Religious Climate and Health Risk Behaviors in Sexual Minority Youths: A Population-Based Study

Abstract

Objectives. We examined whether the health risk behaviors of lesbian, gay, and bisexual (LGB) youths are determined in part by the religious composition of the communities in which they live.

Methods. Data were collected from 31 852 high school students, including 1413 LGB students, who participated in the Oregon Healthy Teens survey in 2006 through 2008. Supportive religious climate was operationalized according to the proportion of individuals (of the total number of religious adherents) who adhere to a religion supporting homosexuality. Comprehensive data on religious climate were derived from 85 denominational groups in 34 Oregon counties.

Results. Among LGB youths, living in a county with a religious climate that was supportive of homosexuality was associated with significantly fewer alcohol abuse symptoms (odds ratio [OR] = 0.58; 95% confidence interval [CI] = 0.40, 0.85) and fewer sexual partners (OR = 0.77; 95% CI = 0.60, 0.99). The effect of religious climate on health behaviors was stronger among LGB than heterosexual youths. Results remained robust after adjustment for multiple confounding factors.

Conclusions. The religious climate surrounding LGB youths may serve as a determinant of their health risk behaviors.

Stigma operates through both discrete events and pervasive environments in threatening the health of its targets.¹ Explanations of lesbian, gay, and bisexual (LGB) youths’ higher level of engagement in health risk behaviors relative to their heterosexual peers have largely relied on reports of encounters with discrete stigmatizing events, such as parental rejection and peer victimization.² Yet, when social structures, institutions, and norms transmit disapproval or invalidation of LGB identities and life experiences, they have the potential to threaten LGB youths’ health independent of individual-level events.^1,3,4

The social climate surrounding LGB individuals both worldwide⁵ and in the United States⁶ is shaped by the moral debate regarding the legitimacy of homosexuality. Given that social ideology is closely intertwined with religious ideology in the United States, expressed attitudes toward homosexuality are largely aligned with degree and type of religious affiliation.^6,7 A majority of US residents report that advancement of LGB rights, such as the legal recognition of same-sex marriage, clashes with their religious beliefs.⁸ Approximately 85% of individuals in the United States identify as religious, and more than half of the US population in 2003 believed that homosexuality is sinful.⁹

Although condemnation of homosexuality has long been a part of many religious doctrines, not all religions hold similar beliefs toward homosexuality.¹⁰ Indeed, a 2011 survey of the US population showed that about half of religious adherents believe that society should accept homosexuality,¹¹ highlighting a shift from the results of earlier studies and mirroring the changing attitudes toward same-sex marriage that also exist across some religious groups.¹² In all, the US public's view of homosexuality is changing and varies as a function of several characteristics, including religious affiliation.

LGB youths construct their sexual identities within social climates shaped by these religious influences, with some LGB youths encountering more threatening climates than others given that religious demography varies widely by geographical region.¹³ The health risk behaviors of LGB youths may thus vary according to the religious composition of the communities in which they live, in that encountered stigma influences health-impairing coping behaviors.¹⁴ However, existing research has focused almost exclusively on associations between personal religiosity and health behaviors among LGB youths.^15–17 Thus, research on the intersection of religious climate and the health behaviors of LGB youths remains limited.

To address this gap in the literature, we used an objective index of community religiosity to predict tobacco use, alcohol abuse, and sexual behaviors in a sample of LGB and heterosexual youths. We hypothesized that religious climate would be associated with these health outcomes over and above the influence of other environmental factors surrounding LGB youths (e.g., school bullying policies, presence of Gay-Straight Alliances) and that the association would be stronger for LGB youths than for heterosexual youths. We also expected that religious climate would exert these health effects independent of established individual-level psychosocial predictors of health risk behaviors.

METHODS

We derived our data from the Oregon Healthy Teens (OHT) study. Annual OHT surveys are administered to more than one third of Oregon's 8th- and 11th-grade public school students. Sexual orientation is assessed only in the survey for 11th graders. We pooled data from the years 2006 (when sexual orientation was first assessed) to 2008 (the most recent data available) to increase the sample size of LGB participants. Participating students came from 297 schools in 34 counties. The questionnaire was available in both English and Spanish. All participants were assured that the survey was anonymous and voluntary, and parents provided passive consent for their children to participate.

