Is it getting better? An analytical method to test trends in health disparities, with tobacco use among sexual minority vs. heterosexual youth as an example
© The Author(s). 2016
Received: 16 October 2015
Accepted: 19 May 2016
Published: 23 May 2016
Previous studies have documented higher health risks for lesbian, gay, and bisexual youth compared to heterosexual youth. However, none has reported whether the sexual orientation-based gaps have widened, narrowed, or remained unchanged over time. The purpose of this study was to develop a way to test differences in trends between sexual minority and heterosexual youth cohorts in population-based studies, with cigarette smoking as an exemplar.
We analysed the Minnesota Student Survey of 1998–2010, a repeated, cross-sectional census of adolescent health in grades 9 and 12. Our sample was students with recent sexual experience (Ns = 17,376–19,617). Sexual orientation was measured by gender of sexual partners in the past 12 months: students with only opposite-gender partner(s) (OPPOS), students with both male and female partners (BOTH), students with only same-gender partner(s) (SAME). We used logistic regressions to examine trends in prevalence of past-month cigarette smoking from 1998 to 2010, separately for each orientation group. We then applied novel interaction analyses to test whether disparities in smoking prevalence between OPPOS and SAME/BOTH changed over time.
Recent smoking rates decreased over time among all orientation groups. BOTH adolescents were more likely than OPPOS adolescents to report past 30-day smoking, but there were no significant differences between SAME adolescents and OPPOS adolescents. Year-by-orientation interactions indicated the gap between BOTH adolescents and OPPOS adolescents widened from 1998 to 2004, then persisted between 2004 and 2010. No significant interaction effects were observed between SAME adolescents and OPPOS adolescents.
All orientation groups had decreasing trends in recent cigarette smoking; however, disparities in smoking rates remain between heterosexual adolescents and bisexual adolescents. These results provide a new method of not just documenting trends within minority groups, but examining whether health equity is improving for them compared to dominant groups.
A substantial body of population-based evidence in North America, Europe, and elsewhere has documented significant health disparities for lesbian, gay, and bisexual youth (sexual minority or LGB youth) compared to their heterosexual peers . They experience disproportionately higher rates of discrimination and violence exposure ; higher prevalence of alcohol and other drug use [3–5]; equal or higher rates of sexual health behaviours, including adolescent pregnancy involvement [6, 7] and for some, greater risk of sexually transmitted infections such as HIV [8, 9]; and higher rates of depression and suicide attempts . Sexual minority youth have also been more likely to report tobacco use than their peers in studies in the USA [4, 11], in the UK  and in Canada [13, 14].
Much of the research examining these health disparities links them to stress as a result of stigma, discrimination, and violence targeted toward sexual minority peoples . Yet the past decade has seen sweeping improvements in the social and legal status of sexual minority populations in a number of countries, including changes in human rights law  and in legalization of same-sex marriage . There have even been safe school policies enacted to reduce homophobic harassment and bullying among youth in schools . This raises the question: is it getting better? Are these health risks beginning to improve among LGB youth? There are a limited number of regularly repeating population-based adolescent health surveys that include measures of sexual orientation, and have been repeated over a long enough time period to begin to document trends in health and risk behaviours [6, 18]. Declines in health-compromising behaviours or improvements in health outcomes among sexual minority groups, while good news in themselves, cannot tell us whether the disparities between them and heterosexual youth are also improving. For example, tobacco use has been declining among adolescents and adults throughout North America and much of Europe over the past decade , so if there are declines in tobacco use among LGB youth, they may be matched by similar declines among heterosexual youth, and thus the gap between them remains. Alternately, tobacco use may be declining at a slower rate for sexual minority youth, widening the gap between them and heterosexual youth, or at a faster rate, narrowing the gap in health equity. Our search of the literature turned up no studies that examine trends over time in health equity between heterosexual and sexual minority people, whether adults or adolescents. Indeed, in searching the literature, we found no articles that directly tested trends in health gaps between any groups, on any health issue.
