Sexual Orientation and Suicide Ideation, Plans, Attempts, and Medically Serious Attempts: Evidence From Local Youth Risk Behavior Surveys, 2001–2009

Abstract

Objectives. We examined the associations between 2 measures of sexual orientation and 4 suicide risk outcomes (SROs) from pooled local Youth Risk Behavior Surveys.

Methods. We aggregated data from 5 local Youth Risk Behavior Surveys from 2001 to 2009. We defined sexual minority youths (SMYs) by sexual identity (lesbian, gay, bisexual) and sex of sexual contacts (same- or both-sex contacts). Survey logistic regression analyses controlled for a wide range of suicide risk factors and sample design effects.

Results. Compared with non-SMYs, all SMYs had increased odds of suicide ideation; bisexual youths, gay males, and both-sex contact females had greater odds of suicide planning; all SMYs, except same-sex contact males, had increased odds of suicide attempts; and lesbians, bisexuals, and both-sex contact youths had increased odds of medically serious attempts. Unsure males had increased odds of suicide ideation compared with heterosexual males. Not having sexual contact was protective of most SROs among females and of medically serious attempts among males.

Conclusions. Regardless of sexual orientation measure used, most SMY subgroups had increased odds of all SROs. However, many factors are associated with SROs.

Suicide is the 10th leading cause of death overall and the 3rd leading cause of death among youths aged 10 to 24 years. In 2010, more than 157 000 people in this age group visited US emergency departments because of attempted suicide or other self-harm injuries.¹ Media reports convey the message that sexual minority youths (SMYs) have much greater rates of suicide (i.e., self-inflicted death) than do non-SMYs; however, the unavailability of sexual orientation information on death certificates makes this impossible to confirm or refute using archival data. What is known is that studies document large disparities in key indicators of suicide risk among SMYs, including suicidal ideation (i.e., considering suicide) and suicide attempts (i.e., nonfatal self-directed potentially injurious behavior with the intent to die).^2–5 In early studies, often using small convenience samples without comparison groups, 20% to 40% of SMYs reported suicidal ideation and attempts.^6–8 Later population-based surveys confirmed these reports and found odds of suicidal ideation and attempts up to 5 to 6 times greater among SMYs than among non-SMYs.^9–16 Researchers understand this increased risk for suicide ideation and attempts in the context of minority stress,^17,18 whereby a hostile social environment characterized by stigma, prejudice, and discrimination may be associated with increases in individual risk factors for suicide, including depression, substance abuse, social isolation, peer conflict, and victimization^{4,5,11,19–32} and decreases in protective factors such as supportive relationships with peers and family.^33–35

We sought to expand what is known about the risk of suicide among SMYs by addressing gaps in the research related to the measurement of both sexual orientation and suicide risk outcomes (SROs). That is, most studies on the topic measure only suicide ideation or attempts.^{11–15,36,37} Less is known about the full range of suicidal behaviors, including suicide planning and medically serious attempts (MSAs). These outcomes are important, as they indicate sustained injury and may indicate increased risk for future suicide, suicide attempt, or repeat attempts.^38–41 Indeed, some research suggests that SMYs not only make more medically serious attempts but also have greater intent to die.^10,42,43 Other research contests these findings.⁴⁴ The Youth Risk Behavior Survey (YRBS) is a population-based study of high school students administered nationally, statewide, and locally that measures ideation, plans, attempts, and MSAs. Four state or local sites have published studies examining sexual orientation and SROs.^{9,16,37,45–48} All studies examined suicide attempts, 3 measured ideation,^37,46,47 4 measured plans,^37,45–47 4 measured MSAs,^16,37,46,47 and 1 measured all.³⁷ In this last study, the sample size precluded conducting adjusted analyses. A recent Centers for Disease Control and Prevention study reported prevalence rates of all SROs by sexual orientation for each of the selected state and local sites that collected sexual orientation information.⁴⁹ We aggregated data across local sites providing adequate power to test the associations between sexual orientation and SROs while accounting for a range of risk factors and demographic variables. The use of data across local sites that are also urban areas adds a unique aspect to this study, as little is known about the associations between sexual orientation and SROs among urban populations.

In addition to expanded measurement of SROs and a unique sample, we have provided multiple measures of sexual orientation. Most studies measure a single dimension of sexual orientation, typically sexual identity (e.g., lesbian, gay, bisexual [LGB]) or sexual behavior (sexual contact with opposite, same, or both sexes).^29,32,50 This assumes that dimensions of sexual orientation are interchangeable and that 1 measure correctly identifies all SMYs. Sexual orientation, however, is multidimensional and dimensions may not overlap.^9,49,51 For example, SMYs of color may engage in sexual contact with same-sex partners but not identify as sexual minorities because of social stigma.⁵² We have added to the research base and examined the associations between sexual orientation and SROs using 2 of 3 recommended dimensions of sexual orientation—sexual identity and sexual behavior⁵³—with sexual attraction being the third and currently unavailable recommended measure. Finally, to avoid obscuring important within-group differences imposed by dichotomous measures of sexual orientation (e.g., LGB vs heterosexual),^{9,10,12,15,16,45,54} we analyzed subgroups of males and females on the basis of sexual identity and sex of sexual contacts, including the less studied population of youths who are unsure of their sexual identity.^15,42 Using data from a unique urban sample and with expanded measures of sexual orientation and SROs, we asked the following questions:

1. On average, do youths who report their sexual identity as LGB or unsure have increased odds of suicide ideation, plans, suicide attempts, and medically serious attempts compared with heterosexual youths, controlling for a range of individual-level risk factors and demographic variables?