Measures

Data on demographic characteristics including gender and race/ethnicity were obtained via self-report. Sexual orientation was assessed with a single item asking respondents to indicate “which of the following best describes you.” Four response options were given: (1) heterosexual (straight), (2) gay or lesbian, (3) bisexual, and (4) not sure. Of the 33 714 respondents, 30 439 (90.3%) self-identified as heterosexual, 301 (0.9%) self-identified as gay or lesbian, and 1112 (3.3%) self-identified as bisexual. We excluded 653 (1.9%) participants who indicated that they were “not sure” about their sexual orientation, consistent with previous studies.¹⁸ An additional 1209 respondents did not complete the sexual orientation item and were excluded from the analyses. Consequently, the final sample size was 31 852. Sociodemographic data for the 2006 through 2008 OHT sample have been described in detail elsewhere.¹⁹

Independent variable. We created a religious climate variable that measured the degree of support for homosexuality across the 34 counties included in the OHT survey. We obtained data from the Glenmary Research Center,²⁰ which collects information on religious affiliation in the United States each census year. The 2000 data set included information on religious adherents (defined as all full members, their children, and others who regularly attend services or participate in the congregation) at the county level for 85 religious groups in Oregon. These groups’ official denominational Web sites were searched for information about their stance toward homosexuality. In addition, we examined doctrinal statements, resolutions, and position statements on homosexuality and determined whether the denomination blessed same-sex unions (or made reference to only heterosexual unions) and allowed ordinations of gay clergy.

We created a 5-level variable indexing each denomination's position toward homosexuality. The coding scheme was as follows:

Denominations having the most negative view of homosexuality, with links to organizations that promote reorientation therapies (e.g., Exodus International) or political involvement in legislation restricting protections to LGB individuals (e.g., National Organization for Marriage), were coded as 0.

Denominations with explicit reference to homosexuality as sinful or not compatible with God's plan or reference to marriage as being between a man and woman only were coded as 1.

Neutral denominations, as evident by no official position or belief included in denominational statements, were coded as 2.

Denominations that are relatively supportive of homosexuality, but with evidence of division on the issue of gay ordinations and performing same-sex marriages, were coded as 3.

Denominations that are universally accepting and explicitly affirming of gay identities and relationships and that allow gay ordination were coded as 4.

Two coders with expertise in religion and sexuality independently coded each religious denomination (interrater reliability = 0.93).

Denominations receiving a score of 0 to 2 (e.g., Church of Jesus Christ Latter-Day Saints, Church of God in Christ, Assemblies of God) were considered unsupportive. Supportive denominations, those with a score of 3 or 4, included Unitarian/Universalists, Quakers, Presbyterian Church (U.S.A.), United Methodist Church, Episcopal, and Community of Christ. We summed the number of religious adherents in the supportive denominations and created a variable of the proportion of adherents to these denominations (of the total number of religious adherents) in each county. This variable was mean standardized, and the scores ranged from −1.565 to 2.814, indicating significant variation in religious climate (a full description of the coding for the 85 denominations is available on request). Religious climate was dichotomized at the median (0 = low support, 1 = high support).

Outcome variables. First, participants were asked about the frequency of 5 symptoms of alcohol abuse in the preceding 12 months:

1. missing school or class because of drinking alcohol,

2. getting sick to your stomach because of drinking alcohol,

3. not being able to remember what happened while you were drinking alcohol,

4. later regretting something you did while drinking alcohol, and

5. worried that you drank alcohol too much or too often.

Response options were 0 times, 1 or 2 times, 3 to 5 times, 6 to 9 times, and 10 or more times. Responses to the 5 items were summed, and we created a dichotomous variable comparing the presence and absence of alcohol abuse symptoms.

