 130) for those who reported previous HIV testing. Condom use was reported by one-half (52%, n = 204) of the analytic sample.

Adjusted Multivariate Logistic Regression Analyses

Associations between protective factors and HIV risk behaviors.

Significant associations were identified between the two protective factors (safe school environments and academic achievement) and five HIV risk behavior outcomes (early sexual debut, having multiple sexual partners, sex under the influence of drugs or alcohol, condom use, and HIV testing). The adjusted model for the outcome of recent sexual intercourse was not statistically significant (p = 0.37). A display of the results is shown in Table 3. The following paragraphs describe the significant findings that were observed.

Table 3.

Bivariate and Multivariate Logistic Regression Analysis – Associations Between Protective Factors and HIV Risk Behaviors

	Early Sexual Debut		Multiple Sexual Partners		Recent Sexual Intercourse		Sex Under the Influence of Drugs or Alcohol		Condom Use		HIV Testing
	OR [95% CI]	AOR [95% CI]	OR [95% CI]	AOR [95% CI]	OR [95% CI]	AOR a [95% CI]	OR [95% CI]	AOR [95% CI]	OR [95% CI]	AOR [95% CI]	OR [95% CI]	AOR [95% CI]
Academic Achievement	0.41**	0.49*	0.44**	0.54	0.58	0.56	0.44*	0.54	2.41*	2.68*	0.44*	0.42*
	[0.21,0.79]	[0.26,0.92]	[0.24,0.82]	[0.29,1.02]	[0.27,1.24]	[0.25,1.26]	[0.19,0.99]	[0.23,1.29]	[1.01,5.75]	[1.15,6.22]	[0.22,0.88]	[0.20,0.89]
Safe School Environment	0.37**	0.44*	0.35**	0.39**	0.67	0.77	0.35*	0.35*	1.88	1.62	0.78	0.75
	[0.18,0.74]	[0.19,1.00]	[0.19,0.66]	[0.21,0.72]	[0.35,1.25]	[0.40,1.51]	[0.15,0.84]	[0.14,0.88]	[0.90,3.92]	[0.71,3.69]	[0.39,1.55]	[0.36,1.59]
Age ≤ 15 years	Reference	Reference	Reference	Reference	Reference	Reference	Reference	Reference	Reference	Reference	Reference	Reference
Age ≥ 16 years	0.34***	0.32***	0.93	1.18	1.04	1.19	0.89	1.12	2.49**	2.01*	1.77	1.87
	[0.20,0.60]	[0.17,0.61]	[0.52,1.65]	[0.66,2.10]	[0.65,1.67]	[0.73,1.93]	[0.43,1.82]	[0.56,2.26]	[1.33,4.67]	[1.04,3.86]	[0.87,3.59]	[0.85,4.12]
White, Non-Hispanic	Reference	Reference	Reference	Reference	Reference	Reference	Reference	Reference	Reference	Reference	Reference	Reference
Black/African American, Non-Hispanic	0.74	0.63	1.17	1.30	1.49	1.55	0.60	0.69	1.10	1.14	1.10	1.31
	[0.33,1.63]	[0.22,1.84]	[0.59,2.32]	[0.63,2.66]	[0.69,3.22]	[0.68,3.50]	[0.22,1.61]	[0.24,1.95]	[0.39,3.06]	[0.35,3.69]	[0.41,2.96]	[0.50,3.48]
Hispanic/Latino	1.88	1.57	3.11***	2.72**	1.50	1.41	1.59	1.47	0.67	0.99	2.42**	2.34*
	[0.98,3.59]	[0.76,3.25]	[1.79,5.38]	[1.46,5.05]	[0.87,2.60]	[0.77,2.57]	[0.78,3.25]	[0.71,3.04]	[0.34,1.33]	[0.48,2.04]	[1.29,4.53]	[1.13,4.88]
Other Race/Ethnicity	1.56	1.44	1.69	1.97*	0.65	0.55	0.71	0.72	1.15	1.02	1.21	0.96
	[0.71,3.42]	[0.61,3.38]	[0.88,3.27]	[1.02,3.78]	[0.33,1.28]	[0.26,1.20]	[0.27,1.87]	[0.25,2.05]	[0.48,2.75]	[0.44,2.36]	[0.46,3.15]	[0.28,3.27]
Heterosexual, with Same Sex Partners	Reference	Reference	Reference	Reference	Reference	Reference	Reference	Reference	Reference	Reference	Reference	Reference
Unsure or No Answer	0.87	0.75	2.20*	1.87	1.33	1.20	1.50	0.98	0.97	2.08	0.81	0.85
	[0.35,2.18]	[0.26,2.18]	[1.05,4.61]	[0.83,4.23]	[0.63,2.81]	[0.57,2.54]	[0.57,3.97]	[0.35,2.71]	[0.37,2.55]	[0.71,6.06]	[0.35,1.90]	[0.35,2.10]
Bisexual Identifying	0.48	0.49	1.20	1.14	1.08	1.11	0.95	0.68	1.15	1.50	0.93	1.20
	[0.20,1.15]	[0.19,1.31]	[0.51,2.86]	[0.47,2.78]	[0.53,2.19]	[0.55,2.27]	[0.33,2.78]	[0.24,1.92]	[0.51,2.59]	[0.59,3.81]	[0.35,2.47]	[0.43,3.29]
Gay Identifying	0.51	0.42	1.11	0.88	1.11	1.09	0.32*	0.19**	0.94	1.15	1.46	1.84
	[0.19,1.36]	[0.14,1.22]	[0.50,2.45]	[0.37,2.09]	[0.58,2.12]	[0.56,2.11]	[0.11,0.91]	[0.06,0.55]	[0.36,2.43]	[0.41,3.24]	[0.56,3.82]	[0.68,5.00]