2. On average, do youths who report same- or both-sex sexual contact have increased odds of suicide ideation, plans, suicide attempts, and medically serious attempts compared with youths who have sexual contact with opposite-sex partners only, controlling for all other factors?

METHODS

The Youth Risk Behavior Surveillance System monitors health risk behaviors—such as unintentional injuries, violence, tobacco, alcohol, and other drug use and sexual risk behaviors. It includes surveys conducted at the national, state, and local levels. Data for our study came from local sites included in the recent Centers for Disease Control and Prevention study⁴⁹ with weighted data that measured SROs, included the optional items on sexual orientation, and had parallel wording of items across years and sites. These sites included Boston, Massachusetts (2001–2009); Chicago, Illinois (2003–2009); New York City, New York (2005–2009); San Diego, California (2001–2005); and San Francisco, California (2001–2009). Although Milwaukee, Wisconsin, measured sexual orientation, it did not meet all inclusion criteria, and therefore we did not include it. Each local site used an independent, cross-sectional, 2-stage clustered design to produce representative samples of public school students in grades 9 through 12 in their districts.

Measures

Sexual orientation.

Sexual orientation was measured by 2 questions: “Which of the following best describes you?” Response options included heterosexual, lesbian or gay, bisexual, and unsure; and “During your life, with whom have you had sexual contact?” Response options included I have never had sexual contact, females, males, and females and males. We classified students as having had opposite-sex contact only, same-sex contact only, both-sex contact, or as never having had sexual contact.

Suicide risk outcomes.

We measured 4 dichotomous SROs: seriously considered suicide (ideation), made a plan for suicide (plans), attempted suicide 1 or more times (attempt), and attempted suicide that resulted in an injury, poisoning, or overdose that had to be treated by a doctor or nurse (an MSA) during the past 12 months.

Control variables.

We included the following risk factors occurring over the past 12 months or ever: having been threatened or injured with a weapon on school property (threatened); engaging in physical fighting on or off school property (in a fight); having been hit, slapped, or physically hurt on purpose by a boyfriend or girlfriend (physical dating violence); ever being physically forced to have sexual intercourse (sexual violence); smoking, drinking, or using marijuana for the first time before aged 13 years; and ever using other illegal drugs (hard drug use) as indicated by use of any form of cocaine, inhalants, ecstasy, heroin, methamphetamines, steroids without a prescription, or injection drugs.

We also included a variable with some evidence of an association with suicidal behavior: perceiving oneself as being overweight or obese as compared with underweight/normal (weight).^55,56 We did not include risk factors measured in a period shorter than the “past year.” Finally, we included the item “ever feeling so sad or hopeless almost every day for 2 weeks or more in a row that you stopped doing some usual activities” (hopelessness). Demographic variables included sex, race/ethnicity (American Indian or Alaska Native, Asian/Pacific Islander, Black, White [reference group], Hispanic, and multiracial/ethnic or other), and grade (ninth grade served as the reference group). Finally, we included measures of city and year to control for unobserved confounding by place and time; New York City and the year 2005 served as the reference groups, respectively.

Analytic Approach

We merged all years of available data from 2001 to 2009 for each local YRBS site that included measures of sexual orientation. We created 2 analytic samples from these site-specific data sets: sites measuring sexual identity defined 1 sample and sites measuring sex of sexual contacts defined the other. In pooling data across sites and years, we identified some inconsistencies in survey items (e.g., physical dating violence is typically measured relative to the past 12 months; however, from 2001 to 2007, Boston did not specify a time period). We used several approaches to test the robustness of results in such instances, for example, by dropping variables or observations or recoding variables. These alternate approaches resulted in small or almost no changes in key variables of interest, so we retained variables and observations in the original approach wherever possible.

We performed all statistical analyses on weighted data to adjust for student nonresponse and to account for the complex sampling design of the YRBS. We scaled weights on the basis of the number of survey years contributed to the sample. After examining missing data and the sample characteristics across independent and combined sites (data available as a supplement to the online version of this article at http://www.ajph.org.), we compared the distribution of demographic variables (Table 1) and SROs (Table 2) by sexual orientation using the weighted Rao-Scott χ² test of association. We also assessed demographic differences among youths with missing sexual orientation and SRO data. We next conducted unadjusted logistic regression analyses to model the bivariate associations between categories of sexual identity and sex of sexual contacts, respectively, and each SRO—ideation, plans, attempts, and MSAs. We added all control variables simultaneously. We have reported unadjusted and adjusted odds ratios (ORs) and 95% Wald confidence intervals (CIs). We conducted all analyses on complete case data using SAS version 9.3 (SAS Institute, Cary, NC) survey procedures.

