Will All Youth Answer Sexual Orientation and Gender-Related Survey Questions? An Analysis of Missingness in a Large US Survey of Adolescents and Young Adults

Abstract

Some researchers and clinicians may feel hesitant to assess sexual orientation and gender-related characteristics in youth surveys because they are unsure youth will respond to these questions or are concerned the questions are too sensitive and will cause discomfort or offense. This can result in missed opportunities to identify LGBTQ+ youth and address health inequities among this population. The aim of this study was to examine prevalence and sociodemographic patterns of missingness among survey questions assessing current sexual orientation, gender identity and expression (SOGIE) and past change in sexual orientation (sexual fluidity) among a diverse sample of US youth. Participants (N=4,245, ages 14-25 years; 95% cisgender, 70% straight/heterosexual, 53% youth of color), recruited from an online survey panel, completed the Wave 1 survey of the longitudinal Sexual Orientation Fluidity in Youth (SO*FLY) Study in 2021. Current SOGIE, past sexual fluidity, and sociodemographic characteristics were assessed for missingness. Overall, 95.7% of participants had no missing questions, 3.8% were missing one question, and 0.5% were missing ≥2 questions. Past sexual fluidity and assigned sex were most commonly missing. Sociodemographic differences between participants who skipped the SOGIE questions and the rest of the sample were minimal. Missingness for the examined items was low and similar across sociodemographic characteristics, suggesting that almost all youth are willing to respond to survey questions about SOGIE. SOGIE and sexual fluidity items should be included in surveys and clinical assessments of youth to inform clinical care, policy-making, interventions, and resource development to improve the health of all youth.

The population of LGBTQ+ (lesbian, gay, bisexual, transgender, queer, and all sexual and gender minority identities) youth in the US is significant, representing an estimated 9.5% of youth, ages 13-17 years (Conron, 2020). The proportion of LGBTQ+ youth in the US has increased over time (Jones, 2022); for example, 14.3% of youth who responded to the Youth Risk Behavioral Surveillance Survey in 2017 identified as a sexual minority vs. 7.3% of youth in 2009 (Raifman et al., 2020). LGBTQ+ youth are more likely than their heterosexual and/or cisgender peers to report adverse mental, behavioral, and physical health outcomes (Hafeez et al., 2017). Given the health inequities and discrimination LGBTQ+ youth face, it is necessary to understand the breadth of identities and identity trajectories in youth. Improved identity measurement will allow researchers, practitioners, and advocates to accurately identify and address specific needs to improve the health of LGTBQ+ and all youth (Spock et al., 2022).

Health surveys that assess sexual orientation, gender identity, and gender expression (SOGIE) are necessary to inform legislation, government funding, interventions, and resource development (Streed et al., 2020). Despite its importance, research on LGBTQ+ youth health has been limited by a lack of standardized and consistently collected measures of SOGIE constructs (National Academies of Sciences, 2022). Additionally, when sexual orientation is assessed, it is often unidimensional, not considering the three primary dimensions of sexual orientation: identity, attractions, and behavior (Ruberg & Ruelos, 2020), which may not be congruent (Ybarra et al., 2019), and are uniquely linked to health (Dawson et al., 2022; Matthews et al., 2014).

Current research supports sexual orientation assessment in adult surveys (Brabete et al., 2020; Case et al., 2008; Jans et al., 2015; VanKim et al., 2010). An analysis of New Mexico health survey data found low nonresponse rates for sexual orientation questions, suggesting adults are willing to answer these questions, and recommending that future health surveys should assess sexual orientation (VanKim et al., 2010). Another study examining response rates for sexual orientation questions in the Nurses’ Health Study II found these questions did not prompt participants to leave the study (Case et al., 2008). Similarly, an analysis of demographic patterns of missing sexual orientation data in adults in the Canadian Community Health Survey found a low nonresponse rate for sexual orientation questions (Brabete et al., 2020). These findings suggest collecting this information among adults is feasible.