Note. The analytic sample was restricted to those with a male sex assigned at birth, history of sexual intercourse, sexual orientation other than heterosexual with opposite sex partners, and participation in YRBS (years 2015–2019). Adjusted models controlled for age, race/ethnicity, and sexual orientation. OR = Odds Ratio; AOR = Adjusted Odds Ratio; 95% CI = 95% Confidence Interval.

^aAdjusted multivariate model for Recent Sexual Intercourse was not statistically significant, p = 0.37.

Significance levels:

^*p < .05,

^**p < .01,

^***p < .001.

Outcome 1: Early sexual debut.

Those who reported a safe school environment were less likely to report having an early sexual debut (aOR = 0.44, 95% CI [0.19,1.00], p < .05) compared with those who did not report a safe school environment. Similarly, those who reported academic achievement were less likely to report having an early sexual debut (aOR = 0.49, 95% CI [0.26,0.92], p < .05) compared with those who did not report academic achievement. Compared with those who were aged 15 years or younger, those aged 16 years and older were less likely to report an early sexual debut (aOR = 0.32, 95% CI [0.17,0.61], p < .001).

Outcome 2: Multiple sexual partners.

Those who reported a safe school environment were less likely to report having multiple sexual partners (aOR = 0.39, 95% CI [0.21,0.72], p < .01) compared with those who did not report a safe school environment. However, those who were categorized as Hispanic/Latino or Other race/ethnicity were more likely to report having multiple sexual partners (aOR = 2.72, 95% CI [1.46,5.05], p < .01; aOR = 1.97, 95% CI [1.02,3.78], p < .05, respectively) compared with those who were categorized as non-Hispanic White.

Outcome 3: Sex under the influence of drugs or alcohol.

Those who reported a safe school environment were less likely to report sex under the influence of drugs or alcohol (aOR = 0.35, 95% CI [0.14,0.88], p < .05) compared with those who did not report a safe school environment. Those who identified as gay were also less likely to report sex under the influence of drugs or alcohol (aOR = 0.19, 95% CI [0.06,0.55], p < .01) compared with those who identified as heterosexual and had same-sex partners.

Outcome 4: Condom use.

Those who reported academic achievement were more likely to report condom use (aOR = 2.68, 95% CI [1.15,6.22], p < .05) compared with those who did not report academic achievement. Those aged 16 years or older were also more likely to report condom use (aOR = 2.01, 95% CI [1.04,3.86], p < .05) compared with those aged 15 years and younger.

Outcome 5: HIV testing.

Those who reported academic achievement were less likely to report HIV testing (aOR = 0.42, 95% CI [0.20,0.89], p < .05) compared with those who did not report academic achievement. However, those who were categorized as Hispanic/Latino were more likely to report HIV testing (aOR = 2.34, 95% CI [1.13,4.88], p < .05) compared with those who were categorized as non-Hispanic White.