Part of the reason for this lack of research may be a lack of appropriate analytical methods. Much of the time, health disparities are documented using logistic regressions in order to calculate odds ratios or risk ratios between the dominant majority and the marginalized subgroup. Repeating those analyses separately within multiple cohorts over time and displaying the odds ratios in a table, or graphing them in a figure, which has been used in the past to show persistent disparities , is insufficient to accurately evaluate trends in disparities. This is because odds ratios should not be directly compared, not even with confidence intervals, due to underlying heterogeneity that cannot be accounted for in logistic regression models . However, in the past decade, some studies have developed methods to compare odds ratios between independent samples for other purposes, using interaction terms. For example, Altman and Bland  recommended using interactions in logistic regression to help determine whether treatment effects differ between two subgroups in intervention studies. These interaction terms produce a ratio of odds ratios within the logistic model. Our question is, could these methods be adapted to test year by orientation group interactions instead, with heterosexual youth as the referent group for orientation, to determine whether the gap is narrowing, widening, or unchanged? If so, this would offer a new method for examining health equity trends among marginalized groups.
Thus, our primary purpose was to adapt these approaches, in order to test trends in health disparities over time between sexual minority and heterosexual youth cohorts in population-based studies. In this analysis, we used disparities in tobacco use among students in the state of Minnesota in the United States of America as the exemplar case to demonstrate the method.
Data were from the Minnesota Student Survey (MSS), a cross-sectional statewide anonymous census of adolescent health administered every 3 years to public school students in grades 9 and 12 in Minnesota. All school districts were invited to participate and school district participation rates were approximately 90 %. Approximately 75,000 students in grades 9 and 12 participated in each year of the survey. The detailed survey procedure has been described elsewhere . We used a weighted, merged data set from 1998 to 2010 for trend analyses, including only school districts that participated in all survey years from 1998 to 2010. The University of British Columbia Behavioural Research Ethics Board approved the study under which these specific analyses were conducted (certificate # H12-00477).
As a secondary analysis, we were limited in the measure of sexual orientation to a consistent measure across all the years: our sample only included students who provided responses to two questions: “During the last 12 months, with how many different male partners have you had sexual intercourse?” and “During the last 12 months, with how many different female partners have you had sexual intercourse?” and to gender. Thus our sample consisted of recently sexually active students, grouped by gender of sexual partner into three categories: those who had sex with opposite-gender partner(s) only (OPPOS), those who had sex with same-gender sexual partner(s) only (SAME), and those who had sex with both male and female partners (BOTH). Approximately 30 % of students from the original surveys in each year reported having had sexual intercourse with male and/or female partners in the last 12 months (Ns = 17,376–19,617).
Samples Students By Gender of Sexual Partners in the Minnesota Student Surveys
(Grade 12, %)
(Grade 12, %)
(Grade 12, %)
(Grade 12, %)
(Grade 12, %)
(Grade 12, %)
Smoking in the last 30 days was assessed with the item, “During the last 30 days, how frequently have you smoked cigarettes?” Having never smoked a cigarette in the last 30 days was coded as “No” and having smoked less than one cigarette per day or more frequently in the last 30 days as “Yes”.
There were significant gender differences in the prevalence of tobacco use, as well as the prevalence of same-gender or both-gender sexual behaviours (data not shown), so all analyses were stratified by gender.
In order to examine trends over time within each of three orientation groups, we first described the prevalence of cigarette smoking in each of 5 years, separately by sexual orientation. Chi-square test for trend was used to compare 1998, 2004, and 2010 data. Because grade 9 respondents may have completed a subsequent survey when they were in 12th grade, we selected only those 3 years, to ensure sample independence. Likewise, given that there may have been changes over time in the age at which adolescents initiate sexual behaviour or tobacco use, we also conducted logistic regressions, adjusted for grade, to assess changes in the prevalence of cigarette smoking from 1998 to 2004 and from 2004 to 2010 within each orientation group. An odds ratio (OR) greater than 1 indicates an increasing trend, and an OR less than 1 indicates a decreasing trend.