TABLE 1—

Sample Characteristics by Sexual Orientation: Local Youth Risk Behavior Survey Sites, United States, 2001–2009

	Sexual Identity,a No. or %				Sex of Sexual Contacts,b No. or %
Characteristic	Heterosexual	Lesbian or Gay	Bisexual	Unsure	Opposite	Same	Both	No Contact
Females
No.	21 700cd	362	1497	842	10 081	577	1199	11 024
Sexual orientation	89.6	1.4	5.7	3.3	43.9	2.4	5.0	48.6
Grade
9	90.8	1.2	4.8	3.2	33.4	2.3	3.8	60.5
10	88.3	1.1	6.7	3.9	43.4	2.3	5.4	48.9
11	89.2	1.6	6.0	3.2	48.2	2.4	5.6	43.8
12	90.0	1.7	5.5	2.8	57.2	2.6	5.6	34.5
Other or ungraded	81.9	5.2	5.9	7.0	31.1	3.9	16.2	48.8
Race/ethnicity
AI/AN	76.3	2.2	5.1	16.5	26.8	3.5	7.8	61.9
A/PI	92.7	0.6	2.1	4.7	22.0	1.7	2.1	74.2
Black	90.7	1.7	5.0	2.6	51.8	3.0	4.2	40.9
White	90.3	0.5	5.6	3.6	41.7	1.2	6.2	50.9
Hispanic	89.2	1.3	6.8	2.6	42.7	2.2	5.1	50.0
Multiple or other	83.0	2.2	10.3	4.5	43.4	3.0	9.7	43.9
Males
No.	20 643	399	509	600	11 667	551	392	7645
Sexual orientation	92.7	2.0	2.2	3.1	57.0	2.8	2.0	38.2
Grade
9	92.4	1.9	2.4	3.3	52.1	2.0	1.7	44.2
10	93.4	1.3	2.2	3.1	56.4	2.8	2.1	38.8
11	92.2	2.4	2.3	3.1	59.4	2.5	2.3	35.7
12	93.5	2.2	1.6	2.6	64.8	4.2	1.6	29.4
Other or ungraded	83.4	9.5	2.3	4.9	53.5	7.8	6.6	32.0
Race/ethnicity
AI/AN	72.8	2.2	11.9	13.0	46.8	1.7	7.2	44.3
A/PI	93.5	0.8	1.4	4.3	29.9	1.7	1.4	67.0
Black	92.3	2.5	2.1	3.1	67.4	3.4	2.1	27.1
White	94.3	1.0	2.4	2.4	46.5	2.5	1.9	49.1
Hispanic	94.8	1.8	1.8	1.6	60.9	2.3	1.7	35.1
Multiple or other	89.5	3.2	3.4	3.9	61.0	3.5	2.5	33.0

Note. A/PI = Asian/Pacific Islander; AI/AN = American Indian/Alaska Native. Column percentages may not add to 100 because of rounding.

^aSites measuring sexual identity are Boston, MA; Chicago, IL; New York City, NY; and San Francisco, CA.

^bSites measuring sex of sexual contacts are Boston, MA; Chicago, IL; New York City, NY; and San Diego, CA.

^cUnweighted sample size and weighted percent.

^dSample sizes across grade and race/ethnicity vary because of missing data.

TABLE 2—

Prevalence of Suicide Risk Outcomes by Sexual Orientation: Local Youth Risk Behavior Survey Sites, United States, 2001–2009

Variable	Suicide Ideation, No. (%)a	Suicide Plans, No. (%)a	Suicide Attempts, No. (%)a	Medically Serious Attempts, No. (%)a
Females
Sexual identityb
Heterosexual	3197 (14.9)	2079 (11.5)	1590 (8.8)	413 (2.2)
Lesbian	108 (35.4)	53 (26.0)	81 (28.3)	26 (9.0)
Bisexual	580 (42.1)	305 (34.8)	357 (30.1)	110 (8.0)
Unsure	219 (27.3)	148 (22.8)	121 (17.9)	29 (4.4)
Males
Sexual identityb
Heterosexual	1779 (8.5)	1353 (8.1)	1087 (6.8)	391 (2.7)
Gay	82 (24.0)	51 (15.4)	59 (23.4)	25 (8.7)
Bisexual	157 (31.2)	103 (31.0)	112 (26.4)	54 (11.6)
Unsure	122 (20.2)	87 (17.9)	89 (18.2)	43 (9.8)
Sex of sexual contactsc
Females
Opposite	2027 (20.2)	1256 (15.6)	1085 (12.5)	277 (3.1)
Same	141 (28.3)	77 (18.0)	107 (20.5)	28 (6.4)
Both	471 (38.7)	250 (34.1)	286 (26.8)	102 (7.8)
No contact	1459 (12.6)	935 (9.5)	687 (7.0)	158 (1.6)
Males
Opposite	1100 (9.6)	816 (9.1)	720 (8.4)	272 (3.4)
Same	126 (27.9)	71 (17.2)	81 (20.5)	28 (6.8)
Both	151 (33.8)	85 (28.7)	109 (27.3)	56 (16.8)
No contact	645 (8.1)	433 (7.4)	306 (4.8)	111 (1.8)

Note. All χ² test results are significant at P ≤ .001.

^aUnweighted sample size and weighted percentage.

^bSites measuring sexual identity are Boston, MA; Chicago, IL; New York City, NY; and San Francisco, CA.