Despite existing research examining sexual orientation nonresponse in adult surveys (Brabete et al., 2020; Case et al., 2008; Jans et al., 2015; VanKim et al., 2010), little research has examined missingness among gender identity questions among either adults or youth. One exception, a recent study examining missingness of both sexual orientation and gender identity questions among adults in the Behavioral Risk Factor Surveillance System, found gender identity was less likely than sexual orientation identity to be missing and that predictors of missingness varied considerably (Jesdale, 2021). Data collection methods, particularly around demographics, often fail to capture complexities of LGBTQ+ identities and experiences (Ruberg & Ruelos, 2020). For example, sexual orientation questions may be imbued with binary ideas of sexuality and gender (Brabete et al., 2020), which may create difficulty for individuals who have past changes in sexual orientation identity and attractions (i.e., sexual fluidity) and/or are gender fluid. Surveys that only allow participants to select one identity do not capture the multiple terms that some participants use to describe their identities, leaving researchers with an incomplete or inaccurate understanding of LGBTQ+ experiences.

Gaps in literature around missingness of SOGIE and sexual fluidity survey questions are even more pronounced for youth. The US National Academies of Science, Engineering, and Medicine made recommendations about SOGIE measures for adults; however, the report stopped short of reviewing measures for youth, noting a need for further work in this area (National Academies of Sciences, 2022). One study assessing question nonresponse in eight school-based surveys administered from 1986-1999 found that nonresponse for sexual orientation was similar to other sexual-focused questions (Saewyc et al., 2004). However, this research has several limitations. The reliance on school-based samples misses youth who are not in school, either because they matriculated or left for another reason. Furthermore, these data were collected in the 1980’s and 1990’s; due to increasing public acceptance and awareness of LGBTQ+ identities, these findings are likely not representative of the likelihood of today’s youth answering such questions. The majority of prior research has also failed to examine missingness among gender identity and expression survey questions in youth, and missingness in sexual fluidity questions in any age group. Collecting SOGIE and sexual fluidity data is necessary to create systems to improve the health of all youth, particularly LGTBQ+ youth.

The aim of this study was to examine prevalence and sociodemographic patterns of missingness among SOGIE and sexual fluidity survey items administered to youth. Based on previous adult studies (Brabete et al., 2020; Case et al., 2008; Jans et al., 2015; Jesdale, 2021; VanKim et al., 2010), we hypothesized the prevalence of SOGIE missingness in this youth sample would be low. Due to lack of prior evidence to inform specific hypotheses, no hypotheses were proposed about sociodemographic patterns of missingness.

Method

Participants

Participants were 4,245 US youth, ages 14-25 years, who completed Wave 1 of the Sexual Orientation Fluidity in Youth (SO*FLY) Study (Katz-Wise et al., 2022). Eligibility criteria for SO*FLY included being age 14-25 years and living in the US. Eligible individuals were recruited from a participant panel from Prodege, an online survey firm, which recruits from TV, radio, and online advertisements. Potential participants were identified based on sociodemographic information previously provided as part of their participation in the Prodege panel, including sexual orientation, gender identity, and race/ethnicity. The full panel was 29% age 13-25 years, 65% female, 71% white, 15% Latina/o/x and/or Hispanic, 9% Asian, 10% Black or African American, and 10% another race/ethnicity. For the SO*FLY Study, LGBTQ+ youth and youth of color were oversampled and participants were sampled from across the US to enable subgroup analyses. Sociodemographic characteristics are in Table 1.

Table 1.

Sample Sociodemographic Characteristics (N=4,245)

