HIV testing among United States high school students at the state and national level, Youth Risk Behavior Survey 2005–2011

Interaction between race and number of sexual partners

Although number of sexual partners was significantly associated with testing for all students, it was particularly important for White students. One possible explanation could be the higher level of HIV testing among Black or African American students with less than 4 partners compared to White and Hispanic. This could also be explained by the lower use of condom among high school students with higher numbers of sexual partners, in particular among White, as reported by Richter et al. (1993).

Interaction between sex and age at first intercourse

Having first intercourse before 13 was associated with higher odds of HIV testing among males. One possible explanation could be that the share of black students is higher among males who had sex before 13 compared to males who had sex after 13, whereas it is similar for females who had sex either before or after 13 (Cavazos-Rehg et al. 2009). As Black students have higher rates of HIV testing, this could drive the difference between the two male subgroups.

Interaction between sex and use of condom

Not using a condom at last intercourse was associated with higher odds of HIV testing among females compared to males. Young women, who use condoms as contraception and did not use one at last sex, are likely to get tested for pregnancy at a care provider (Daley et al. 2005) and should receive HIV testing, given efforts for integrating pregnancy and STIs prevention among teens (Tran et al. 2010). Besides, young women who did not use condom at last sex, but use oral contraception, are likely to receive counseling and HIV testing from the doctor who prescribes their contraceptive (American College of Obstetricians and Gynecologists (ACOG) 2008).

The comparison between states and the national sample highlighted geographic differences in HIV testing and risk behavior profiles, with the most common difference in prevalence use of condom at last sex. This reinforces previous observations of risk profile variations across locations among older adolescents (Moore et al. 2013).

These geographic differences are very informative for State Health Departments as they point toward areas that need further investments in programs. They can also inform state education agencies and school districts about local adaptations needed in HIV/AIDS educational curriculum (National Association of State Boards of Education 2013). For example, currently, only 33 states and the district of Columbia mandate HIV education and only 13 states require the inclusion of information on the possible negative outcomes of teen sex and pregnancy (Guttmacher Institute 2014). Thus, states could gain great benefits from reinforcing their programs according to the sexual risk behaviors patterns of their local youth populations.

Forty-eight states plus District of Columbia now have HIV testing laws that are consistent with the key features of the CDC’s 2006 recommendations: opt-out HIV testing, no written consent required, and no pretest counseling required (CDC 2013c). Still, some differences remain across states (e.g., minimum age required to consent, availability of anonymous testing, reporting policies of positive HIV tests) and can sometimes limit widespread HIV testing among adolescents (National HIV/AIDS Clinicians’ Consultation Center 2014). Additionally, rapidly aligning practices with new HIV testing laws can be a real challenge in itself (Hyden et al. 2014; Kelley et al. 2011). For example, some HIV locations lack a clear and consistent communication of their confidentiality policies for HIV testing due to a lack of training (Hyden et al. 2014).

This study also highlights the missed opportunity to collect HIV testing information through YRBS. Although weighted data were available for 43 states, only 7 monitored HIV testing for at least one of the years within 2005–2011. National and state use the same standard questionnaire, but state and local officials responsible for the coordination of YRBS can add or delete questions to meet their policy or programmatic needs (CDC 2013a). This can result in gaps in knowledge. Controlling HIV/AIDS epidemic requires measuring progress in HIV testing and prioritizing interventions accordingly. The comprehensive monitoring of HIV testing across all states through YRBS would be an important step forward in terms of HIV prevention among youths. Furthermore, systematic monitoring of sexual preferences would be of great value in targeting populations at increased risk of transmission, such as male having sex with male (CDC 2013d).

This study has some limitations. The cross-sectional design of YRBS prevents identification of temporal order between HIV testing and risk behaviors. Additionally, the self-reported nature of YRBS can introduce bias (e.g., recall, non-response, social desirability) (Sackett 1979). However, several studies evaluated YRBS validity and concluded to its reliability (Brener et al. 2002). Additionally, YRBS is limited to students attending high school; this results in selection bias. Finally, higher dropouts among Hispanics and Black or African American ethnicities can also raise bias issue (National Center for Education Statistics 2012).

Participants and procedure

This study used national and state YRBS 2005–2011. YRBS measures the prevalence of health risk behaviors among high school students, through an anonymous paper-and-pencil survey administered every other year in randomly selected public and private high schools. National and state levels are representative of high school students in the United States and states considered. They come from separate scientific samples of schools and students, follow the same survey methodology and use the same core questionnaire, to which state and local agencies can add or delete questions (CDC 2013a). National YRBS data are publicly accessible and state data are available upon request. After IRB (Emory University) approval, data were retrieved for 43 states. Data were not available for the seven remaining states due to either absence of weighted data (California, Oregon), no YRBS participation (Minnesota, Washington), or no data sharing policy (Hawaii, Indiana, and Vermont). Out of the 43 states, seven collected the information related to HIV testing through their YRBS questionnaire for at least one year in the studied period (Massachusetts: 2005–2011; Connecticut: 2007–2011; New Jersey: 2005, 2009, and 2011; Arkansas, North Carolina, North Dakota: 2009–2011; South Carolina: 2009). This study used 59,793 observations at the national level and 39,421 observations at the state level.

Measures

Statistical analysis

All analyses were conducted using SAS-callable SUDAAN version 9.3, accounted for the complex sampling design, and used a significant level α = 0.05. Domain analyses were used to perform calculation on the subpopulation of high school students who had sexual intercourse. Less than 10% of data were missing across all variables, except for: HIV testing in the 2011 national survey (24%) and in North Carolina (12% for 2009–2011); and lifetime illegal injection drug use in North Carolina (29% for 2009–2011). Missing data were accounted for by using the “not missing completely at random” option.

Descriptive analyses were conducted at national and state levels for the period 2005–2011. Testing prevalence, risk behavior prevalence and demographic characteristics were calculated among students who ever had sex. Finite population correction was applied for the calculation of state prevalence due to the smaller size of state samples (<5,000).

The association of HIV testing with risk behaviors at the national level for the 2005–2011 period was modeled using logistic regression. First, logistic regression was used to assess the crude association between HIV testing and each risk behavior and demographic. Multivariable logistic regression model was then used to assess the adjusted association between HIV testing and each independent variable, controlling for sex, grade, and race/ethnicity. Effect modification by sex, race/ethnicity, grade level, and year of survey was assessed through backward elimination. The Bonferroni method was used to correct multiple testing (Bland and Altman 1995). Four interaction terms were significant at the level 0.05, with k = α/4, and remained in the final model: Race/ethnicity and Number of sexual partner in life (p-value = 0.004), Sex and Age at first sexual intercourse (p-value = 0.004), Sex and Use of condom at last intercourse (p-value = 0.007), and Grade and Ever forced intercourse (p-value = 0.012). No multicolinearity was found between covariates.

State HIV testing and risk behavior prevalence were compared to national estimates. Significant differences were assessed using t-test controlling for sex, grade, and race/ethnicity. Multiple-testing was accounted for through a statewide significance α = 0.05 (k = α/6).