Second, participants were asked the number of days they had smoked cigarettes during the preceding 30 days. Response options were 0 days, 1 or 2 days, 3 to 5 days, 6 to 9 days, 10 to 19 days, 20 to 29 days, and all 30 days. We created a dichotomous variable in which any smoking was compared with no smoking. Third, participants were asked about the number of sexual partners in their lifetime; we created a dichotomous variable comparing no sexual partners and at least one sexual partner. The 3 health behavior measures just described have shown excellent test–retest reliability.^21,22

Community-level covariates. Because religious climate is correlated with the broader community climate surrounding LGB individuals,²³ we wanted to explore the role of community climate in confounding any observed association between religious climate and health behaviors. To accomplish this objective, we used a variable that was recently developed for measuring the community climate surrounding LGB youths.¹⁹

This variable was composed of 5 items:

1. proportion of registered Democrats in each county (data were obtained from the Oregon Secretary of State Election Division),

2. proportion of same-sex partner households by county (from the US Census),

3. proportion of school districts within each of the counties that had anti-bullying policies inclusive of sexual orientation (from the Oregon Department of Education),

4. proportion of school districts within each of the counties that had anti-discrimination policies related to sexual orientation (from the Oregon Department of Education), and

5. proportion of schools within each district that had a Gay-Straight Alliance (from the Gay and Lesbian Education Network).

Previous research has shown that this community climate variable is associated with suicide attempts among LGB youths.¹⁹

A principal components factor analysis indicated that these 5 items loaded onto a single factor (factor loadings ranged from 0.50–0.85); consequently, the items were standardized and summed to create a total score, which was dichotomized at the median. Importantly, the factor analysis also indicated that the measure of religious climate did not strongly load onto the other 5 factors (the factor loading for religious climate was 0.15), suggesting that it is distinct from general community climate. In addition to this measure of the broader community climate surrounding LGB youths, we controlled for the population size of each county (data were obtained from the 2000 US Census), given established associations between urbanicity and attitudes toward gays and lesbians²⁴ as well as health behaviors.²⁵

Individual-level covariates. The OHT survey included several measures of well-established predictors of substance use and sexual behaviors at the individual level. Depressive symptoms were assessed with the following question: “During the past 12 months, did you ever feel so sad or hopeless almost every day for 2 weeks or more in a row that you stopped doing some usual activities?” Physical abuse was assessed with the question “During your life, has any adult ever intentionally hit or physically hurt you?” Finally, sexual abuse was assessed with the question “During your life, has any adult ever had sexual contact with you?” Participants indicated either the presence or absence of depressive symptoms, physical abuse, and sexual abuse.

In addition, the OHT survey asked questions assessing individual-level risk factors specific to the use of substances, including injunctive norms for substance use,²⁶ perceptions of risks related to substance use,²⁷ and peer substance use.²⁸ Injunctive norms were measured with a single item:

How wrong do you think it is for someone your age to: (1) drink beer, wine, or hard liquor (e.g., vodka, whiskey, or gin) regularly and (2) smoke cigarettes.

Responses were dichotomized (1 = very wrong, 0 = wrong, a little bit wrong, or not wrong at all).

Perceived risk was assessed with the question

How much do you think people risk harming themselves (physically or in other ways) if they: (1) smoke one or more packs of cigarettes per day and (2) take one or two drinks of an alcoholic beverage (beer, wine, liquor) nearly every day.

Again, responses were dichotomized (1 = great risk, 0 = no risk, slight risk, or moderate risk).

Finally, peer substance use was assessed as follows:

Think of your best friends (the friends you feel closest to). In the past 12 months, how many of your best friends have: (1) smoked cigarettes and (2) tried beer, wine, or hard liquor (e.g., vodka, whiskey, or gin)?

Responses ranged from 0 to 4; we created a dichotomous variable comparing no friends using cigarettes or alcohol versus at least one friend.

Statistical Analysis

The analytic strategy consisted of 3 steps. First, we used basic descriptive cross tabulations to test for group differences in substance use, sexual behaviors, and associated risk factors between LGB and heterosexual youths. Second, we examined whether the effect of religious climate on health risk behaviors varied as a function of sexual orientation status by using a multiplicative interaction term to test the 2-way interaction between sexual orientation and religious climate.