SOGIE Characteristics	n	%	Missing n (%)
Sexual orientation identity			17 (0.4)
Straight/heterosexual	2,972	70.3
Bisexual	624	14.8
Gay or lesbian	181	4.3
Pansexual	188	4.5
Queer	43	1.0
Asexual	76	1.8
Not sure	112	2.7
Another identity/identities	32	0.8
Attractions			11 (0.2)
Girls/women only	1,509	35.6
Boys/men only	1,750	41.3
Nonbinary people only	32	0.8
Another gender identity only	41	1.0
Multiple gender identities	902	21.3
Sex assigned at birth			28 (0.7)
Female	2,630	62.4
Male	1,587	37.6
Gender identity			10 (0.2)
Girl/woman	2,457	58.0
Boy/man	1,602	36.6
Nonbinary	140	3.3
Another gender identity	3	0.1
Selected multiple gender identities	33	0.8
Transgender status			17 (0.4)
Do not identify as transgender	4,003	94.7
Identify as transgender	116	2.7
Not sure	91	2.2
Don’t know what question is asking	18	0.4
Gender expression			16 (0.4)
Very feminine	663	15.7
Mostly feminine	994	23.5
Somewhat feminine	522	12.3
Equally feminine and masculine	543	12.8
Somewhat masculine	358	8.5
Mostly masculine	693	16.4
Very masculine	456	10.8
Past change in sexual orientation identity
Yes	698	16.6	51 (1.2)
No	3,496	83.4
Past change in attractions			34 (0.8)
Yes	1,386	32.9
No	2,825	67.1
Sociodemographic Characteristics
Age group			0 (0)
14-17 years	440	10.4
18-21 years	1,734	40.9
22-25 years	2,071	48.8
Gender identity (constructed)a			38 (0.9)
Cisgender girl/woman	2,438	58.0
Cisgender boy/man	1,540	36.6
Transgender girl/woman	6	0.1
Transgender boy/man	48	1.1
Nonbinary/additional identity or identities	175	4.2
Race/ethnicity			21 (0.5)
American Indian or Alaska Native	51	1.2
Asian	378	9.0
Black or African American	701	16.6
Latina/o/x and/or Hispanic	206	4.9
Native Hawaiian or Other Pacific Islander	12	0.3
White	2,003	47.4
Another race	23	0.5
Selected multiple races/ethnicities	849	20.1
Urban-rural categoryb			5 (0.1)
Large central metro	1,333	31.4
Large fringe metro	1,051	24.8
Medium metro	885	20.9
Small metro	419	9.9
Micropolitan	338	8.0
Non-core	214	5.1

^aVariable constructed from sex assigned at birth and gender identity.

^bCounty determined by reported zip code, based on the National Center for Health Statistics urban-rural county categories, 2013.

Measures

Sociodemographic Characteristics

The following sociodemographic characteristics were used to describe the sample and examine missing data patterns: age in years, current gender identity, current sexual orientation identity, race/ethnicity, and rurality. Race and ethnicity were assessed with two separate items: “Are you Latina/o/x and/or Hispanic?” (Response options: yes, no) and “Which race(s) best describe you?” Participants could select all that applied from: American Indian or Alaska Native, Asian, Black or African American, White, Native Hawaiian or Other Pacific Islander, another race: (write-in). These items were combined into one race/ethnicity variable for analyses. Rurality was assessed by coding participants’ zip codes into categories (large central metro, large fringe metro, medium metro, small metro, micropolitan, non-core) (Ingram & Franco, 2012).

Current SOGIE and Past Sexual Fluidity

Missingness was examined for current sexual orientation identity, current attractions, current gender identity, sex assigned at birth, transgender status, current gender expression, and past change in sexual orientation identity and attractions.

Sexual orientation identity was assessed with one item adapted from prior studies (Centers for Disease Control and Prevention, 2019; Human Rights Campaign, 2018; Meyer, 2020): “Sexual orientation’ describes who you are attracted to and how you identify yourself based on those attractions. Sexual orientation may change over the course of people’s lives. Which of the following best describes your current sexual orientation?” Response options: straight/heterosexual, bisexual, gay or lesbian, pansexual, queer, asexual, not sure, another identity/identities (write in). Participants could select one response option.

Attractions were assessed with one item adapted from prior studies (Centers for Disease Control and Prevention, 2019; Human Rights Campaign, 2018; Meyer, 2020): “‘Attraction’ describes sexual or romantic feelings toward another person. Attractions may change over the course of people’s lives. Who are you currently attracted to?”. Response options: girls/women, boys/men, nonbinary people (e.g., genderqueer, gender non-conforming, another nonbinary identity), people of another gender identity: (write-in). Participants could select all that applied. The following mutually exclusive categories were created for analyses: girls/women only, boys/men only, nonbinary people only, people of another gender identity only, and multiple gender identities (attraction to ≥2 identities).

Gender identity was assessed with one item adapted from prior research (Bauer et al., 2017): “Which of the following best describes your current gender identity?” Response options: girl/woman; boy/man; nonbinary, e.g., genderqueer, gender non-conforming, another nonbinary identity; another gender identity (write-in). Participants could select all that applied. Sex assigned at birth was assessed with one recommended item (Williams Institute Scholars, 2020): “What sex were you assigned at birth, on your original birth certificate?” Response options: female, male. Gender identity categories were created from these items as follows: cisgender girl/woman (sex assigned at birth: female, gender identity: girl/woman), cisgender boy/man (sex assigned at birth: male, gender identity: boy/man), transgender girl/women (sex assigned at birth: male, gender identity: girl/woman), transgender boy/man (sex assigned at birth: female, gender identity: boy/man), nonbinary (gender identity: nonbinary with or without another identity or a nonbinary identity written in with or without girl/woman or boy/man).