Next, to examine sexual orientation-based disparities in tobacco use within each of the five survey years, we conducted grade-adjusted logistic regressions. The OPPOS group was used as a reference category. Thus, an OR of > 1 indicates that SAME or BOTH students were more likely than OPPOS students to report having smoked a cigarette in the last 30 days. An OR of < 1 indicates a lower likelihood of recent smoking among SAME or BOTH than among OPPOS.
Interpretation of Odds Ratios for Analyses of Trends in Disparities
ORs for interaction terms
Sexual orientation-based disparities
Year 1998 or 2010
Year 2004 (reference)
Year 1998 or 2010
Year 2004 (reference)
A question may arise, why are these complex interaction terms needed for testing trends in disparities? Couldn’t we just compare the trends in the prevalence (i.e., the percents) between heterosexual and sexual minority youth, rather than comparing trends in the odds ratios or adjusted odds ratios? Indeed, Asada  notes the importance of examining absolute measures of inequalities that take into account the starting point, or the absolute level of the health issue in each group, when comparing populations. Asada asserts that by calculating the absolute change in each group by subtracting the last year percent from first year percent, the absolute size of the change can hint at the trend in the inequality, i.e., the inequality is greater when the dominant group improves more than the minority group, and it is smaller when the minority group improves more than the minority group. The problem with directly comparing trends in prevalence between heterosexual and sexual minority youth as an absolute difference in percentage points is that these results may be somewhat misleading, because simple percents do not account for age differences in the two samples that can provide a competing explanation for the disparities. Because sexual orientation is a developmental task of adolescence, and development itself has a wide range of normal variation in timing, not all young people go through puberty, develop attractions, or begin romantic or sexual relationships at the same age, and if the attractions or identity are stigmatized, it tends to be longer before young people publicly disclose that, even if they act on the knowledge earlier . Thus, in secondary school population surveys throughout the world, researchers have generally noted significant age differences between heterosexual and sexual minority youth, no matter how sexual orientation is measured [1, 24, 25], although those age differences are not always in the same direction. For example, youth engaging in same-gender or both-gender sexual behaviour tend to report that at slightly younger ages than those engaging in opposite-sex sexual behaviours , which appears to be due in part to their greater risk for being targeted for sexual abuse during early adolescence [2, 8].
In contrast, among those with same-sex attractions, stigma may delay their recognition of and public disclosure of such attractions . And when it comes to identity labels, because of our heteronormative society, many adolescents who have not yet developed attractions or engaged in sexual behaviour identify as the default heterosexual, and it is only later during adolescence that they identify as lesbian, gay, or bisexual . Thus, in population-based surveys, the average age of sexual minority youth tends to be different from the average age among heterosexual youth, but may be older or younger, depending on whether the sample is limited to sexually experienced youth, or all youth. And since smoking itself has a maturational trend (older youth are more likely to smoke than younger adolescents), unless the analyses controls for age (or grade in studies where the sample is only within discrete grades), the prevalences for heterosexual and sexual minority youth are not truly comparable, they're confounded by demographic differences in age. Therefore, age- or grade- adjusted regressions are potentially more appropriate comparisons to use in order to identify trends in disparities.
Documenting the strength of disparities between groups, and testing trends in disparities, however, are only part of the information needed to understand and address health equity issues. Odds ratios, even adjusted odds, are a measure of the relative strength of the disparity between two groups, but provide no information about the scope or magnitude of the problem. For example, an odds ratio of 2.0 could describe a disparity in tobacco use between two groups when the difference is 1.5 % vs. 3.0 %, or when it is 45 % vs. 90 %, but most people would find the second disparity far more urgent a concern. As well, the size of the percentage change within a group over time may provide a clue as to which group’s improvement or decline might be driving the widening or narrowing (or unchanged) gap between the groups. Therefore, it is important to report the prevalence to give readers a sense of the size of the issue, while using age-adjusted regressions to describe the strength of the disparity between both groups as an adjusted odds ratio, and to test the trends in the disparity using the ratio of adjusted odds ratios.