Finally, we used generalized estimating equations (GEE) methodology to test whether religious climate remained significantly associated with health behaviors after adjustment for multiple individual- and community-level risk factors.²⁹ GEE is a method developed for handling clustered data in which the observations within each cluster are correlated with each other. Given that OHT survey respondents were nested within their county of residence, we used GEE to account for the correlations among observations from each individual within the same county. We ran 3 separate models for tobacco use, alcohol abuse symptoms, and sexual risk behaviors. Given power considerations, we combined lesbian and gay youths with bisexual youths, as well as boys and girls, similar to other studies.³⁰

RESULTS

As can be seen in Table 1, rates of tobacco use (F₁ = 539.43; P < .001) and alcohol abuse symptoms (F₁ = 13.48; P < .001) were significantly higher among LGB respondents than among heterosexual youths, and LGB youths had more sexual partners than did heterosexual youths (F₁ = 519.09; P < .001). The magnitude of these disparities was considerably larger for tobacco and sexual partners than for alcohol abuse symptoms. Also, with the exception of perceived risk of drinking and smoking, rates of all established risk factors for these health behaviors were significantly higher among LGB youths.

TABLE 1—

Sexual Orientation–Related Disparities in Substance Use, Sexual Behaviors, and Associated Risk Factors: Oregon Healthy Teens Study, 2006–2008

Outcome	Lesbian, Gay, and Bisexual (n = 1413), Mean ±SD or No. (%)	Heterosexual (n = 30 439), Mean ±SD or No. (%)	Group Difference F (df = 1)
No. of d smoked cigarettes in past 30 d (range = 0–6)	1.43 ±2.17	0.51 ±1.41	539.43*
No. of alcohol abuse symptoms in past 12 mo (range = 0–25)	4.75 ±3.74	4.40 ±3.27	13.48*
No. of lifetime sexual partners (range = 0–6)	2.19 ±2.19	1.10 ±1.67	519.09*
Depressive symptoms	607 (43.0)	5192 (17.1)	653.70*
Lifetime sexual abuse	420 (29.7)	2465 (8.1)	832.25*
Lifetime physical abuse	657 (46.5)	8353 (27.4)	283.58*
No perceived risk of drinking	199 (14.1)	4063 (13.3)	1.552
≥ 1 friend drinks	918 (65.0)	18 279 (60.1)	30.40*
No injunctive norms against drinking	167 (11.8)	2776 (9.1)	13.73*
No perceived risk of smoking	100 (7.1)	2310 (7.6)	0.158
≥ 1 friend smokes	682 (48.3)	11 516 (37.8)	83.91*
No injunctive norms against smoking	688 (48.7)	13 378 (44.0)	20.06*

*P < .001.

\After controlling for community- and individual-level covariates, the interaction between LGB status and religious climate was statistically significant for any alcohol abuse symptoms (odds ratio [OR] = 0.63; 95% confidence interval [CI] = 0.43, 0.92) and any sexual partners (OR = 0.87; 95% CI = 0.76, 0.98) but not for any tobacco use (P > .05). Given the significant interactions for alcohol abuse symptoms and sexual partners, we stratified the sample by LGB status to examine the main effect of religious climate on these outcomes.

As can be seen in Figure 1, LGB youths living in counties with a religious climate that is less supportive of homosexuality had higher levels of alcohol abuse symptoms and more sexual partners than did LGB youths living in supportive religious climates. These results were also observed for heterosexual youths, although the effect of religious climate on alcohol abuse and sexual partners was stronger among LGB than heterosexual youths, as indicated by the significant interactions.

FIGURE 1—

Relationship between religious climate and health risk behaviors: Oregon Healthy Teens Study, 2006–2008.

Note. Hetero = heterosexual; LGB = lesbian, gay, and bisexual; RC = religious climate. Religious climate was divided at the median (low = low support, high = high support).

We next examined whether the associations between religious climate and health behaviors remained significant after adjustment for multiple community- and individual-level risk factors (Tables 2 and 3). Among LGB youths, a supportive religious climate was significantly associated with fewer alcohol abuse symptoms (OR = 0.58; 95% CI = 0.40, 0.85) and fewer sexual partners (OR = 0.77; 95% CI = 0.60, 0.99) after control for established risk factors. Similarly, among heterosexual youths, a supportive religious climate was significantly associated with fewer alcohol abuse symptoms (OR = 0.83; 95% CI = 0.77, 0.89) and fewer sexual partners (OR = 0.89; 95% CI = 0.85, 0.94). There was no main effect of religious climate on tobacco use for either sexual orientation group (data not shown).