^cSites measuring sex of sexual contacts are Boston, MA; Chicago, IL; New York City, NY; and San Diego, CA.

RESULTS

Between 1.1% and 19.7% of youths did not answer the SRO questions and between 6.0% and 6.2% did not answer the sexual orientation questions, and we therefore excluded them. Between 0.2% and 2.1% had missing data on both. Females, students in advanced grades, and White youths were more likely to be included than were males, ninth graders, and all racial/ethnic minority students. Racial/ethnic minorities comprised more than 85.0% of the sample, and most youths were in 9th or 10th grade (site-specific demographic characteristics are available as a supplement to the online version of this article at http://www.ajph.org).

Four sites measured sexual identity: Boston, Chicago, New York, and San Francisco (Table 1). Nearly 93.0% of males and 90.0% of females identified as heterosexual. Two percent of males and 1.4% of females identified as gay or lesbian, and 2.2% of males and 5.7% of females identified as bisexual. Heterosexual identity ranged from 72.8% among American Indian or Alaska Native males to 94.8% among Hispanic males. Four cities measured sex of sexual contacts: Boston, Chicago, New York, and San Diego. More than 60.0% of males and more than half of females ever had sexual contact. Almost 3.0% of males and 2.4% of females reported same-sex contact and more than twice as many females (5.0%) as males (2.0%) reported both-sex contact.

The prevalence of SROs varied significantly on the basis of sexual identity and sex of sexual contacts among both males and females (Table 2). LGB and unsure youths indicated approximately 2 to 3 times the prevalence of all SROs compared with heterosexual peers. On the basis of sexual identity, heterosexual males reported the lowest prevalence of ideation (8.5%), plans (8.1%), and attempts (6.8%); and bisexual females reported the highest prevalence of these outcomes, 42.1%, 34.8%, and 30.1%, respectively. MSAs ranged from 2.2% among heterosexual females to 11.6% among bisexual males. On the basis of sexual contact, youths without any sexual contact had the lowest prevalence of SROs (range: 1.6% [MSAs] to 12.6% [ideation]), and youths with both-sex contact had the greatest prevalence of all SROs (range: 7.8% [MSAs] to 38.7% [ideation]).

In unadjusted logistic regression analyses examining the associations between sexual identity and SROs, LGB and unsure youths had significantly increased odds of all SROs compared with heterosexual peers, with the ORs ranging from between 2.02 (95% CI = 1.03, 3.96) for MSAs among unsure females to 5.11 (95% CI = 3.16, 8.25) for planning among bisexual males (Table 3). When controlling for suicide risk factors and demographic variables in adjusted analyses, the odds of SROs among LGB youths declined but remained significantly elevated in most cases compared with heterosexual peers. ORs ranged between 1.72 (95% CI = 1.11, 2.68) for MSAs among bisexual females and 4.24 (95% CI = 2.09, 8.59) for suicidal ideation among gay males. The odds of suicide planning among lesbians did not remain significantly elevated, nor did the odds of MSAs among gay males differ from heterosexual males. Adjusting for all other factors, the odds of SROs among unsure youths did not differ from heterosexual peers, except for suicidal ideation among males, which remained elevated. Demographically, racial/ethnic minority groups compared with White, non-Hispanic youths generally did not have increased risk of SROs except suicide attempt (results not shown in Table 3). Among females, all risk factors except marijuana use were independently associated with at least 1 SRO, and 4 risk factors (being threatened with a weapon, dating violence, hopelessness, and hard drug use) were associated with all SROs. Drinking was associated with most SROs among females but was associated only with ideation among males. Also, unlike among females, among males marijuana use was a risk factor for all SROs, except planning. Sexual violence, hopelessness, and hard drug use were associated with all SROs for males.

TABLE 3—

Unadjusted and Adjusted ORs of Suicide Risk Outcomes by Sexual Identity: Local Youth Risk Behavior Survey Sites, United States, 2001–2009