Results and discussion
Trends in past-month cigarette smoking within each orientation group
Trends in Prevalence of Recent Cigarette Smoking Across Years, within Sexual Orientation Groups
Ever smoked cigarettes in the last 30 days (%)
Trend 1998 – 2004b
Trend 2004 – 2010c
ORd (95% CI)
ORd (95% CI)
0.48 (0.45, 0.51)
0.66 (0.62, 0.71)
0.59 (0.49, 0.70)
0.59 (0.50, 0.70)
0.48 (0.30, 0.76)
0.83 (0.54, 1.25)
0.46 (0.43, 0.48)
0.52 (0.49, 0.56)
0.63 (0.46, 0.85)
0.51 (0.40, 0.64)
0.74 (0.35, 1.59)
0.53 (0.30, 0.96)
Disparities in smoking between sexual minority and heterosexual groups in each year
Sexual Orientation Disparities in Recent Cigarette Smoking, within Year: Odds Ratiosa and 95 % Confidence Intervals
1.36 (1.19, 1.55)
1.43 (1.25, 1.64)
1.73 (1.51, 1.99)
1.33 (1.19, 1.48)
1.60 (1.43, 1.80)
0.86 (0.62, 1.19)
0.63 (0.44, 0.90)
0.87 (0.63, 1.20)
1.12 (0.86, 1.47)
1.10 (0.84, 1.44)
1.66 (1.30, 2.13)
1.76 (1.45, 2.14)
2.36 (1.96, 2.84)
2.27 (1.92, 2.68)
2.33 (2.00, 2.71)
0.67 (0.36, 1.24)
0.56 (0.33, 0.97)
1.08 (0.69, 1.70)
0.79 (0.52, 1.20)
1.24 (0.87, 1.78)
Trends in sexual orientation disparities in smoking: Is it getting better?
Trends in Disparities in Recent Cigarette Smoking: Interactions Between Sexual Orientation and Year
ORa (95% CI)
ORa (95% CI)
OPPOS by Year 2004
BOTH by Year 1998
0.81 (0.67, 0.98)
0.73 (0.54, 0.99)
BOTH by Year 2010
0.91 (0.76, 1.08)
0.98 (0.77, 1.24)
SAME by Year 1998
1.00 (0.63, 1.58)
0.65 (0.30, 1.38)
SAME by Year 2010
1.25 (0.82, 1.91)
1.15 (0.65, 2.04)
In this paper, we have documented trends in recent cigarette use among sexually active heterosexual and sexual minority adolescents in Minnesota schools between 1998 and 2010. As demonstrated in recent studies of tobacco use among general populations of adolescents in high-income countries, we found steady declines in tobacco use within each orientation group across the years. In each year of the survey, we found disparities in past-month cigarette smoking between youth with bisexual vs. heterosexual sexual behaviours, which are similar to findings from other cross-sectional studies. In contrast to other studies, however, we did not find disparities between youth who report only same-gender sexual behaviour and those with heterosexual behaviour. This may be partly explained by differences in how sexual orientation is measured in the Minnesota Student Survey as compared to other studies, as some studies combine lesbian, gay and bisexual adolescents, thus potentially masking differences within sexual minority groups. Likewise, studies that use attraction or identity as the sexual orientation measure, although generally considered a more valid measure than behaviour alone [1, 26], may end up combining youth with solely same-gender sexual partners and those with both-gender sexual partners, because attraction, identity and actual sexual behaviour are not necessarily concordant among adolescents [9, 24, 25]. However, other studies have documented higher levels of health challenges among bisexual youth and adults, in part because of lack of acceptance in gay and lesbian communities as well as heterosexual communities, and these disparities in tobacco use for youth with bisexual behaviour are of concern.