TABLE 2—

Associations Between Religious Climate and Health Risk Behaviors Among Lesbian, Gay, and Bisexual Youths: Oregon Healthy Teens Study, 2006–2008

	Alcohol Abuse Symptoms		Sexual Partners
	OR (95% CI)	P	OR (95% CI)	P
Religious climate	0.58 (0.40, 0.85)	.005	0.77 (0.60, 0.99)	.044
Community climate	1.35 (0.87, 2.10)	.186	1.03 (0.76, 1.38)	.872
Population size	1.14 (0.74, 1.76)	.546	0.95 (0.71, 1.27)	.72
Sexa	0.64 (0.43, 0.95)	.027	0.83 (0.63, 1.08)	.166
Race/ethnicityb	1.21 (0.78, 1.86)	.399	0.79 (0.58, 1.07)	.134
Depressive symptoms	1.18 (0.79, 1.75)	.429	1.02 (0.79, 1.33)	.86
Physical abuse	1.71 (1.15, 2.56)	.009	1.34 (1.04, 1.74)	.026
Sexual abuse	1.35 (0.84, 2.17)	.212	2.95 (2.15, 4.05)	<.001
Peer substance use	1.88 (1.26, 2.82)	.002	…	…
Injunctive norms	2.31 (1.07, 4.98)	.032	…	…
Perceived risk	1.30 (0.74, 2.29)	.365	…	…

Note. CI = confidence interval; OR = odds ratio. Shown are the results of a final generalized estimating equations model predicting behavioral and sexual health outcomes.

^aFemale = 0, male = 1.

^bNon-White = 0, White = 1.

TABLE 3—

Associations Between Religious Climate and Health Risk Behaviors Among Heterosexual Youths: Oregon Healthy Teens Study, 2006–2008

	Alcohol Abuse Symptoms		Sexual Partners
	OR (95% CI)	P	OR (95% CI)	P
Religious climate	0.83 (0.77, 0.89)	<.001	0.89 (0.85, 0.94)	<.001
Community climate	0.73 (0.67, 0.79)	<.001	0.91 (0.85, 0.96)	<.001
Population size	1.24 (1.14, 1.35)	<.001	0.83 (0.78, 0.88)	<.001
Sexa	1.03 (0.96, 1.11)	.409	0.96 (0.91, 1.01)	.109
Race/ethnicityb	0.91 (0.83, 0.99)	.027	0.92 (0.87, 0.98)	.006
Depressive symptoms	1.09 (0.99, 1.21)	.081	1.49 (1.39, 1.59)	<.001
Physical abuse	1.25 (1.14, 1.36)	<.001	1.57 (1.48, 1.66)	<.001
Sexual abuse	1.31 (1.12, 1.52)	<.001	3.36 (3.03, 3.72)	<.001
Peer substance use	2.85 (2.65, 3.06)	<.001	…	…
Injunctive norms	2.21 (1.88, 2.60)	<.001	…	…
Perceived risk	1.02 (0.93, 1.13)	.638	…	…

Note. CI = confidence interval; OR = odds ratio. Shown are the results of a final generalized estimating equations model predicting behavioral and sexual health outcomes.

^aFemale = 0, male = 1.

^bNon-White = 0, White = 1.

DISCUSSION

The central finding of this study is that the religious climate surrounding LGB youths serves as a determinant of their health risk behaviors. Indeed, religious climate remained a significant correlate of alcohol abuse symptoms and sexual partners for LGB youths even after adjustment for multiple potential confounders at both the individual and community levels. Although religious climate was also associated with health behaviors among heterosexual youths, it was more strongly associated with the health behaviors of LGB youths.

Previous research has documented that LGB youths have higher rates of tobacco use, alcohol abuse symptoms, and unprotected sexual intercourse than their heterosexual peers.² Although negative aspects of the social environment may explain these health disparities, existing research with LGB youths has rarely addressed the influence of social determinants of health that are measured at the ecological level. This study therefore provides novel information regarding how specific characteristics of the social context in which LGB youths reside (namely, the religious climate) may contribute to health disparities related to sexual orientation.

