Who counts as family? Family typologies, family support, and family undermining among young adult gay and bisexual men

Abstract

Gay and bisexual men may form chosen families in addition to or in place of families of origin. However, the characteristics of these diverse families remain largely unexamined in the quantitative literature. The purpose of this study was to develop a family typology based on responses from a racially and ethnically diverse sample of young adult gay and bisexual men (YGBM) recruited from the Detroit Metropolitan Area (N=350; 18–29 years old). To explore the role of family, we then examined family social support and social undermining in relation to YGBM psychological distress within different family types. A series of multivariate regressions were used to examine associations between family social support and social undermining with depression and anxiety outcomes. The majority (88%) of YGBM included family of origin in their definitions of family and 63% indicated having chosen families. Associations between family social processes and psychological outcomes varied by type of family, suggesting that family composition shapes how perceptions of support and undermining relate to experiencing symptoms of depression and anxiety. Chosen families play a prominent role in the lives of YGBM and should not be overlooked in family research. Findings also highlight the importance of examining co-occurring family social support and social stress processes to further address psychological distress symptoms among YGBM.

Psychological distress, often characterized by symptoms of anxiety and depression (Veit & Ware, 1983), is more prevalent among sexual minority populations compared to heterosexual populations (Balsam et al., 2005; Bostwick et al., 2010; Cochran & Mays, 2000; Cochran et al., 2003; Gilman et al., 2001; King et al., 2008). In a nationally representative sample of adults, Bostwick and colleagues (2010) found that lesbian, gay, and bisexual (LGB) individuals were more likely to have experienced a variety of mood (including major depression) and anxiety disorders in the past year or at any point in their lifetime. Smaller community samples, consisting of adolescents and young adults, show similar trends (Hatzenbuehler et al., 2008; Spencer and Patrick, 2009; Ueno, 2010). Psychological distress is an important risk factor and correlate of externalizing behaviors such as substance abuse and suicide attempt, which may also explain the higher prevalence of these behaviors among LGB populations (King et al., 2008). These disparities exist across various stages of life, posing particular challenges to the health and well-being of sexual minorities from adolescence through young and older adulthood.

The family is a source of both stress and support, and is an important yet understudied area of research with respect to the mental health of sexual minority youth and young adults (Carpineto et al., 2008; Goldfried & Goldfried, 2001; Ryan et al., 2009, 2010; Willoughby et al., 2010). The primary focus of this research has been on parents’ reactions to an adolescent child’s initial disclosure of an LGB identity and the subsequent strain on the parent-child relationship, often characterized as the rejection or loss of social support experienced and reported by LGB youth (Bouris et al., 2010; D’Augelli et al., 2005; Dew et al., 2006; Jadwin-Cakmak et al., 2015; Willoughby et al., 2006). Recent evidence suggests that parental influences extend beyond adolescence to play an important and unique role in an LGB child’s transition to young adulthood and in adjustment during young adulthood (Barker et al., 2006; Needham & Austin, 2010; Mustanski et al., 2010; Sheets & Mohr, 2009). Furthermore, heterosexual and sexual minority peers also play important roles for LGB youth in light of challenges they face at home due to negative family reactions or intolerance of their sexual identity (Doty et al., 2010; Ryan et al., 2009). These challenges may shape LGB’s children’s perceptions of whom they see as family; however, even LGB youth in supportive families may still seek to develop family-like relationships with others if they have no sexuality-specific support within the home (Doty et al., 2010).

Family Research among Sexual Minority Youth

As interest in the families and familial relationships of LGB people continues to grow, family and sexual minority health scholars also draw attention to the theoretical and methodological challenges in this area of research. In her seminal work, Weston (1991) presents a rich ethnographic account of the family-like networks of friends and current or former romantic partners, referred to as chosen family, that gays and lesbians form in addition to or in place of biological family, known as family of origin. A significant body of literature captures this practice as it occurs within specific populations, referred to as fictive kinship among African-Americans and compadrazgo among Mexican-American populations (Chatters et al., 1994; Gill-Hopple and Brage-Hudson, 2012). In subsequent qualitative studies, Nardi (1992) explored gay men’s relationships as the building blocks of chosen families and gay communities while Weeks and colleagues (2001) argue that the emergence of chosen families reflects larger shifts in the definition and role of family in society. Applying a queer and feminist lens, Oswald (2009) would call these shifts a “queering” of the family, or a challenge to the heteronormative conceptualizations of family that give recognition to family ties only defined by biology or acknowledged through legal recognition (i.e., “biolegal” ties) (Oswald, 2002).