	Suicide Ideation, No. or OR (95% CI)	Suicide Plans, No. or OR (95% CI)	Suicide Attempts, No. or OR (95% CI)	Medically Serious Attempts, No. or OR (95% CI)
Females
Unadjusted	24 212a	18 379	20 545	20 491
Sexual identity
Heterosexual (Ref)	1.00	1.00	1.00	1.00
Lesbian	3.14*** (1.97, 5.03)	2.71** (1.46, 5.02)	4.09*** (2.49, 6.71)	4.34*** (2.21, 8.50)
Bisexual	4.16*** (3.37, 5.15)	4.10*** (3.14, 5.36)	4.46*** (3.57, 5.57)	3.84*** (2.78, 5.32)
Unsure	2.15*** (1.60, 2.89)	2.27*** (1.57, 3.27)	2.26*** (1.59, 3.20)	2.02* (1.03, 3.96)
Adjusted	19 110	14 398	16 532	16 517
Sexual identity
Heterosexual (Ref)	1.00	1.00	1.00	1.00
Lesbian	2.35* (1.22, 4.53)	1.62 (0.73, 3.60)	3.95*** (1.95, 8.02)	3.33** (1.54, 7.21)
Bisexual	2.78*** (2.09, 3.69)	2.45*** (1.73, 3.45)	2.59*** (1.87, 3.59)	1.72* (1.11, 2.68)
Unsure	1.49 (0.97, 2.28)	1.31 (0.78, 2.19)	1.21 (0.79, 1.85)	0.72 (0.29, 1.79)
Risk factors
Threatenedb	1.42** (1.11, 1.83)	1.39* (1.03, 1.87)	1.68*** (1.24, 2.28)	2.75*** (1.84, 4.11)
In a fight	1.24* (1.05, 1.46)	1.33** (1.10, 1.61)	1.32** (1.09, 1.61)	1.31 (0.90, 1.89)
Physical dating violence	1.45*** (1.17, 1.80)	1.78*** (1.41, 2.24)	1.76*** (1.38, 2.26)	1.92** (1.22, 3.00)
Sexual violence	1.91*** (1.50, 2.44)	1.66*** (1.24, 2.24)	1.51** (1.12, 2.05)	1.53 (0.98, 2.41)
Health risk behavior
Smokingc	1.45** (1.11, 1.90)	0.99 (0.73, 1.34)	1.44** (1.13, 1.83)	1.80** (1.19, 2.72)
Drinking	1.41** (1.15, 1.73)	1.79*** (1.45, 2.20)	1.36** (1.09, 1.70)	1.19 (0.83, 1.71)
Marijuana	0.98 (0.75, 1.27)	0.78 (0.53, 1.15)	1.01 (0.75, 1.38)	0.92 (0.53, 1.61)
Hard drug used	1.44*** (1.18, 1.77)	1.84*** (1.39, 2.44)	1.76*** (1.37, 2.26)	1.93** (1.30, 2.86)
Hopelessness	4.37*** (3.70, 5.16)	3.17*** (2.50, 4.02)	4.37*** (3.44, 5.56)	3.36*** (2.31, 4.87)
Weighte	1.22** (1.08, 1.38)	1.02 (0.86, 1.22)	1.03 (0.84, 1.27)	0.92 (0.65, 1.31)
Males
Unadjusted	21 874	17 057	17 723	17 549
Sexual identity
Heterosexual (Ref)	1.00	1.00	1.00	1.00
Gay	3.40*** (2.19, 5.28)	2.06** (1.20, 3.56)	4.22*** (2.72, 6.55)	3.42*** (1.74, 6.72)
Bisexual	4.89*** (3.50, 6.82)	5.11*** (3.16, 8.25)	4.95*** (3.38, 7.25)	4.73*** (2.80, 8.01)
Unsure	2.73*** (1.94, 3.83)	2.48*** (1.65, 3.75)	3.06*** (2.20, 4.26)	3.94*** (2.70, 5.74)
Adjusted	16 080	12 534	13 332	13 276
Sexual identity
Heterosexual (Ref)	1.00	1.00	1.00	1.00
Gay	4.24*** (2.09, 8.59)	2.19* (1.07, 4.48)	2.38** (1.27, 4.45)	0.97 (0.29, 3.25)
Bisexual	2.97*** (1.83, 4.80)	2.56** (1.39, 4.71)	3.44*** (1.91, 6.22)	3.15** (1.43, 6.94)
Unsure	1.77* (1.09, 2.88)	0.93 (0.48, 1.79)	1.33 (0.66, 2.71)	2.32 (0.93, 5.77)
Risk factors
Threatened	1.06 (0.78, 1.46)	1.47* (1.04, 2.09)	1.53* (1.08, 2.17)	1.80* (1.04, 3.12)
In a fight	1.04 (0.84, 1.29)	1.20 (0.92, 1.56)	0.95 (0.69, 1.30)	0.95 (0.55, 1.64)
Physical dating violence	1.07 (0.76, 1.50)	1.05 (0.76, 1.46)	1.17 (0.81, 1.71)	1.50 (0.91, 2.48)
Sexual violence	2.51*** (1.46, 4.31)	2.24*** (1.42, 3.53)	2.41*** (1.56, 3.72)	2.46** (1.39, 4.35)
Health risk behavior
Smoking	0.89 (0.64, 1.24)	0.82 (0.55, 1.22)	0.99 (0.62, 1.59)	1.25 (0.63, 2.50)
Drinking	1.39** (1.08, 1.78)	1.07 (0.81, 1.41)	1.00 (0.70, 1.45)	0.55 (0.27, 1.11)
Marijuana	1.40* (1.01, 1.93)	1.43 (0.91, 2.24)	2.48** (1.39, 4.43)	5.52*** (2.25, 13.55)
Hard drug use	2.20*** (1.73, 2.79)	2.76*** (1.96, 3.89)	1.97*** (1.39, 2.79)	2.29** (1.22, 4.30)
Hopelessness	6.46*** (5.25, 7.93)	4.86*** (3.75, 6.30)	2.45*** (1.77, 3.39)	2.14** (1.33, 3.45)
Weighte	1.17 (0.89, 1.53)	1.01 (0.75, 1.35)	1.04 (0.78, 1.37)	0.77 (0.45, 1.32)

Note. CI = confidence interval; OR = odds ratio. Sites measuring sexual identity are Boston, MA; Chicago, IL; New York City, NY; and San Francisco, CA. Adjusted analyses control for race/ethnicity, grade, city, and year.

^aSample sizes differ because of missing data.

^bThreatened or injured with a weapon on school property 1 or more times in the past year.