We also demonstrated a novel technique for testing whether these disparities between sexual minority and heterosexual youth tobacco use were significantly increasing, declining, or unchanged, and found that the gap between bisexual and heterosexual youth’s tobacco use widened between 1998 and 2004 for both boys and girls reporting both-gender sexual partners, and these disparities continued between 2004 and 2010. This analysis suggests that while population-wide interventions to reduce tobacco initiation among adolescents is having some effect on sexual minority youth, among those with bisexual behaviour, the interventions are not as effective as on those with exclusively monosexual behaviour, whether opposite-sex only or same-sex only. The reasons for such widening disparities over the past decade are not clear, but one possible explanation is that bisexual invisibility, or biphobia in both mainstream community and LGBTQ communities increase the minority stress for bisexual adolescents, and reduce their social supports. Thus, additional targeted interventions may be needed, either to reduce biphobia and stigma-related stress or improve social supports, in order to further reduce the prevalence of tobacco use among bisexual adolescents.
Strengths and limitations
As with all studies, this analysis has both strengths and limitations. A key strength is the use of regularly repeated large-scale population surveys of adolescents in school, with one of the few such surveys that covers more than a decade and includes a measure of sexual orientation, with large enough sample sizes to disaggregate those with same-gender only and both-gender sexual partners. As well, the use of an established statistical technique, although applying it for a novel analytical purpose, is a clear innovation that may be useful for measuring trends in health equity for sexual minority youth and potentially other marginalized groups.
At the same time, it must be acknowledged that the measure of sexual orientation itself is sub-optimal. Since the majority of adolescents in North America are not sexually active, a measure of sexual orientation that relies on sexual behaviours will inevitably exclude a significant proportion of the population, either heterosexual or sexual minority youth. Sexual orientation has been more accurately assessed among adolescents as attractions or identity labels  as there is noted discordance between attractions/identity and actual sexual behaviour [9, 24, 25]. As well, focusing on the gender of past year sexual partners may misclassify youth with only one sexual partner in the past year, whether same-gender or other-gender; if any previous partners in earlier years were of a different gender, they would more accurately belong in the both-gender category. These trend analyses should be replicated with other population-based surveys that also assess sexual orientation through attraction or identity measures, so that youth who are not sexually experienced are included in the study. These results are from a single state in the US, Minnesota; different environments, laws, culture and history in other regions of the US might influence trends in health disparities for sexual minority youth in different directions than we found in Minnesota. Similarly, trends in disparities may look different in Canada, or in other countries outside of North America; these methods should be used where possible to document whether disparities are getting better in different regions globally, wherever it is feasible given existing population surveys. Finally, as a study of youth in school, the results cannot be generalized to youth who are not attending school.
This is among the first few studies to document trends in health and risk behaviours among sexual minority adolescents, and to our knowledge, is the first to actually test trends in health disparities for them. The results show the importance of not only documenting trends among sexual minority populations, but also documenting whether health equity is improving for them compared to dominant groups. This information can help guide policies, practices, and resource allocation to reduce health inequities.
This innovative method of testing trends in health disparities, however, to see if the gap is narrowing, widening, or unchanged, has relevance for health equity beyond that of sexual minority people. The technique could be applied within national, regional, and local population studies to evaluate trends in health disparities among other groups who experience significantly higher health risks or poorer health outcomes as a result of social marginalization; for example, disparities between ethnocultural minority groups and the dominant groups in societies, or between low-income groups and those with higher socio-economic status. It may not be enough to track changes in health outcomes for Indigenous populations, or for groups of children in foster care, for example, without placing these trends in the context of the wider community’s trends in health improvements. This new approach to testing the trend in disparities is one way to take context into account, and will provide important information to guide health equity initiatives for marginalized groups.
BOTH, adolescents who had sex with both male and female partners; LGB, lesbian, gay, and bisexual; OPPOS, adolescents who had sex with opposite-gender partner(s) only; SAME, adolescents who had sex with same-gender partner(s) only.
