High-School Students: Sexual Orientation and Pregnancy Risk
High-School Students: Sexual Orientation and Pregnancy Risk
Conducted every 2 years, the Centers for Disease Control and Prevention's YRBSS monitors priority health risk behaviors, including sexual behaviors that contribute to unintended pregnancy, among representative samples of 9th- through 12th-grade students. Most states (48) and more than 20 large urban school districts participate in the YRBSS; however, only 8 sites—5 states (Delaware, Maine, Massachusetts, Rhode Island, and Vermont) and 3 large urban school districts (Boston, MA; Chicago, IL; and New York City, NY)—included questions about students' sexual identity and gender of sexual partners. Of these 8 sites, only 4 (2 cities and 2 states) asked whether students had "ever been pregnant or gotten someone pregnant." New York City had the largest and most racially diverse sample of these 4 remaining sites. To eliminate possible confounding because of regional, city, or district-level variability, we selected the New York City school district for our study.
We used data from the 2005, 2007, and 2009 New York City YRBSS. The New York City YRBSS employs a stratified, 2-stage probability sample designed to produce a representative sample of New York City students attending grades 9 through 12 in each survey year. In the first sampling stage, schools were randomly selected proportional to school enrollment. In the second stage, classrooms meeting during a selected time period or covering a required subject were randomly selected and all students within that classroom were asked to complete a self-administered YRBSS questionnaire. Students who did not want to participate were allowed to opt out.
Since 2005, schools served by the Department of Health and Mental Hygiene's District Public Health Offices located in 3 high-risk neighborhoods (i.e., South Bronx, East and Central Harlem, and North and Central Brooklyn) were oversampled to ensure that these subareas were represented in the sample. In 2009, the New York City YRBSS also included an oversample of schools served by School-Based Health Centers, which provide onsite primary care to students in the 5 boroughs of New York City. Schools were selected for citywide and borough-level representation. Overall response rates, which were averaged across school and student response rates, ranged from 68% in 2005 and 2007 to 79% in 2009.
In 2005, 8140 New York City students completed the YRBSS questionnaire. Sample sizes in 2007 and 2009 were 9080 and 11 887, respectively. We restricted our sample to students who reported having had vaginal intercourse (n = 10 827; 5278 female and 5549 male students). Finally, we excluded students who had missing data on any of the variables included in our analyses, resulting in a final sample size of 4892 female and 4811 male students.
Students were asked how many times they had ever been pregnant or gotten someone pregnant (dependent variable). Students who responded that they had been pregnant or had gotten someone pregnant (i.e., pregnancy involvement) 1 or more times were coded 1; otherwise they were coded 0.
We measured sexual orientation with 2 items—gender of sexual partners and sexual identity. Students were asked, "During your life, with whom have you had sexual contact?" For female students, gender of sexual partners was categorized as male partners only or male and female partners. For male students, gender of sexual partners was categorized as female partners only or male and female partners. Students were asked, "Which of the following best describes you? Heterosexual (straight), gay or lesbian, bisexual, or unsure." We categorized sexual identity as (1) heterosexual; (2) gay, lesbian, or bisexual; or (3) unsure. We collapsed gay or lesbian with bisexual because the risk profiles for both groups were similar and because there were issues of data sparseness (n < 10 in key cells).
To examine the extent to which differences in sexual behaviors explain the association between sexual orientation and pregnancy, we included a number of indicators of sexual behaviors in our models. Age at first intercourse was measured as aged 13 years or younger, 14 or 15 years, and 16 years or older. We measured the number of sexual partners the students reported in their lifetime as 1 partner, 2 or 3 partners, 4 or 5 partners, or 6 or more partners. Respondents also reported if they had ever been physically forced to have sexual intercourse.
We categorized race/ethnicity as non-Hispanic White, non-Hispanic Black, Hispanic (any race), Asian/Pacific Islander, or other race. We included age of respondent as a continuous variable ranging from 12 to 18 years (the age variable was bottom coded at 12 and top coded at 18). We categorized survey year as 2005, 2007, or 2009.
We stratified all analyses by gender, as the relationship between sexual orientation, sexual behaviors, and pregnancy may differ for girls and boys. We began with descriptive statistics. Next, we examined bivariate associations between sexual identity and the covariates and dependent variable, as well as between gender of sexual partners and the covariates and dependent variable. We employed multivariable logistic regression to examine the association between sexual orientation, measured as sexual identity and gender of sexual partners, and risk of pregnancy, adjusting for demographics and sexual behaviors. We weighted all analyses to adjust for the New York City YRBSS's sampling design and unit nonresponse by using the SVY command in Stata version 13 (StataCorp LP, College Station, TX).
Because of modest correlations between sexual identity and gender of sexual partners (r = 0.56, female students; r = 0.61, male students), we examined the data for possible multicollinearity. We found no indication of multicollinearity; none of the variables in the final model had variance inflation factors larger than 4.07 or tolerances smaller than 0.25. Furthermore, gender of sexual partners and sexual identity had variance inflation factors no larger than 2.5 and tolerances no smaller than 0.40. Generally, a cutpoint of 10 (variance inflation factor) and 0.10 (tolerance) are used to indicate multicollinearity.