^cSmoking, drinking, and marijuana use before aged 13 years.

^dHard drug use is a composite variable measured by lifetime use of any form of cocaine, inhalants, ecstasy, heroin, methamphetamines, steroid use without a prescription, or injection drug use.

^ePerceived overweight or obesity.

*P ≤ .05; **P ≤ .01; ***P ≤ .001.

In unadjusted analyses examining sexual orientation defined by sex of sexual contacts, we found that most youths with same-sex contact and both-sex contact compared with opposite-sex contact had increased odds of all SROs, with the ORs ranging from 1.56 (95% CI = 1.09, 2.21) for ideation among females with same-sex contact to 5.66 (95% CI = 3.23, 9.91) for MSAs among males with both-sex contact (Table 4). Females with same-sex contact did not vary in suicide planning compared with females with opposite-sex contact only. Not having sexual contact was associated with reduced odds of SROs except in the case of suicide planning among males, in which we found no statistical difference.

TABLE 4—

Unadjusted and Adjusted ORs of Suicide Risk Outcomes by Sex of Sexual Contacts: Local Youth Risk Behavior Survey Sites, United States, 2001–2009

	Suicide Ideation, No. or OR (95% CI)	Suicide Plans, No. or OR (95% CI)	Suicide Attempts, No. or OR (95% CI)	Medically Serious Attempts, No. or OR (95% CI)
Females
Unadjusted	22 720a	16 867	19 283	19 223
Sex of sexual contacts
Opposite sex (Ref)	1.00	1.00	1.00	1.00
Same sex	1.56* (1.09, 2.21)	1.19 (0.80, 1.78)	1.81*** (1.28, 2.56)	2.12* (1.16, 3.87)
Both sexes	2.49*** (2.02, 3.07)	2.80*** (2.09, 3.75)	2.57*** (1.99, 3.32)	2.63*** (1.89, 3.66)
No contact	0.57*** (0.51, 0.64)	0.57*** (0.49, 0.67)	0.53*** (0.45, 0.63)	0.51*** (0.37, 0.72)
Adjusted	17 904	13 207	15 474	15 448
Sex of sexual contacts
Opposite sex (Ref)	1.00	1.00	1.00	1.00
Same sex	1.53* (1.09, 2.14)	0.94 (0.61, 1.44)	2.03*** (1.37, 2.99)	1.31 (0.49, 3.50)
Both sexes	1.62** (1.21, 2.18)	1.85*** (1.31, 2.62)	1.63** (1.16, 2.31)	1.53* (1.02, 2.31)
No contact	0.75*** (0.64, 0.89)	0.66*** (0.52, 0.84)	0.78** (0.65, 0.93)	0.86 (0.56, 1.31)
Risk factors
Threatenedb	1.48*** (1.18, 1.85)	1.29 (0.98, 1.70)	1.69*** (1.28, 2.12)	2.73*** (1.91, 3.90)
In a fight	1.23** (1.06, 1.43)	1.31** (1.10, 1.57)	1.36** (1.13, 1.63)	1.41* (0.99, 2.02)
Physical dating violence	1.47*** (1.19, 1.82)	1.66*** (1.36, 2.03)	1.74*** (1.38, 2.20)	2.00*** (1.32, 3.01)
Sexual violence	1.80*** (1.43, 2.28)	1.64*** (1.24, 2.16)	1.53** (1.14, 2.05)	1.50 (0.98, 2.30)
Health risk behavior
Smokingc	1.44** (1.12, 1.85)	1.00 (0.75, 1.33)	1.47** (1.16, 1.86)	1.79** (1.21, 2.66)
Drinking	1.41*** (1.17, 1.70)	1.72*** (1.42, 2.08)	1.32** (1.08, 1.61)	1.16 (0.83, 1.62)
Marijuana	0.92 (0.73, 1.17)	0.75 (0.53, 1.05)	0.98 (0.74, 1.29)	0.96 (0.57, 1.60)
Hard drug used	1.49*** (1.23, 1.81)	1.81*** (1.40, 2.34)	1.82*** (1.44, 2.29)	1.92*** (1.31, 2.80)
Hopelessness	4.37*** (3.74, 5.11)	3.10*** (2.49, 3.85)	4.37*** (3.51, 5.44)	3.27*** (2.28, 4.69)
Weighte	1.26*** 1.13, 1.42)	1.07 (0.91, 1.25)	1.10 (0.91, 1.33)	1.01 (0.73, 1.41)
Males
Unadjusted	20 039	15 199	16 194	16 051
Sex of sexual contacts
Opposite sex (Ref)	1.00	1.00	1.00	1.00
Same sex	3.66*** (2.54, 5.26)	2.08* (1.19, 3.64)	2.83*** (2.00, 3.99)	2.06* (1.17, 3.62)
Both sexes	4.82*** (3.05, 7.62)	4.03*** (2.20, 7.39)	4.11*** (2.72, 6.20)	5.66*** (3.23, 9.91)
No contact	0.84* (0.70, 1.00)	0.80 (0.63, 1.01)	0.55*** (0.45, 0.68)	0.53*** (0.38, 0.72)
Adjusted	14 723	11 191	12 185	12 138
Sex of sexual contacts
Opposite sex (Ref)	1.00	1.00	1.00	1.00
Same sex	2.85*** (1.90, 4.29)	1.02 (0.53, 1.95)	1.58 (0.95, 2.65)	0.82 (0.32, 2.07)
Both sexes	2.64** (1.47, 4.73)	1.19 (0.57, 2.49)	2.18* (1.20, 3.96)	2.40* (1.16, 4.97)
No contact	1.09 (0.81, 1.46)	1.03 (0.79, 1.34)	0.76 (0.51, 1.13)	0.46* (0.23, 0.94)
Risk factors
Threatened	1.10 (0.84, 1.44)	1.55** (1.14, 2.10)	1.47* (1.05, 2.07)	1.83* (1.08, 3.10)
In a fight	0.99 (0.80, 1.23)	1.18 (0.90, 1.53)	0.89 (0.66, 1.22)	0.84 (0.49, 1.44)
Physical dating violence	1.13 (0.83, 1.54)	1.10 (0.82, 1.48)	1.04 (0.73, 1.49)	1.38 (0.87, 2.18)
Sexual violence	2.41*** (1.48, 3.94)	2.30*** (1.55, 3.42)	2.38*** (1.57, 3.61)	2.43*** (1.43, 4.13)
Health risk behavior
Smoking	1.02 (0.75, 1.38)	0.98 (0.71, 1.34)	1.05 (0.68, 1.63)	1.12 (0.60, 2.08)
Drinking	1.35* (1.07, 1.71)	1.08 (0.83, 1.40)	1.04 (0.74, 1.47)	0.65 (0.33, 1.26)
Marijuana	1.33 (0.99, 1.78)	1.39 (0.94, 2.07)	2.08** (1.23, 3.52)	4.22*** (1.86, 9.56)
Hard drug use	2.25*** (1.80, 2.82)	2.86*** (2.01, 4.09)	2.03*** (1.45, 2.87)	2.24** (1.23, 4.09)
Hopelessness	6.92*** (5.69, 8.41)	5.20*** (4.01, 6.74)	2.56*** (1.87, 3.50)	2.24*** (1.43, 3.51)
Weighte	1.22 (0.97, 1.54)	1.17 (0.89, 1.53)	1.08 (0.82, 1.42)	0.74 (0.46, 1.19)