Extant data on the relationship between religion and health risk behaviors demonstrate that personal religiosity does not protect against tobacco use, alcohol consumption, or risky sexual behavior among LGB youths, as it consistently does for heterosexual youths.^15–17 By contrast, our results showed that LGB youths living in counties with more supportive religious climates exhibited fewer health risk behaviors, indicating that religion can be protective for LGB youths. One possible reason for the discrepant findings is that whereas previous studies have examined religion as an individual-level measure of religious belief and adherence (i.e., personal religiosity), we investigated religious climate as a contextual-level measure of the social environment. Unfortunately, the OHT survey does not include measures of personal religiosity, so we were unable to examine relationships between personal religiosity and religious climate.

Exploring the religious contexts under which personal religiosity either promotes or undermines LGB health represents an important avenue for further inquiry. Future studies should also examine the extent to which ongoing demographic changes in religious attitudes toward homosexuality¹¹ influence the health of LGB youths. Population-based studies of US adolescents began including measures of sexual orientation in the 1990s. Consequently, age–period–cohort analyses can be used to examine whether increasingly tolerant religious climates are associated with a concurrent reduction in health disparities related to sexual orientation.

Limitations

Although measuring the religious climate at the ecological level represents a novel contribution to the literature on social determinants of sexual minority health, our measure has a number of limitations. First, we did not have access to the individual attitudes of the members included in the religious data set; instead, we used an aggregate measure of attitudes based on denominations’ official stance toward homosexuality. Given the diversity of attitudes toward homosexuality within religious denominations, our measure could be subject to misclassification.

Second, we used national, rather than regional, data on religious denominations’ attitudes toward homosexuality. There is likely to be local variation in religious attitudes and doctrines, which could have led to some misclassification of religious denominations. However, when attempting to acquire information on positions toward homosexuality from the Web sites of local denominations in Oregon, we found that most provided links to the national denomination, suggesting similarity of beliefs. Moreover, national denominational positions are likely to be a valid indicator among highly conservative denominations, which have considerably less variability on the issue of homosexuality.³¹

Third, the religious data set did not allow for Jewish congregations to be divided into separate categories (e.g., Reform vs Orthodox), despite the fact that Jewish congregations differ with respect to their views on homosexuality. In addition, the religious data set undercounted African American denominations. However, because African American congregations largely (although not uniformly) express negative attitudes toward homosexuality,³² this undercount would bias our results toward the null.

Fourth, the most recently available data on religious denominations were from 2000, whereas health outcomes were measured from 2006 to 2008. If religious adherence changed significantly either within or between denominations during that period, this could have introduced misclassification. Researchers can use more recent data on religious climate and youth health to test the robustness of the current results.

In addition to measurement limitations, our data describe only one US state, which may restrict generalizability. However, most states resemble Oregon in terms of having wide variation in religious climate, suggesting that similar results may be observed in other geographic areas. Nevertheless, our findings require replication across diverse social contexts. Finally, our data were cross-sectional, precluding the ability to establish causal inferences.

Strengths

Our study also included several important strengths. Using an objective measure of religious climate at the ecological level avoids the methodological limitations of existing studies on environmental risk factors and health risk behaviors among LGB individuals, which have relied on individual-level measurements of the environment (e.g., perceived discrimination).³³ These measures are confounded with health status and therefore may lead to biased estimates of the association between social climate and LGB health.³⁴

In addition, given that youths cannot select into different social environments as easily as adults, the present study largely avoids the possibility of attributing any association between religious environment and health risk behaviors to LGB youths selecting into environments that encourage such behaviors. Finally, our creation of an objective index of community religiosity that was linked to individual health outcomes circumvents the ecological fallacy,³⁵ or the methodological limitation of drawing inferences about the effect of ecological influences on health when the measures of health rely solely on aggregated reports.

Conclusions

In addition to these methodological strengths, our study contributes to a nascent body of work suggesting that social factors measured at the ecological level exert demonstrable effects on the health of LGB populations.^3,19 This research could ultimately aid public health efforts to reduce health disparities among LGB adolescents by providing information regarding the targeting of prevention intervention programs to LGB youths living in higher-risk environments.

In particular, the health of LGB youths who live in unsupportive religious climates can potentially be protected by psychosocial services aimed at facilitating an optimal integration of sexual identities and religious beliefs³⁶ or by affirmative school climates that include Gay-Straight Alliances, for example.¹⁹ Developing interventions that diminish the negative psychosocial consequences of living in social environments that stigmatize homosexuality remains an important public health priority.