A contemporary example of chosen families among queer Black and Latino men in particular are those of the House and Ball communities in which “houses,” or groups of queer men with family-like ties headed by a house parent, compete through performance art in events called balls (Bailey, 2013; Kubicek et al., 2013a-b; Kuvalanka, 2012). House and Ball communities have provided spaces for queer Black and Latino youth to build confidence and earn money through competition but also to find social support and relative safety in an otherwise homophobic and racist society (Bailey, 2013). As such, House and Ball communities have become important sites among researchers to understand how these chosen family and peer networks impact the health and well-being, including vulnerability to HIV infection, of sexual and racial/ethnic minority youth (Arnold & Bailey, 2009; Arnold et al., 2016; Galindo, 2013; Kubicek et al., 2013a-b; Wong et al., 2013). With a few notable exceptions, chosen family relationships have rarely been studied in the context of sexual minority health (Dewaele et al., 2011; Muraco, 2006). This leaves many gaps in the literature regarding optimal methods in quantitative research for measuring the presence and role of chosen family networks and family support. Therefore, critical questions around who counts as family and how family interactions impact psychological health and well-being remain unanswered.

We sought to develop a family typology based on answers from young adult gay and bisexual men (YGBM) who were asked to define their families as being comprised by their family of origin, relationships formed via romantic partnerships, and/or roommate or friendship relationships. Across different family types, we examined YGBM’s experiences of family social support and family social undermining in relation to psychological distress. While some researchers have included separate parent, family, and peer/friend social support constructs in quantitative studies of LGB psychological health and distress, very few have examined how LGB individuals themselves define family and subsequently how support or non-support from these families is associated with psychological status (Detrie & Lease, 2007; Doty et al., 2010; Sheets & Mohr; 2009; Shilo & Sava, 2011). Nonetheless, these studies provide important information about the unique role of support by source and type on various mental health outcomes. Both Sheets and Mohr (2009) and Shilo and Savaya (2011) report that support from friends and family uniquely predicted less depressive symptoms and greater life satisfaction; sexuality-specific support was especially predictive of sexuality-specific outcomes such as internalized binegativity and LGB self-acceptance. In contrast, Doty and colleagues (2010) found no main effect of sexuality-specific support (from family and friends) on emotional distress once they accounted for sexuality stress, however; support moderated the relationship between emotional distress and sexuality distress. Only Sheets and Mohr (2009) tested interaction terms between family and friend support, one general and one sexuality-specific, but they were not significant. Conspicuously absent from these studies of supportive networks are measures of support from romantic partners despite evidence that significant numbers of sexual minority youth seek and form romantic partnerships and benefit psychologically from them (Bauermeister et al., 2010; Bauermeister et al, 2011; Glick & Golden, 2013). A general focus of existing research on support within GBM romantic partnerships has been on HIV-related social support and sexual risk behavior (Bauermeister et al., 2011; Fergus et al., 2009; Goldenberg and Stephenson, 2015). Thus gaps remain in understanding YGBM partner support as embedded in a network of support and in relation to overall health and well-being. In the current study, we avoid partitioning of support source and instead ask YGBM to report on perceptions of general support and non-support based on their defined family typology.