Transgender status was assessed with one item adapted from prior research (Johns et al., 2022): “Do you identify as transgender?” Response options: no, I do not identify as transgender; yes, I do identify as transgender; I am not sure if I am transgender; I do not know what this question is asking. Participants could select one option.

Gender expression was assessed with one item adapted from prior research (Wylie et al., 2010): “A person’s appearance, style, dress, or the way they walk or talk may affect how people describe them. On average, how do you think other people would describe you?” Response options: Very feminine, mostly feminine, somewhat feminine, equally feminine and masculine, somewhat masculine, mostly masculine, very masculine. Participants could select one option.

Past sexual fluidity was assessed with two items adapted from prior research (Katz-Wise, 2015): 1) “Have you ever experienced a change in your sexual orientation identity? For example, identifying one way, then identifying another way.” 2) “Have you ever experienced a change in your attractions to others over time? For example, feeling attracted to only one gender, then feeling attracted to a different gender or more than one gender.” Response options for both items: yes, no.

Procedure

Participants completed Wave 1 surveys online via the Prodege survey platform in August 2021. Eligible individuals from the Prodege panel were presented with the survey length and incentive amount on a Prodege dashboard accessed either via an app or through a website. After clicking on the survey, participants saw the following introductory text: “Welcome! The following questions will ask about your sexual orientation and attraction(s). If you are not sure what a term means, please answer to the best of your ability. All responses will remain confidential.” Then participants provided informed consent and continued with the survey. Surveys assessed sociodemographic characteristics, including current SOGIE and past sexual fluidity (see Measures). Upon survey completion, all participants received mental health and support resources, and an incentive within Prodege’s structure of panel currency. Study procedures were approved by the Boston Children’s Hospital Institutional Review Board.

Statistical Analysis

We first examined sociodemographic characteristics of the overall sample (N=4,245), reporting the frequency and proportions for dichotomous and categorical variables. We then reported the proportion of data missing across individual variables. Finally, we qualitatively compared the proportion of missingness (for SOGIE variables with missing n ≥20) for specific sociodemographic subgroups to the full sample. Statistical tests of difference were not conducted due to small subgroup sample sizes. Analyses were conducted using SAS 9.4. This study was not preregistered. The data and study materials can be made available upon request.

Results

Sociodemographic Characteristics

Descriptive characteristics of the analytic sample are in Table 1. Participants were age 14-25 years; 48.8% were age 22-25 years. Participants were 47.4% white, 16.6% Black, 9.0% Asian, 4.9% Latina/o/x, and 1.2% American Indian or Alaska Native; 20.1% identified with multiple racial/ethnic groups. Over half lived in a large central (31.4%) or large fringe metro (24.8%) county; micropolitan and non-core (more rural counties) represented 13.1% of the sample.

Sexual orientation identities included: 70.3% straight/heterosexual, 14.8% bisexual, 4.5% pansexual, and 4.3% gay or lesbian. Most participants reported attraction to one gender (41.3% boys/men only, 35.6% girls/women only, 0.8% nonbinary only, and 1.0% to another gender); 21.3% reported attraction to multiple genders. Most participants identified as a cisgender girl/woman or cisgender boy/man with a little over 5% identifying as transgender (girl/woman, boy/man), nonbinary, or another identity/identities.

Prevalence of Missingness

Overall, 95.7% (4,063/4,245) of participants had complete data (no missing variables); 3.8% (160/4,245) were missing only one question. Only 22 (0.5%) were missing ≥2 questions (range of missing questions 0-4). The most commonly missing questions were past change in sexual orientation identity (n=51 missing), past change in attractions (n=34 missing), and sex assigned at birth (n=28 missing, Table 1). Missingness increased for questions toward the end of the survey (Table 2).

Table 2.