This study was funded by grants #CPP 86374 and #MOP 119472 from the Canadian Institutes of Health Research (Saewyc, PI). The authors acknowledge the Minnesota Departments of Health, Human Services, and Education for permission to access the Minnesota Student Surveys.
YH, ES, and BDZ conceptualized and designed the study. YH analyzed the data, interpreted the results, and drafted much of the manuscript. ES secured access to the Minnesota Student Survey data, contributed to interpreting the results, wrote portions of the manuscript, and revised the final manuscript. BDZ guided data analysis, interpreted the results, and edited the final manuscript. All authors read and approved the final manuscript.
The authors declare that they have no competing interests.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
- Saewyc EM. Research on adolescent sexual orientation: Development, health disparities, stigma and resilience. J Res Adolesc. 2011;21(1):256–72.View ArticlePubMedPubMed CentralGoogle Scholar
- Friedman MS, Marshal MP, Guadamuz TE, Wei C, Saewyc E, Wong CF, Stall R. A meta-analysis to examine disparities in childhood sexual abuse, parental physical abuse, and peer victimization among sexual minority and non-sexual minority individuals. Am J Public Health. 2011;101(81):1481–94. doi:10.2105/AJPH.2009.190009.View ArticlePubMedPubMed CentralGoogle Scholar
- Marshal MP, Friedman MS, Stall R, King KM, Miles J, Gold MA, Bukstein OG, Morse JQ. Sexual orientation and adolescent substance use: A meta-analysis and methodological review. Addiction. 2008;103:546–56.View ArticlePubMedPubMed CentralGoogle Scholar
- Russell ST, Driscoll AK, Truong N. Adolescent same-sex romantic attractions and relationships: Implications for substance use and abuse. Am J Public Health. 2002;92:198–202.View ArticlePubMedPubMed CentralGoogle Scholar
- Konishi C, Saewyc E, Homma Y, Poon C. Population-level evaluation of school-based interventions to prevent problem substance use among gay, lesbian and bisexual adolescents in Canada. Prev Med. 2013;57(6):929–33. doi:10.1016/j.ypmed.2013.06.031.View ArticlePubMedPubMed CentralGoogle Scholar
- Saewyc EM, Poon C, Homma Y, Skay CL. Stigma management? The links between enacted stigma and teen pregnancy trends among gay, lesbian and bisexual students in British Columbia. Can J Hum Sex. 2008;17(3):123–31. PMC2655734.PubMedPubMed CentralGoogle Scholar
- Rossen FV, Lucassen MFG, Denny S, Robinson E. Youth ‘07 The health and wellbeing of secondary school students in New Zealand: Results for young people attracted to the same sex or both sexes. Auckland: The University of Auckland; 2009.Google Scholar
- Saewyc E, Skay C, Richens K, Reis E, Poon C, Murphy A. Sexual orientation, sexual abuse, and HIV-risk behaviors among adolescents in the Pacific Northwest. Am J Public Health. 2006;96(6):1104–10. doi:10.2105/AJPH.2005.065870.View ArticlePubMedPubMed CentralGoogle Scholar
- Kann L, Olsen EO, McManus T, et al. Sexual identity, sex of sexual contacts, and health-risk behaviors among students in grades 9-12--Youth Risk Behavior Surveillance, selected sites, United States, 2001–2009. MMWR Surveill Summ. 2011;60(7):1–133.PubMedGoogle Scholar
- Marshal MP, Dietz LJ, Friedman MS, Stall R, Smith HA, McGinley J, Thoma BC, Murray PJ, D’Augelli AR, et al. Suicidality and depression disparities between sexual minority and heterosexual youth: A meta-analytic review. J Adolesc Health. 2011;49:115–23.View ArticlePubMedPubMed CentralGoogle Scholar