Additive effect by number of protective factors experienced.

Significant associations were found based on the number of protective factors experienced (no protective factors, one protective factor, or two protective factors) and four HIV risk behavior outcomes (early sexual debut, having multiple sexual partners, sex under the influence of drugs or alcohol, and condom use); however, there were no associations with the outcome of HIV testing. The adjusted model for the outcome of recent sexual intercourse was not statistically significant (p = 0.14). The results are shown in Table 4, and the significant findings are summarized as follows.

Table 4.

Multivariate Logistic Regression Analysis – Additive Effect By Number of Protective Factors Experienced

	Early Sexual Debut	Multiple Sexual Partners	Recent Sexual Intercourse	Sex Under the Influence of Drugs or Alcohol	Condom Use	HIV Testing
	AOR [95% CI]	AOR [95% CI]	AOR a [95% CI]	AOR [95% CI]	AOR [95% CI]	AOR [95% CI]
Protective Factors (Range 0–2)
No Protective Factors	Reference	Reference	Reference	Reference	Reference	Reference
1 Protective Factor	0.52	0.14*	0.07	0.20*	2.59	0.65
	[0.23,1.18]	[0.03,0.71]	[0.01,0.61]	[0.04,0.95]	[0.60,11.26]	[0.19,2.27]
2 Protective Factors	0.23**	0.08**	0.06	0.11**	5.16*	0.34
	[0.10,0.56]	[0.02,0.43]	[0.01,0.50]	[0.02,0.54]	[1.34,19.85]	[0.10,1.16]
Age ≤ 15 years	Reference	Reference	Reference	Reference	Reference	Reference
Age ≥ 16 years	0.32***	1.13	1.22	1.08	1.96*	1.95
	[0.17,0.59]	[0.62,2.07]	[0.75,1.98]	[0.53,2.23]	[1.04,3.69]	[0.89,4.30]
White, Non-Hispanic	Reference	Reference	Reference	Reference	Reference	Reference
Black/African American, Non-Hispanic	0.64	1.31	1.52	0.67	1.17	1.28
	[0.22,1.85]	[0.64,2.66]	[0.67,3.45]	[0.23,1.91]	[0.37,3.70]	[0.49,3.37]
Hispanic/Latino	1.58	2.74**	1.38	1.53	1.01	2.31*
	[0.77,3.23]	[1.45,5.19]	[0.76,2.52]	[0.72,3.23]	[0.48,2.15]	[1.13,4.73]
Other Race/Ethnicity	1.41	1.94*	0.59	0.73	1.00	1.02
	[0.58,3.41]	[1.02,3.72]	[0.29,1.22]	[0.26,2.07]	[0.43,2.31]	[0.31,3.30]
Heterosexual, with Same Sex Partners	Reference	Reference	Reference	Reference	Reference	Reference
Unsure or No Answer	0.74	1.82	1.19	0.91	1.98	0.89
	[0.26,2.16]	[0.79,4.18]	[0.57,2.51]	[0.31,2.62]	[0.69,5.67]	[0.37,2.17]
Bisexual Identifying	0.49	1.18	1.14	0.69	1.49	1.17
	[0.18,1.32]	[0.48,2.94]	[0.56,2.29]	[0.24,2.00]	[0.58,3.81]	[0.42,3.27]
Gay Identifying	0.42	0.91	1.12	0.20**	1.14	1.82
	[0.14,1.21]	[0.38,2.17]	[0.58,2.16]	[0.07,0.58]	[0.40,3.24]	[0.67,4.96]

Note. The analytic sample was restricted to those with a male sex assigned at birth, history of sexual intercourse, sexual orientation other than heterosexual with opposite sex partners, and participation in YRBS (years 2015–2019). Adjusted models controlled for age, race/ethnicity, and sexual orientation. OR = Odds Ratio; AOR = Adjusted Odds Ratio; 95% CI = 95% Confidence Interval.

^aAdjusted multivariate model for Recent Sexual Intercourse was not statistically significant, p = 0.14.

Significance levels:

^*p < .05,

^**p < .01,

^***p < .001.

Outcome 1: Early sexual debut.