Proportion Missing Response by Survey Question Order (N=4,245)

Variable	Survey Order	Missing Count	% Missing
Age	1	0	0.0
Race/ethnicitya	2	21	0.5
Urban-rural categoryb	3	5	0.1
Gender identity	4	10	0.2
Sex assigned at birth	5	28	0.7
Transgender status	6	17	0.4
Gender expression	7	16	0.4
Attractions	8	11	0.3
Current SOI	9	17	0.4
Past change in attractions	10	34	0.8
Past change in SOI	11	51	1.2

^aVariable constructed from two consecutive questions: Are you Latina/o/x and/or Hispanic? (n=13 missing) and What race(s) best describe you? (n=23 missing).

^bCounty was determined based on reported zip code, using the National Center for Health Statistics urban-rural county categories, 2013.

Sociodemographic Patterns of Missingness

We examined sociodemographic patterns of missing responses to the most commonly missing questions (Table 3). Qualitatively, there was limited variation in the sociodemographic characteristics of those missing the sexual orientation identity change question (n=51) compared to the overall sample, with small overrepresentation of: age 18-21 years, cisgender girls/women, participants identifying as Latina/o/x and/or Hispanic, and participants living in large central metro or small metro areas. Limited sociodemographic variability was also found among those missing the attraction change question compared to the overall sample, with small overrepresentation of age 18-21 years; cisgender boys/men; straight/heterosexual participants; American Indian or Alaska Native, Black or African American, or Latina/o/x and/or Hispanic participants; and participants living in large central metro or micropolitan/non-core areas. Those missing the sex assigned at birth question were similar to the overall sample with a slight overrepresentation of individuals identifying as pansexual. However, sociodemographic variations were small across these three questions, suggesting minimal differences between those who skipped these questions and the full sample.

Table 3.

Sociodemographic Differences in Missingness in Sexual Orientation and Past Sexual Fluidity (N=4,245)

	Overall Sample (N=4,245)		Missing Sex Assigned at Birth (N=28)		Missing Attraction Change (N=34)		Missing Sexual Orientation Identity Change (N=51)
	n	%	n	%	n	%	n	%
Age group
14-17 years	440	10.4	2	7.1	4	11.8	7	13.7
18-21 years	1,734	40.9	12	42.9	18	52.9	28	54.9
22-25 years	2,071	48.8	14	50.0	12	35.3	16	31.4
Missing	0	0.0	0	0.0	0	0.0	0	0.0
Gender identity (constructed)b
Cisgender girl/woman	2,438	57.4	-		20	58.8	31	60.8
Cisgender boy/man	1,540	36.3	-		14	41.2	19	37.3
Transgender girl/woman	6	0.1	-		0	0.0	0	0.0
Transgender boy/man	48	1.1	-		0	0.0	0	0.0
Nonbinary/additional identity or identities	175	4.1	-		0	0.0	1	2.0
Missing	38	0.2	28	100.0	0	0.0	0	0.0
Sexual orientation identity
Straight/heterosexual	2,972	70.0	20	71.4	25	73.5	34	66.7
Bisexual	624	14.7	3	10.7	3	8.8	7	13.7
Gay or lesbian	181	4.3	1	3.6	2	5.9	3	5.9
Pansexual	188	4.4	3	10.7	0	0.0	0	0.0
Queer	43	1.0	0	0	0	0.0	0	0.0
Asexual	76	1.8	0	0	0	0.0	0	0.0
Not sure	112	2.6	1	3.6	0	0.0	2	3.9
Another identity/identities	32	0.8	0	0	0	0.0	0	0.0
Missing	17	0.4	0	0	4	11.8	5	9.8
Race/ethnicity
American Indian or Alaska Native	51	1.2	0	0	3	8.8	1	2.0
Asian	378	8.9	1	3.6	2	5.9	2	3.9
Black or African American	701	16.5	5	17.9	8	23.5	7	13.7
Latina/o/x and/or Hispanic	206	4.9	1	3.6	7	20.6	7	13.7
Native Hawaiian or Other Pacific Islander	12	0.3	0	0	0	0.0	1	2.0
White	2,003	47.2	14	50.0	12	35.3	17	33.3
Another race	23	0.5	0	0	0	0.0	1	2.0
Selected multiple races/ethnicities	849	20.0	6	21.4	2	5.9	12	23.5
Missing	21	0.5	1	3.6	0	0.0	3	5.9
Urban-rural categoryb
Large central metro	1,333	31.4	9	32.1	13	38.2	19	37.3
Large fringe metro	1,051	24.8	7	25.0	4	11.8	8	15.6
Medium metro	885	20.9	7	25.0	5	14.7	10	19.6
Small metro	419	9.9	3	10.7	5	14.7	7	13.7
Micropolitan	338	8.0	2	7.1	5	14.7	5	9.8
Non-core	214	5.0	0	0	2	5.9	2	3.9
Missing	5	0.1	0	0	0	0.0	0	0.0

^aVariable constructed from sex assigned at birth and gender identity.