- Austin SB, Ziyadeh N, Fisher LB, Kahn JA, Colditz GA, Frazier AL. Sexual orientation and tobacco use in a cohort study of US adolescent girls and boys. Arch Pediatr Adolesc Med. 2004;158:317–22.View ArticlePubMedGoogle Scholar
- Hagger Johnson G, Taibjee R, Semlyen J, Fitchie I, Fish J, Meads C, Varney J. Sexual orientation identity in relation to smoking history and alcohol use at age 18/19: cross-sectional associations from the Longitudinal Study of Young People in England (LSYPE). MJ Open. 2013;3:e002810. doi:10.1136/bmjopen-2013-002810.Google Scholar
- Busseri MA, Willoughby T, Chalmers H, Bogaert AF. On the association between sexual attraction and adolescent risk behavior involvement: Examining mediation and moderation. Dev Psych. 2008;44(1):69–80.View ArticleGoogle Scholar
- Saewyc E, Poon C, Wang N, Homma Y, Smith A, the McCreary Centre Society. Not Yet Equal: The Health of Lesbian, Gay, & Bisexual Youth in BC. Vancouver, BC: McCreary Centre Society; 2007. ISBN #: 978-1-895438-84-5.Google Scholar
- Hurley M. Sexual orientation and legal rights: A chronological overview. PRB 04-13E. Ottawa: Library of Parliament. Revised September 2005.Google Scholar
- Supreme Court of the United States of America, Obergefell v. Hodges, 576 U.S. *****, No. 14–556. 2015.Google Scholar
- Russell ST, Kosciw J, Horn S, Saewyc E. Safe schools policy for LGBTQ students. SRCD Social Policy Report. 2010;24(4):1–17.Google Scholar
- Lucassen MFG, Clark TC, Denny SJ, Fleming TM, Rossen FV, Sheridan J, Bullen P, Robinson EM. What has changed from 2001 to 2012 for sexual minority youth in New Zealand? J Paed Child Health. epub ahead of print 2014, doi:10.1111/jpc.12727.
- Dick B, Ferguson J, Saewyc E, Baltag V, Bosek K, for the World Health Organization. Health for the World’s Adolescents: A Second Chance in the Second Decade of Life. Geneva: World Health Organization; 2014. http://www.who.int/adolescent/second-decade.Google Scholar
- Allison PD. Comparing logit and probit coefficients across groups. Sociol Methods Res. 1999;28:186–208.View ArticleGoogle Scholar
- Altman DG, Bland JM. Interaction revisited: The difference between two estimates. BMJ. 2003, 326:219. doi: http://0-dx.doi.org.brum.beds.ac.uk/10.1136/bmj.326.7382.219.
- Minnesota Departments of Education, Health, Human Services, & Public Safety (2010). Minnesota Student Survey: 1992 – 2010 trends. Retrieved from http://www.health.state.mn.us/divs/chs/mss/trendreports/msstrendteport2010.pdf.
- Asada Y. On the choice of absolute or relative inequality measures. Milbank Q. 2010;88(4):612–22.View ArticleGoogle Scholar
- Goodenow C, Szalacha LA, Robin LE, Westheimer K. Dimensions of sexual orientation and HIV-related risk among adolescent females: Evidence from a statewide survey. Am J Public Health. 2008;98(6):1051–8. doi:10.2105/AJPH.2005.080531.View ArticlePubMedPubMed CentralGoogle Scholar
- Igartua K, Thombs BD, Burgos G, Montoro R. Concordance and discrepancy in sexual identity, attraction, and behavior among adolescents. J Adol Health. 2009;45:602–8. doi:10.1016/j.jadohealth.2009.03.019.View ArticleGoogle Scholar
- Saewyc EM, Bauer GR, Skay CL, Bearinger LH, Resnick MD, Reis E, Murphy A. Measuring sexual orientation in adolescent health surveys: Evaluation of eight school-based surveys. J Adol Health. 2004;35:345e. 1-e.16, on-line at http://download.journals.elsevierhealth.com/pdfs/journals/1054-139X/PIIS1054139X04001612.pdf.View ArticleGoogle Scholar