Compared with those who reported experiencing no protective factors, those who reported experiencing two protective factors had a decreased likelihood of reporting an early sexual debut (aOR = 0.23, 95% CI [0.10,0.56], p < .01). Those who were aged 16 years and older were less likely to report this outcome (aOR = 0.32, 95% CI [0.17,0.59], p < .001) compared with those who were aged 15 years and younger.

Outcome 2: Multiple sexual partners.

When there was one protective factor experienced, the likelihood of reporting multiple sexual partners decreased (aOR = 0.14, 95% CI [0.03,0.71], p < .05), and the likelihood further decreased when there were two protective factors experienced (aOR = 0.08, 95% CI [0.02,0.43], p < .01), when compared with those who reported experiencing no protective factors. Those who were categorized as Hispanic/Latino (aOR = 2.74, 95% CI [1.45,5.19], p < .01) or Other race/ethnicity (aOR = 1.94, 95% CI [1.02,3.72], p < .05) were more likely to report this outcome when compared with those categorized as non-Hispanic White.

Outcome 3: Sex under the influence of drugs or alcohol.

Those who reported experiencing one protective factor were less likely to report sex under the influence of drugs or alcohol (aOR = 0.20, 95% CI [0.04,0.95], p < .05), and the likelihood was further reduced among those who reported experiencing two protective factors (aOR = 0.11, 95% CI [0.02,0.54], p < .01), compared with those who reported experiencing no protective factors. Those who identified as gay were less likely to report this outcome (aOR = 0.20, 95% CI [0.07,0.58], p < .01) when compared with those who identified as heterosexual with same sex partners.

Outcome 4: Condom use.

Those who reported experiencing two protective factors were more likely to report condom use (aOR = 5.16, 95% CI [1.34,19.85], p < .05) compared with those who reported experiencing no protective factors. Those who were aged 16 years and older were more likely to report this outcome (aOR = 1.96, 95% CI [1.04,3.69], p < .05), when compared with those who were aged 15 years and younger.

Outcome 5: HIV testing.

Although models examining experiences with multiple protective factors were not associated with HIV testing, those who were categorized as Hispanic/Latino were significantly more likely to report HIV testing (aOR = 2.31, 95% CI [1.13,4.73], p < .05). This was in comparison with those who were categorized as non-Hispanic White.

Discussion

Using nationally representative YRBS data (years 2015–2019), the findings from this secondary data analysis are among the first to examine the associations between protective factors (safe school environments and academic achievement) and HIV risk behaviors among sexually experienced AGBM, a population that exhibits the highest prevalence of undiagnosed HIV infections (Centers for Disease Control and Prevention, 2023b). These findings lend to a greater understanding of the interrelation between safe school environments and academic achievement as highly significant protective factors in the context of adolescent HIV prevention, providing much needed evidence for future studies and interventions.

A summary of our study findings in the context of recent literature is provided below, followed by a discussion on recommendations for strengthening safe school environments and academic achievement for adolescent HIV prevention. More specifically, the policy implications of our findings for educational settings are discussed. This entails support for a school-based model that incorporates sexual health education and services, professional development, and positive youth development, as these programs are described in the evidence base as supportive of healthy adolescent behaviors and improved HIV prevention (Wilkins et al., 2022).

Protective Factor Associations

The school-based protective factors examined in this secondary data analysis were associated with a reduced likelihood of multiple HIV risk behaviors. The results from our study showed that those who reported having a safe school environment were less likely to report an early sexual debut, multiple sexual partners, and sex under the influence of drugs or alcohol. Similarly, those who reported academic achievement were also less likely to report having an early sexual debut, as well as condomless sex. Additionally, multivariate logistic regression models demonstrated an additive protective effect when these individual protective factors were aggregated into a categorical composite score (i.e., no protective factors, one protective factor, or two protective factors). For example, compared to those who did not experience any protective factors, those who experienced one protective factor had lower odds of reporting multiple sexual partners and sex under the influence of drugs or alcohol. In comparison, those who experienced two protective factors had even lesser odds.