Social Support and Social Undermining

Social support may be particularly advantageous for LGB individuals given their marginalized status in society and the related physical and psychological effects of sexuality-based discrimination (Hatzenbuehler, 2009; Hatzenbuehler et al, 2011). A challenge in understanding the influence of social support on health is that relationships often consist of supportive and non-supportive social exchanges, a dynamic often overlooked in social network and social support research (Abbey et al., 1985; Rook, 1984). These non-supportive exchanges (also called social undermining, social strain, social conflict, and negative social support) consist of attitudes or behaviors, directed from one person to another, that serve to intimidate, criticize and devalue, or create obstacles (Vinokur & van Ryn, 1993). They may be subtle and occur over prolonged periods of time rather than as short bursts of overt aggression. Evidence generally suggests that social undermining (term used herein) is associated with psychological distress and decreased well-being whereas social support is primarily associated with well-being (Newsom et al., 2005; Vinokur & van Ryn, 1993). The relative strength of these associations, however, varies across relationship conditions (e.g. type and closeness) such that support and undermining may both correlate strongly with distress under one set of conditions while either support or undermining alone is the more significant predictor of distress under other conditions (Abbey et al., 1985; Gant et al., 1993; Walen & Lachman, 2000). Abbey and colleagues (1985), for example, found that support and undermining were associated with four measures of well-being (depression, anxiety, interpersonal sensitivity, quality of life) when the support source was phrased as people and the undermining source as some one person but not as the person closest to you. Walen and Lachmen (2000) compared support and strain from family, friend, and partner and found that support was associated with psychological status (life satisfaction, positive mood, negative mood) regardless of source but that strain was associated only when coming from family and partner. Therefore, the link between perceptions of support and undermining to psychological outcomes may depend on the type of relationship but also on the specificity of the phrasing.

A family social network framework that incorporates support and undermining may facilitate better understanding of LGB youth’s experiences with their families during emerging adulthood. When LGB youth disclose their sexual identity to family of origin, disclosure may lead to increased support and closeness between family members or it may cause conflict and even lead to abuse or complete rejection of an LGB child (Carpineto et al., 2008; D’Augelli et al., 2010; Gonzalez et al., 2013; Savin-Williams and Ream, 2003; Willoughby et al., 2006). Therefore, LGB youth must carefully weigh the risks and benefits of disclosure. They must decide whether to disclose but also when, where, and to which family members. The initial reactions from family members may have prolonged effects on the quality of the relationships and on the physical and mental health of LGB individuals (Ryan et al., 2009; 2010). While disclosure to family of origin is an important milestone for many youth, considerably less attention has been given to family dynamics post-disclosure as LGB youth and family address issues related to dating and romantic partners, faith and spirituality, and disclosure to non-parental family (D’Augelli, 2005; Floyd et al., 2006; Phillips & Ancis, 2008; Rothblum et al., 2005). Thus, families remain an important influence on LGB young adults’ well-being yet social support and social undermining functions across family types remain largely undefined in the literature.

Study Aims and Hypotheses

Based on the concept of chosen families and on theories of social networks and social support, the current study focuses on the association between psychological distress and family social support and social undermining. In previous quantitative studies of LGB mental health and social support from family and friends, definitions of family networks have rarely been explored and have often been seen as mutually exclusive from friendship networks. According to previous research, particularly ethnographic and qualitative studies, sexual minorities form chosen families consisting of relationships between people not necessarily related by biology (i.e., family of origin) nor through legally-recognized partnerships. Therefore, the first aim of this study is to describe a typology of families based on the definitions of YGBM, and then to examine the levels of social support and social undermining across family types. The second aim of the current study is to examine the relationships between family social support and family social undermining with psychological distress while accounting for internalized homophobia and degree of sexual orientation disclosure. Findings from previous research suggest that these two latter variables represent markers of sexual identity development that reflect real or perceived experiences of rejection and affect access to social support. To further understand the relationship between social support, social undermining, and sexual identity development with psychological distress, we examine these relationships within and across family types.

Methods

Sampling

Data come from the United for HIV Integration and Policy (UHIP) study, an academic-community partnership conducted by the University of Michigan School of Public Health in collaboration with community partners in the Detroit Metropolitan Area (DMA). The overall study goal of UHIP is to examine the social and structural determinants of vulnerability to HIV infection among young adult men who have sex with men (MSM) living in the DMA (Bauermeister et al., 2014). Participants were eligible if they indicated a male sex assignment at birth, were between 18–29 years old, were currently living in the DMA, and indicated ever having sex with men. Researchers used a convenience sampling technique, recruiting study participants at LGBTQ-related events and venues, including bars, clubs, health fairs and other community events, and on-line social networking sites in collaboration with community partners. Potential participants were approached and received an explanation of the study in addition to a palm card outlining the basic eligibility requirements with a link to the on-line survey and mention of the $30 incentive for completing the survey.