^bCounty was determined based on zip code, using the National Center for Health Statistics urban-rural county categories, 2013.

Discussion

The aim of this study was to examine prevalence and sociodemographic patterns of missingness among SOGIE and sexual fluidity survey questions in a large sample of US youth. This study represents, to our knowledge, the first exploration of missingness patterns across a broad range of SOGIE-related survey items and the first to explore these missingness patterns among adolescents. Further, it fills a gap in the literature regarding missingness in survey questions assessing gender expression and provides the first exploration of missingness in sexual fluidity survey questions in any age group.

Consistent with previous adult literature (Brabete et al., 2020; Case et al., 2008; Jans et al., 2015; Jesdale, 2021; VanKim et al., 2010), our study found low rates of missingness (<5%) in SOGIE and sexual fluidity questions. These findings directly challenge the idea that SOGIE-related questions are too sensitive to include in youth surveys. Considering that most of our sample identified as heterosexual (70.3%) and cisgender (94.6%), our findings also challenge assumptions that these youth will not answer SOGIE-related questions due to discomfort or offense. This is consistent with prior research on patient and clinician perspectives on collecting SOGIE data (Maragh-Bass et al., 2017).

In contrast with a prior adult study (Brabete et al., 2020), we found only small sociodemographic differences in missingness patterns. This may be due in part to small subgroup sample sizes, likely making proportions sensitive to small sociodemographic differences across questions; results should be interpreted with this limitation in mind. We also did not assess income or education levels, which have previously been cited alongside increased age as potential factors in data missingness patterns (Brabete et al., 2020). Further research is needed to understand the influence of these sociodemographic characteristics on SOGIE response probabilities in youth.

To our knowledge, no existing literature has assessed youth willingness to answer questions regarding past sexual fluidity. Importantly, we found the number of individuals missing sexual fluidity data was small (n=51 missing identity change and n=34 missing attraction change out of 4,245, respectively), and appeared to be influenced by survey fatigue (i.e., these questions were toward the end of the survey) rather than hesitance to answer. This suggests that youth are open to answering questions about sexual fluidity, providing support for their inclusion in surveys.

While we were unable to conduct statistical tests of differences in sexual fluidity missingness among sociodemographic subgroups due to small sample sizes, missingness was more common among cisgender individuals and those residing in rural areas. This may be a function of experience; some research has found prevalence of sexual fluidity among cisgender individuals is lower than among transgender and nonbinary individuals (Katz-Wise et al., 2016, 2022). It is possible that cisgender individuals are more likely to skip these questions because they lack these experiences or are unsure how to respond. Similarly, individuals in rural areas may be less familiar with the concept of sexual fluidity (and other aspects of sexual orientation) than individuals in urban areas (Page, 2017), and may be more likely to skip these questions. More research is needed to understand why these groups may be slightly less likely to respond to sexual fluidity questions.

Study Strengths and Limitations

Our study has several strengths. Our sample included responses from a large, racially/ethnically and gender diverse sample of US youth, ages 14-25 years. Previous studies of missingness typically collected data from adults aged 18 years and older; our findings provide new insight into the unique experiences of adolescents. Additionally, our measures of sexual orientation identity, attractions, and gender identity were inclusive of nonbinary identities, allowed for the selection of multiple response options, and included an open-ended option for self-identification.

Online data collection is a strength, contributing to the diversity of our sample. However, the willingness of individuals to respond to SOGIE questions anonymously and online may not reflect the willingness of individuals to respond to similar questions in-person or in other non-anonymized settings (e.g., clinics or schools) that may compromise safety or confidentiality (Meckler et al., 2006; Robertson et al., 2018; Temkin et al., 2017). In addition, the current study oversampled LGBTQ+ youth and recruitment materials described the study as focusing on sexual orientation and attractions. Thus, individuals who chose to participate in this study may be more willing to answer SOGIE and sexual fluidity questions. Use of a convenience sample from an existing online participant panel has limitations. The Prodege panel may have specific unknown characteristics that are not shared by all US youth. For example, individuals are recruited to participate in the Prodege panel via advertisements on TV, radio, and online. Although most US youth have access to the internet, particularly via smartphones (Vogels et al., 2022), the panel does not include youth who were not exposed to Prodege advertisements on one of these platforms. Thus, our findings may not be generalizable to the broader US youth population.