Prevalence of Protective Factors

Prevalence estimates from our study showed that 82% of AGBM reported a safe school environment, while 80% reported academic achievement. This is similar to samples in other studies in which 85% reported a safe school environment (Kaczkowski et al., 2022) and 87% reported academic achievement (Hazel et al., 2019). However, the samples examined in these comparison studies aggregated lesbians with AGBM, which may account for the slightly higher prevalence estimates. There are likely many factors that contributed to the observed resilience of those in our sample, which were not measured in the primary data source utilized in this secondary data analysis. One such explanation is social support within the academic setting that enabled students to achieve academic success and overcome obstacles, such as stigma from sexual orientation. Nevertheless, 18% of AGBM in our sample reported missing school because they felt unsafe, and 20% reported unsatisfactory academic achievement. This is congruent with existing research that indicates AGBM exhibit disproportionate experiences with violence and victimization at school compared to heterosexual adolescents, contributing to unsafe school environments and decreases in academic achievement (Fields & Wotipka, 2022).

Additionally, studies have shown that AGBM have lower odds of graduating high school and attending college, stemming from experiences with discrimination, fewer educational expectations, and less of a sense of school belonging (Sansone, 2019). Similarly, chronic absenteeism has been associated with less educational attainment and success (i.e., dropping out of school, delayed graduation), poorer health, substance use, and risky sexual behaviors (Johnson et al., 2023). Moreover, high school graduation is a known predictor of adult success and is associated with better employment opportunities, adult physical and mental health, and less criminal justice involvement (Hoagwood et al., 2023).

However, the majority in our sample reported a safe school environment, a factor that has been associated with school connectedness (Rose et al., 2022), which in turn has been associated with decreased substance use (De Pedro et al., 2017; Ethier et al., 2018), violence (Basile et al., 2020), suicide (Eisenberg et al., 2021), and HIV risk behaviors (Ethier et al., 2018), as well as academic achievement (Baams et al., 2020), lending support to our overall findings.

Sexual Health Education and Services

Safe school environments also have implications for helping students receive effective sexual health education and adolescent HIV prevention (Centers for Disease Control and Prevention, 2019). This is because when students are absent from school due to feeling unsafe, they are less likely to be provided with vital information on sexual health education and services.

Additionally, a systematic review of high-quality sexual health education curricula, inclusive of sexual diversity (i.e., various sexual orientations and gender identities), found that these curricula improved experiences with homophobia and homophobic bullying (i.e., bullying victimization based on perceived sexual orientation or gender identity), safety and school attendance, adverse mental health, and academic achievement (Goldfarb & Lieberman, 2021). Similarly, these curricula are protective in decreasing the number of sexual partners, the use of drugs or alcohol before sex, and dating and intimate partner violence. It is also worth noting that sexual health education has been shown to be most effective when started early, before sexual debut, and before negative attitudes regarding differences in sexual orientation and gender identity are developed.

However, not all forms of sexual health education are associated with safer school environments, specifically curricula that are non-inclusive of sexual diversity, as these have been shown to increase homophobic bullying and adverse mental health (Rabbitte, 2020). As such, curricula such as the National Sex Education Standards exist to foster healthy adolescent sexual development and growth into adulthood (Future of Sex Education Initiative, 2020).

Sexual health education is also important for facilitating parent-adolescent communication, parental monitoring, and parental engagement in schools, all of which are associated with decreasing many adolescent health risks (Wilkins et al., 2022). This is particularly true when sexual health education programs provide homework assignments so students can engage with their parents and caregivers. For families and caregivers of sexually diverse adolescents who do not have access to relevant school-based training materials, several resources are available online (Family Acceptance Project, n.d.; Los Angeles LGBT Center, 2023).

Professional Development

Evidence shows that strengthening safe school environments is achievable through annual professional development of school staff, particularly regarding classroom management, as effective classroom management is associated with higher levels of school connectedness (Centers for Disease Control and Prevention, 2019). Additionally, classroom management is positively associated with motivation for learning and academic achievement (Adedigba & Sulaiman, 2020).

Professional development in supporting sexually diverse students is also recommended to increase school connectedness (Centers for Disease Control and Prevention, 2019). This includes education on language, health risks, effects of a negative school environment, supportive policies, unique needs, and bystander intervention skills. Supportive educators, trained to intervene effectively when homophobic bullying takes place, have been found to have a strong effect on academic success and decrease the occurrence of student harassment, assaults, and absenteeism (Johns et al., 2019). Additionally, educators are more likely to be proactive in developing safe and supportive environments when they have received professional development that includes topics pertinent to sexually diverse students.