The original UHIP dataset had 429 participants. For the purposes of this analysis, we focus on 350 cismale participants who reported identifying as gay or bisexual, and who did not have any missing data on our variables of interest. Data describing the psychosocial experiences of transgender participants (N=33; 8%) are reported elsewhere (Bauermeister et al., 2016). Thus, the final sample for this analysis consisted of 350 gay (89%) and bisexual (11%) cisgender male participants between the ages of 18 and 29 (M = 23.0, SD = 2.9).

Procedures

Upon entering the on-line study site, users were asked to provide an e-mail address they could use to return and complete the survey later, if necessary, and receive a $30 e-gift certificate upon completion. If eligible, as determined by responses to screening questions at the beginning of the survey, users were subsequently invited to participate and complete an on-line informed consent form. Consented participants completed a 30-minute on-line survey that including the following general domains: sociodemographic characteristics, general and physical health and health behavior, sexuality and gender identity and expression, social networks and social support, general and LGBTQ-specific community involvement, sexual health and sexual behavior, substance use, experiences with discrimination, and mental health and other psychosocial assessments.

All study data were protected with a 128-bit SSL encryption and kept within a university-based firewalled server. During data collection, data were reviewed for duplicates and falsified entries were removed by examining participants’ e-mail and IP addresses. The researchers obtained a Certificate of Confidentiality to further protect study data. Our Institutional Review Board reviewed and approved all study procedures.

Measures

Family Typology, Social Support & Social Undermining

We first asked participants, “Whom do you consider to be your family?” Participants could choose from any of the following categories (check all that apply) to define their family: Your family of origin; Roommates or a group of friends; A family formed by marriage or partnering; Other. Participants were grouped into 6 mutually exclusive categories of family types based on their responses.

We then asked participants to keep their definition of family in mind as they answered a subsequent set of questions about their experiences of support and undermining. Participants were asked to indicate the extent to which they received social support from their families. A 10-item scale was used to capture the four functional domains of social support proposed by House (1981): appraisal, emotional, informational, and instrumental. Together, these items represent an underlying factor of universal social support (Vinokur and Vinokur-Kaplan, 1990; Vinokur et al., 1987). Participants indicated how often their families behaved in a certain way or made certain comments. Sample items include “Say things that raise your self confidence,” “Show that they care about you as a person,” “Give you useful informational or advice when you need it,” and “Provide you with direct help … how much they do things for you.” Participants responded on a 4-point scale (1=Never, 2=Rarely, 3=Sometimes, 4=Often). A family social support (Family-SS) score was calculated by averaging the responses to the ten items, where a higher score was indicative of greater perceived social support from family (α = .94).

Social undermining from the family was assessed using a 7-item scale. Social undermining consists of behaviors that diminish a person’s self-worth and counteracts the benefits of supportive behaviors (Abbey et al., 1985; Vinokur and van Ryn, 1993; Vinokur and Vinokur-Kaplan, 1990). Participants indicated how often their families behaved in a certain way or made certain comments. Sample items include “Act in an unpleasant or angry manner toward you,” “Criticize you,” and “Get on your nerves.” Participants responded on a 4-point scale (1=Never, 2=Rarely, 3=Sometimes, 4=Often). A family social undermining (Family-SU) score was calculated by averaging the responses to the seven items, where a higher score was indicative of greater social undermining from family (α = .91).

Psychological Distress

Depressive symptoms were assessed using a 10-item scale adapted from the Center for Epidemiologic Studies Depression (CES-D) Scale, a validated survey of clinically significant distress as a marker for clinical depression (Radloff, 1991). Sample items include “I felt that everything I did was an effort,” “I felt hopeful about the future,” and “I felt lonely.” Participants responded on a 4-point scale (1=Rarely or none, 2=Occasionally, 3=Some or a little time, 4=All of the time) for how they had felt in the past week. Two of the ten items were reverse-coded to match the remaining items in valence. A depression score was calculated by averaging the responses to the ten items (α = .80).