Our survey was US-based and only available in English. Given the potential influence of location, language, and cultural background on societal norms and understandings of SOGIE and LGBTQ+ identities (Jans et al., 2015; Michaels et al., 2017; Reisner et al., 2014), more research is needed to assess missingness patterns in other geographic and cultural contexts. Finally, the measure used to assess attraction did not assess degrees of attraction to each gender, which did not allow participants who were attracted to more than one gender to indicate that they were more or less attracted to each gender group (e.g., being mostly attracted to one gender, but sometimes attracted to other genders). Future research could assess attraction with greater nuance to capture a wider diversity of youths’ experiences with attraction.

The SO*FLY study is a prospective cohort study with seven waves of data collection with the same participants. Thus, we are well-positioned to conduct future analyses examining prospective changes in SOGIE, which can play a key role in improving inclusivity of SOGIE questions. The language individuals use to identify their sexual orientation and gender identity is vast and can change over time (Eliason & Streed, 2017; Russell et al., 2009). Prospectively tracking these changes with open-ended questions may lead to faster implementation of questions with response options that better reflect participants’ identities. Doing so may increase likelihood of answering these questions, though to our knowledge, this has not been examined.

Implications for Research and Practice

Relatively low levels of missingness in this study suggest youth are willing to respond to questions about SOGIE and sexual fluidity. Thus, SOGIE questions should be included in youth surveys, as well as in clinical practice with youth. The Youth Risk Behavior Surveillance System includes some SOGIE items (e.g., sexual orientation identity and gender identity) but could be improved by adding items assessing attractions and sexual fluidity (Centers for Disease Control, 2022). Given the prevalence of sexual fluidity among youth, it is pertinent to assess these questions at multiple time points. Asking SOGIE and sexual fluidity questions in research and clinical settings acknowledges the multidimensional, dynamic nature of sexual identity, behavior, and attraction, and improves our understanding of how these factors develop and shift over time (Ruberg & Ruelos, 2020; Suen et al., 2020). It also allows researchers and clinicians to identify and address the unique needs of LGBTQ+ individuals and the health inequities experienced within this population (National Academies of Sciences, 2020).

Findings from this study also have implications for methods used to address missingness in survey data, particularly the use of multiple imputations (MI). While MI is a common method to address bias due to missingness, one should consider the magnitude of missingness, type of missingness, and variables with missingness for MI use with SOGIE and sexual fluidity questions. First, MI may not be necessary if the levels of missingness are already low among these questions. Although guidance varies, MI provides negligible benefit with less than 5% missingness (Schafer, 1999). However, when there are large amounts of missingness, MI can introduce bias not present in a complete case analysis. Second it can be challenging to assume that data in SOGIE and sexual fluidity questions is missing at random (MAR) as the missingness only depends on observed data. The use of multiple imputation for these questions may not fully address non-response bias if they are missing not at random (MNAR), even after conditioning on observed data (Li et al., 2015). Third, missingness on the type of variables (exposure, confounders, covariates, or outcome) also affects the role of missingness on bias and efficiency, thus influencing MI’s performance in improving them (Lee & Carlin, 2012). Finally, in addition to empirical considerations, there may be ethical concerns related to imputing data for social identity-related questions (Brown et al., 2021; Randall et al., 2021; Woods et al., 2023). We encourage researchers who are considering use of MI for missingness of SOGIE and sexual fluidity questions to first address these considerations.

Conclusions

Our findings represent a significant contribution to literature exploring SOGIE and sexual fluidity. Missingness for the examined questions was low in this sample and differences in patterns of missingness across sociodemographic subgroups were small. Higher levels of missingness found toward the end of the survey suggests survey fatigue, rather than lack of willingness to respond. SOGIE-related questions should be included in surveys and clinical assessments to inform clinical care, policy-making, interventions, and resource development to improve the health of all youth.