Professional development has been recognized as a best practice for school psychologists, counselors, social workers, and nurses (American Psychological Association & National Association of School Psychologists, 2015). However, nearly 70% of these professionals report an absence of competency training on how to work with sexually diverse students (GLSEN et al., 2019). This poses a barrier to adolescent HIV prevention, particularly as AGBM have reported being most willing to discuss HIV testing and condom use with school nurses (Rasberry et al., 2015). This is particularly salient, as 21% of the AGBM in our study reported previously testing for HIV, and 52% reported condom use.

To strengthen the capacity of school health professionals, multiple professional organizations have collaborated to develop a curriculum that promotes the health and well-being of sexually diverse students (Greenberg et al., 2021). This includes goals for increasing student awareness of sexual health services, such as increased HIV testing, either at school or in the community (Centers for Disease Control and Prevention, 2019). The availability of partnerships with school-based health centers (SBHCs) has been shown to support adolescents’ health needs and academic achievement (Kjolhede et al., 2021). These SBHCs also improve access to health services by decreasing barriers, such as those that arise from financial, geographic, age, and cultural factors.

One paradoxical finding from our study was that those who reported academic achievement were less likely to report the outcome of HIV testing, meaning there was no protective association observed. However, protective associations were observed between academic achievement and the outcomes of early sexual debut and condom use. A second remarkable finding occurred among Hispanic/Latino AGBM, in which this demographic group displayed both a greater likelihood of reporting the outcomes of HIV testing and having multiple sexual partners, compared to non-Hispanic Whites. These findings indicate an important opportunity for SBHCs and community organizations to provide important prevention services, such as access to condom availability programs, HIV testing, pre- and post-exposure prophylaxis, and antiretroviral therapy when necessary (Andrzejewski et al., 2019; Hsu et al., 2021). These programs are vital for reducing HIV transmission, particularly among key populations where an increase in HIV incidence has been observed, such as Hispanics/Latinos (Guilamo-Ramos et al., 2020).

Policy development is also important for strengthening safe school environments, and this includes the implementation of policies that address homophobic bullying and victimization. Evidence shows that the presence of environments that are supportive of sexually diverse students, either in schools or communities, is also associated with lower odds of substance use (Eisenberg et al., 2020). However, zero-tolerance disciplinary policies have been shown to be ineffective and are associated with lower feelings of school safety (Huang & Cornell, 2021). In contrast, policies that utilize a pro-social behavioral approach, such as demonstrating kindness and caring to oneself and others, provide favorable student outcomes and are associated with improved classroom behavior, confidence, and interactions, as well as decreases in student anxiety and negative mental health (Preston & Rew, 2022). An example of this approach includes training staff to respond positively to inappropriate behaviors, such as requiring immediate apologies to peers and reviewing classroom expectations (Centers for Disease Control and Prevention, 2019).

Positive Youth Development

Evidence also shows that positive youth development (PYD) programs are an additional tool for improving school connectedness, academic achievement, and school attendance, as these programs facilitate relationships with supportive adults, bonds with peers, a sense of belonging, and self-efficacy (Curran & Wexler, 2017). Similarly, these programs are associated with reduced substance use, violence, mental health conditions, and HIV risk behaviors (Catalano et al., 2019). Examples of programs with PYD models include 4-H, scouting, and Boys and Girls Clubs (Fish, 2020).

Recent meta-analyses show that mentorship programs that promote PYD by pairing adolescents with caring non-parental adults significantly affect many youth outcomes (Raposa et al., 2019). More specifically, mentorship programs that employed targeted skills-based approaches, compared to non-specific relational approaches, were found to have a greater effect on academic, psychological, and social functioning (Christensen et al., 2020). The combination of digital technology with mentoring relationships, such as e-mentoring, has also grown in popularity. However, a recent systematic review of e-mentoring interventions shows that although e-mentoring has great potential for reach, the current strength of the evidence is weak (Kaufman et al., 2022).

Safe environments for communication are particularly important for AGBM, as these individuals increasingly experience rejection or lack of support from their families or caregivers at home (Smith et al., 2020). Additionally, AGBM have reported reluctance to talk with school staff when they were uncertain how staff perceive issues surrounding sexual orientation or when they perceived staff as lacking in knowledge regarding issues relevant to sexually diverse students (Rasberry et al., 2015).