Anxiety symptoms were assessed and scored using a 6-item subscale from the Brief Symptom Inventory (BSI), which has been validated with sexual minority youth populations (D’Augelli, 2002; Derogatis & Melisarato, 1983; Rosario et al., 2009). Participants responded to 6 statements to indicate how they had felt or what they had thought in the past week. Samples items include “Nervousness or shakiness inside” and “Feeling so restless you couldn’t sit still.” Participants responded on a 5-point Likert scale (1=Never, 2=Almost never, 3=Sometimes, 4=Fairly often, and 5=Very often). An anxiety score was calculating by averaging the responses to the 6 items (α =.92).

Comfort With Sexual Identity and Disclosure of Same-Sex Attractions

Overcoming internalized homophobia and disclosing one’s sexual orientation represent important internal and external processes undertaken by LGB youth as part of their identity development (D’Augelli, 1994; Meyer, 1995). IH is associated with other symptoms of psychological distress such as depression and anxiety, which can influence or be influenced by perceptions of support and conflict in personal relationships (Mohr & Fassinger, 2003; Newcomb & Mustanski, 2010; Sheets & Mohr, 2009). Furthermore, degree of sexual orientation disclosure is also associated with IH and can similarly impact or reflect access to and perception of social support including initial or long-term conflict resulting from disclosure (Cox et al., 2011; Mohr & Fassinger, 2003; Shilo & Savaya, 2011).

Discomfort with one’s sexual identity was measured as internalized homophobia using a 9-item scale adapted from the Internalized Homophobia Scale (Meyer, 1995). Participants were asked to indicate the extent to which they agreed with statements about the level of comfort with their gay/bisexual identity or attraction towards men. Sample items include “I wish I weren’t[gay or bisexual],” “I have tried to stop being attracted to men in general,” and “I feel alienated from myself because of being [gay or bisexual].” The survey was programmed to substitute the bracket in each statement with the participant’s previously selected sexual identity. Participants responded on a 4-point scale (1=Strongly disagree, 2=Disagree, 3=Agree, 4=Strongly agree). We calculated an IH score by averaging the responses to the nine items, where a higher mean score was indicative of internalized homophobia (α = .92).

Sexual orientation disclosure (SOD) was measured by asking participants whether they had disclosed their feelings of same-sex attractions, to specific people in their social network: mother, father, siblings, other family members, friends, and co-workers (Waldo, 1999). Participants had the options of responding Yes, No, or Not Applicable for each category for a total of six responses. A proportion was calculated by dividing the number categories in which disclosure had occurred by the total number of possible categories, excluding categories from the denominator where a participant responded Not Applicable. This proportion indicates the extent to which a participant had disclosed their feelings of same-sex attractions given their specific social network. On average, participants were out to 75% of their social network.

Sociodemographic Characteristics

We collected demographic information from participants that included their age, sexual identity, race and ethnicity, highest level of education completed, student status, and employment status. To further assess participants’ living arrangements, we asked if they lived with other people, how many people lived in their household, and whom they lived with based on a list of eight relationship types (e.g., friend/roommate, mother, sibling, other male relative). We also asked whether participants were currently in a relationship and asked them to qualify their relationship as serious, casual, or other. Finally, providing a 4-point response scale, we asked participants how many of their friends shared their same sexual orientation (1=None of them; 4=All of them).

Data Analytic Plan

Descriptive statistics are provided to elaborate on the sociodemographic characteristics (e.g., household composition, relationship status, peer network homogeneity) of study participants. We conducted chi-squared and ANOVA tests to examine any bivariate differences across family types. Bivariate correlations were used to initially examine general associations between study variables. Three sets of regression analyses were conducted to examine (1) main effects of support and undermining in the whole sample, (2) interaction effects between support and undermining in the whole sample, (3) main effects of support and undermining when stratified by family type. Depression and anxiety were modeled as separate outcomes in our regression analyses.