Note. CI = confidence interval; OR = odds ratio. Sites measuring sex of sexual contacts are Boston, MA; Chicago, IL; New York City, NY; and San Diego, CA. Adjusted analyses control for race/ethnicity, grade, city, and year.

^aSample sizes differ because of missing data.

^bThreatened or injured with a weapon on school property 1 or more times in the past year.

^cSmoking, drinking, and marijuana use before aged 13 years.

^dHard drug use is a composite variable measured by lifetime use of any form of cocaine, inhalants, ecstasy, heroin, methamphetamines, steroid use without a prescription, or injection drug use.

^ePerceived overweight or obesity.

*P ≤ .05; **P ≤ .01; ***P ≤ .001.

In adjusted analyses, the pattern of risk changed, most notably for males. Neither males with same-sex contact nor those with both-sex contact differed significantly in their odds of planning a suicide compared with males with opposite-sex contact only. Males with same-sex contact also did not vary in their odds of attempts or MSAs. The protective effect of not having sexual contact remained only for MSAs (OR = 0.46; 95% CI = 0.23, 0.94) among males. Among females, the increased odds of ideation, plans, and attempts found among youths with same-sex contact and both-sex contact in unadjusted analyses remained in adjusted analyses. However, females with same-sex contact no longer had increased odds of MSAs. Unlike for males, the protective effect of not having sexual contact remained in adjusted analyses, except as pertaining to MSAs, for which it no longer remained (OR = 0.86; 95% CI = 0.56, 1.31). Except for suicide planning among males, all youths with both-sex contact had increased odds of all SROs compared with youths with opposite-sex contact only. Similar to results on the basis of sexual identity, racial/ethnic minority groups had increased risk for suicide attempts compared with White non-Hispanic youths (results not shown). The pattern of risk factors significantly associated with SROs among males and females was largely the same on the basis of sex of sexual contacts as on the basis of sexual identity.

DISCUSSION

We examined the risk of suicide ideation, plans, attempts, and MSAs among a diverse sample of urban high school youths between 2001 and 2009. In response to our first research question, we found that when accounting for all other factors and compared with heterosexuals, LGB youths had significantly increased odds of each SRO, except lesbians did not have increased odds of planning and gay males did not have increased odds of MSAs. Unsure males had significantly elevated odds of suicidal ideation.

We found mixed results in response to our second question, which was focused on sex of sexual contacts. All youths with same-sex contact and both-sex contact had increased odds of suicidal ideation; however, the odds of suicide plans, attempts, and MSAs varied among SMY subgroups, with youths with both-sex contact having increased odds of all SROs except for planning among males, which was not elevated. Females with same-sex contact and both-sex contact had more consistent elevations in odds of SROs than did males.

Our results suggesting increased associations between SMYs and suicidal ideation and attempts concur with results from other adolescent and adult population-based studies that measured sexual orientation on the basis of sexual identity, sex of sexual contacts, and sexual attraction.^{11,32,50,54,57} The associations between SMYs and suicide planning and MSAs are less robust, with more consistently elevated associations found among bisexuals and youths with both-sex contact.