In-school programs such as gay-straight alliances (GSAs), also known as gender and sexualities alliances or gender-sexuality alliances, are consistent with PYD models. These programs allow adults and role models to support youth-led activities in a safe and supportive environment and help foster a sense of agency, self-esteem, and empowerment (Johns et al., 2019). Meetings allow students to check in with others, follow up on topics from prior meetings, and discuss salient topics such as mental health, substance use, sexual health, and community health resources (Centers for Disease Control and Prevention, 2019). Studies have shown that students in schools with GSAs report lower levels of bullying and suspensions, and higher perceived school safety and grades (Lessard et al., 2020). In a 2019 survey, 62% of students reported having access to GSAs, an increase from 31% in 2001 (Truong et al., 2021), and guidance is available online for starting a GSA (GLSEN, 2023; GSA Network, 2023).

In summary, emerging evidence has shown that a school-based model consisting of sexual health education and services, professional development, and positive youth development programs can be used to strengthen safe school environments and academic achievement, two protective factors from our study that were associated with a reduction in several HIV risk behaviors. However, much progress is needed in this field, as few schools currently incorporate these programs, with only 8% of all adolescent students, or approximately two million, being reached (Wilkins et al., 2022).

Strengths and Limitations

A major strength and contribution of this secondary data analysis is the examination of associations between protective factors and HIV risk behaviors, rather than an exclusive focus on identifying risk behaviors. Also, the analysis drew from nationally representative data using a relatively large sample size from three timepoints (years 2015–2019), resulting in a pre-COVID-19 cross-section that can be used for post-pandemic comparisons as 2021 data begin to be analyzed.

Limitations of this study include the use of self-reported data, which can increase threats from bias. However, literature is available that demonstrates the reliability of self-reported data on adolescent sexual behaviors, such as recent condom use (Craker et al., 2019), an outcome of interest in this study. Second, the sampling frame omitted students enrolled in alternative, special, or vocational education programs, as well as those who may no longer attend high school, possibly due to the very factors that served as variables in this study (i.e., not feeling safe, poor academic achievement, and/or use of drugs or alcohol). Similarly, student participation in the YRBS can be affected by parental consent, as some states, districts, and schools may require signed permission forms to participate (i.e., active parental permission) (Centers for Disease Control and Prevention, 2023a). This is in contrast to consent that allows students to opt out of the survey by returning a signed form denying permission (i.e., passive parental permission). When active parental permission is required, it is estimated that participation may be reduced by half. Therefore, it is important to note that the generalizability of this study is limited to AGBM who attended high school and obtained parental permission to participate during survey administration. Third, the dataset did not contain geographic or socioeconomic data, i.e., zip codes, census tracts, or access to free or reduced school lunches, precluding the analysis of social and structural determinants of health, which are known drivers of HIV transmissions in the United States. Finally, the protective factors examined in this study were not comprehensive and were limited to those available in the YRBS dataset. However, our findings support mounting evidence that suggests that strong positive relationships with parents, caring adults, and teachers at school are protective of health-related outcomes and promote positive development (Brown et al., 2020; Hong et al., 2023; Sieving et al., 2017), providing direction for much needed future research and longitudinal studies.

Conclusion

The findings from this secondary data analysis provide robust evidence for the significance of school-based protective factors and their associations with reducing several HIV risk behaviors. Given the hours that students spend in school each week, strategies to strengthen safe school environments and academic achievement offer tremendous potential for adolescent HIV prevention. This is especially timely as the benchmark of the EHE initiative is to reduce new HIV infections by 90% by the year 2030. Moving forward, we recommend that future administrations of the YRBS include additional survey items that are designed to examine programmatic issues. Adding items on current practices related to safety and academic support would lend much needed data to educational stakeholders responsible for the structure of day-to-day activities and curricula design. In summary, achieving EHE benchmarks can be facilitated by generating evidence supporting policies, programs, investments, and funding for school-based programs that aim to improve social and structural determinants of health and reduce adolescent HIV burdens.

Key Considerations:

• When students are absent from school because they feel unsafe, they are less likely to be provided sexual health education and services.

• One barrier to adolescent HIV prevention is the absence of competency training for school professionals (i.e., psychologists, counselors, social workers, and nurses) on how to work with sexually diverse students.

• School-based health clinics can improve access and decrease barriers to sexual health services.

Data Accessibility:

No new data were used; therefore, data sharing is not applicable.