Methods
Conducted every 2 years, the Centers for Disease Control and Prevention's YRBSS monitors priority health risk behaviors, including sexual behaviors that contribute to unintended pregnancy, among representative samples of 9th- through 12th-grade students. Most states (48) and more than 20 large urban school districts participate in the YRBSS; however, only 8 sites—5 states (Delaware, Maine, Massachusetts, Rhode Island, and Vermont) and 3 large urban school districts (Boston, MA; Chicago, IL; and New York City, NY)—included questions about students' sexual identity and gender of sexual partners. Of these 8 sites, only 4 (2 cities and 2 states) asked whether students had "ever been pregnant or gotten someone pregnant." New York City had the largest and most racially diverse sample of these 4 remaining sites. To eliminate possible confounding because of regional, city, or district-level variability, we selected the New York City school district for our study.
Sample
We used data from the 2005, 2007, and 2009 New York City YRBSS. The New York City YRBSS employs a stratified, 2-stage probability sample designed to produce a representative sample of New York City students attending grades 9 through 12 in each survey year. In the first sampling stage, schools were randomly selected proportional to school enrollment. In the second stage, classrooms meeting during a selected time period or covering a required subject were randomly selected and all students within that classroom were asked to complete a self-administered YRBSS questionnaire. Students who did not want to participate were allowed to opt out.
Since 2005, schools served by the Department of Health and Mental Hygiene's District Public Health Offices located in 3 high-risk neighborhoods (i.e., South Bronx, East and Central Harlem, and North and Central Brooklyn) were oversampled to ensure that these subareas were represented in the sample. In 2009, the New York City YRBSS also included an oversample of schools served by School-Based Health Centers, which provide onsite primary care to students in the 5 boroughs of New York City. Schools were selected for citywide and borough-level representation. Overall response rates, which were averaged across school and student response rates, ranged from 68% in 2005 and 2007 to 79% in 2009.
In 2005, 8140 New York City students completed the YRBSS questionnaire. Sample sizes in 2007 and 2009 were 9080 and 11 887, respectively. We restricted our sample to students who reported having had vaginal intercourse (n = 10 827; 5278 female and 5549 male students). Finally, we excluded students who had missing data on any of the variables included in our analyses, resulting in a final sample size of 4892 female and 4811 male students.
Measures
Students were asked how many times they had ever been pregnant or gotten someone pregnant (dependent variable). Students who responded that they had been pregnant or had gotten someone pregnant (i.e., pregnancy involvement) 1 or more times were coded 1; otherwise they were coded 0.
We measured sexual orientation with 2 items—gender of sexual partners and sexual identity. Students were asked, "During your life, with whom have you had sexual contact?" For female students, gender of sexual partners was categorized as male partners only or male and female partners. For male students, gender of sexual partners was categorized as female partners only or male and female partners. Students were asked, "Which of the following best describes you? Heterosexual (straight), gay or lesbian, bisexual, or unsure." We categorized sexual identity as (1) heterosexual; (2) gay, lesbian, or bisexual; or (3) unsure. We collapsed gay or lesbian with bisexual because the risk profiles for both groups were similar and because there were issues of data sparseness (n < 10 in key cells).
To examine the extent to which differences in sexual behaviors explain the association between sexual orientation and pregnancy, we included a number of indicators of sexual behaviors in our models. Age at first intercourse was measured as aged 13 years or younger, 14 or 15 years, and 16 years or older. We measured the number of sexual partners the students reported in their lifetime as 1 partner, 2 or 3 partners, 4 or 5 partners, or 6 or more partners. Respondents also reported if they had ever been physically forced to have sexual intercourse.
We categorized race/ethnicity as non-Hispanic White, non-Hispanic Black, Hispanic (any race), Asian/Pacific Islander, or other race. We included age of respondent as a continuous variable ranging from 12 to 18 years (the age variable was bottom coded at 12 and top coded at 18). We categorized survey year as 2005, 2007, or 2009.
Data Analysis
We stratified all analyses by gender, as the relationship between sexual orientation, sexual behaviors, and pregnancy may differ for girls and boys. We began with descriptive statistics. Next, we examined bivariate associations between sexual identity and the covariates and dependent variable, as well as between gender of sexual partners and the covariates and dependent variable. We employed multivariable logistic regression to examine the association between sexual orientation, measured as sexual identity and gender of sexual partners, and risk of pregnancy, adjusting for demographics and sexual behaviors. We weighted all analyses to adjust for the New York City YRBSS's sampling design and unit nonresponse by using the SVY command in Stata version 13 (StataCorp LP, College Station, TX).
Because of modest correlations between sexual identity and gender of sexual partners (r = 0.56, female students; r = 0.61, male students), we examined the data for possible multicollinearity. We found no indication of multicollinearity; none of the variables in the final model had variance inflation factors larger than 4.07 or tolerances smaller than 0.25. Furthermore, gender of sexual partners and sexual identity had variance inflation factors no larger than 2.5 and tolerances no smaller than 0.40. Generally, a cutpoint of 10 (variance inflation factor) and 0.10 (tolerance) are used to indicate multicollinearity.