Only 4 other adolescent population-based studies, to our knowledge, have measured suicide planning among SMYs.^37,45–47 Using YRBS data from San Francisco, researchers found that victimization moderated the effect of sexual orientation on suicide planning among heterosexuals but not among LGB.⁴⁵ Similarly, in unadjusted analyses, Faulkner et al. did not find increased planning among students with any same-sex contact compared with opposite-sex contact only.³⁷ A recent Centers for Disease Control and Prevention study found higher prevalence rates of suicide planning among SMYs than among non-SMYs.⁴⁹ The small number of studies assessing suicide planning and their mixed results supports the need for further research. Although some clinical studies suggest that planning indicates increased suicide intent,⁵⁸ other research indicates that adolescent suicides and suicide attempts are often impulsive.^59–61

MSAs are perhaps the greatest risk indicator of future suicide considering the potential associated intent.^58,62 Our results suggesting elevated odds of MSAs among lesbians, bisexual youths, and those with both-sex contact concur in part with the only known population-based studies examining this association. A study of early Massachusetts YRBS data found that in unadjusted analyses youths with same-sex contact versus opposite-sex contact had increased odds only of MSAs.³⁷ Using data from both the Massachusetts and Vermont YRBSs, researchers found increased odds of MSAs only among youths with both-sex contacts, not same-sex contacts.¹⁶ Finally, a Rhode Island study found double the odds of MSAs among LGB youths.⁴⁷ Together, these results suggest that youths with both-sex contact may have increased minority stress or less support in the face of minority stress than do youths with same-sex contact.¹⁷

Limitations

The following limitations should be taken into consideration. First, regarding sexual orientation measurement, although we used 2 measures, we could not draw any conclusions about which of them may be the more important marker of suicide risk. Some research suggests that sexual identity is a more important predictor of risk than are sexual attraction or behavior.⁶³ We were unable to test this hypothesis in the current set of analyses, as not all sites included both measures. Three local sites (Boston, Chicago, and New York City) included both measures of sexual orientation, so future research that uses data from the 3 sites and compares sexual identity, sexual behavior, and a new measure combining information from sexual identity and sexual behavior as predictors of suicide risk would shed light on this issue. Another sexual orientation measurement issue is that neither measure captured transgender youths or youths who did not identify with any of the labels provided.⁶⁴

Second, we did not control for social factors, such as school and community climate or connectedness, which may buffer the effect of sexual minority stress.^65,66 Unfortunately, these measures were either not collected or not consistently available across sites and years. Third, some risk factors (e.g., being threatened, hopelessness) included in our analyses may mediate the associations between sexual orientation and SROs, but we cannot rigorously test the mediation in this study because we do not know the exact time sequences of those risk factors compared with the timing of sexual orientation and SROs. Future research (e.g., well-designed longitudinal research) is needed to further investigate the mediation issue. Fourth, missing data among males and younger students may have led to overestimation of results. Fifth, results are subject to self-report bias, and the repeated cross-sectional study design does not allow us to infer causation.

Strengths

We analyzed data from numerous angles, including by sexual identity, sex of sexual contacts, and gender. This enabled us to make finer distinctions of risk within the heterogeneous population of SMYs instead of masking effects by dichotomizing sexual orientation. Although aggregating data did not allow site-specific inferences, it did provide greater statistical power to identify and better understand the complex relationship between sexual orientation subgroups and the range of SROs not typically studied, including suicide planning and MSAs.

This large sample size also allowed us to examine relationships while taking into account a range of risk factors. Finally, our study results suggest that the risk present among SMYs found in nationally representative samples also occurs at the local level. This is important because local data can be more specific to unique manifestations of the outcomes, allowing more tailored, more relevant, and potentially more effective prevention activities.⁶⁷

Implications

These findings showing increased risk of SROs among SMYs regardless of the measure used and controlling for risk factors have important implications for adolescent suicide prevention programs and practices. SMYs have increased stress, including discrimination, victimization, and internalized homophobia.¹⁸ Some SMYs may respond to these stressors in unhealthy ways (e.g., drug and alcohol use) or isolate themselves. These behaviors further increase risk of suicide.

Greater awareness among families, schools, providers, and other adults of SMYs’ stressors and how to respond may help reduce SROs. Additionally, prevention efforts focused on increased tolerance and decreased stigma may be beneficial across all levels of the social ecology, between individuals, among families and communities, and across the broader society and social institutions. Schools and communities can promote protective factors such as positive youth engagement, help seeking, and conflict resolution skills. Fostering connectedness between individuals, families, schools, and SMY-serving organizations may help promote resilience and reduce risk of SROs for all youths.^{33,54,65,68,69}

Others have recommended improved training for school staff that takes into account youths’ sexual orientation, better defined clinical guidelines sensitive to issues of SMYs, and more inclusive social policies in schools and communities to reduce risk for SROs among SMYs.^29,70 Prevention development and evaluation and research that takes into account overlapping risk factors are recommended. Finally, expansion of national-, state-, and local-level surveillance of sexual orientation, SROs, and protective factors can help tailor local prevention efforts, monitor program effectiveness, and track efforts to decrease health disparities among SMYs more generally.