In the first set of regression analyses, our multivariate regressions included age and race/ethnicity as covariates and support, undermining, IH, and SOD as predictors. To examine the role of family type in relation to psychological distress, family type was also included as five dummy variables with family-of-origin serving as the excluded reference category. In the second set of analyses, we examined the relationship between support and undermining in relation to the outcomes. The previous regressions models were repeated with the inclusion of an interaction term between support and undermining in the final step. Significant interactions were plotted and simple slopes were estimated to gain a better understanding of the roles that support and undermining have when they occur simultaneously. In the third set of analyses, we examined how the relationships between predictors and outcomes varied across family types. In sensitivity analyses, we tested for differences in the magnitude of the associations observed (i.e., test of slopes) between the different family type models.

Results

Preliminary Analyses

Table 1 displays participants’ sociodemographic characteristics. The vast majority of participants identified as gay (89%) while nearly half the sample identified their race/ethnicity as Black or African-American (48%) followed by White (29%), Hispanic or Latino (15%), and Multi-racial/ethnic (9%). Nearly 60% of the sample had at least some college education; approximately 45% of the sample indicated they were currently a full or part-time student. The majority of participants had full or part-time employment (69%) while a significant proportion indicated they were unemployed but looking for a job (25%).

Table 1.

Sample Demographics (N=350)

	N	%
Age category (years)
18 – 20	74	21.1
21 – 23	127	36.3
24 – 26	106	30.3
27 – 29	43	12.3
Sexual Identity
Gay	313	89.4
Bisexual	37	10.6
Race/Ethnicity
Black or African-American	168	48.0
White	101	28.9
Hispanic or Latino	51	14.6
Multi-racial/ethnic	30	8.6
Education Reached
Some HS	27	7.7
HS/GED	87	24.9
Technical/Associates	32	9.1
Some college	118	33.7
College	53	15.1
Graduate school	33	9.4
Current Student
Full-time	104	29.7
Part-time	52	14.9
No	194	55.4
Current Employment
Full-time	136	38.9
Part-time	104	29.7
Unemployed - looking	88	25.1
Other	22	6.3
Group(s) identified as family
(1) Origin	130	37.1
(2) Partners	10	2.9
(3) Friends/roommates	32	9.1
(1) – (2)	25	7.1
(1) – (3)	101	28.9
(1) - (2) - (3)	52	14.9
Lives with others	249	71.1
Average number of people in household (n=249)	3.06	6.68
Household members (n=249)
Mother	102	41.0
Father	57	22.9
Sibling(s)	65	26.1
Children	2	0.8
Other relatives	31	12.4
Friend(s)/roommate(s)	106	42.6
Partner(s)	19	7.6
Other	7	2.8
Has a partner/boyfriend	145	41.4
Partner type (n=145)
Serious	108	74.5
Casual	33	22.8
Other	4	2.8
Friends who share the same sexual orientation
Almost all of them	78	22.3
Some of them	140	40.0
A few of them	117	33.4
None of them	15	4.3

When asked to describe their families, participants constructed their family networks in a variety of ways. The largest group of participants (37%) stated that their family consisted only of their family of origin. The second largest group (29%) said it was both their family of origin and their friends or roommates. The third largest group (15%) was those who indicated all three family type options: family of origin, friends or roommates, and family by partnership. Smaller proportions chose only their friends/roommates (9%), their family of origin and partners (7%), or their family by partnership only (3%) to represent their families. A single individual defined his family as family by partnership and friends/roommates (0.3%). There were also eight individuals who indicated that their families were not defined by any of the three given family type categories (2%). These latter two groups were excluded from further analyses.

Among the 249 participants (71%) who said that they lived with other people, 29% indicated living with one other person, 28% with two other people, 23% with three other people, and 21% with four or more other people. When asked to describe who these household members were, most participants identified them as their friends or roommates (43%) and their mother (41%). About one-quarter (26%) stated they lived with a sibling and 23% stated they lived with their father. A smaller proportion said they lived with another relative (10%) and their partners (8%). In general, 42% of the entire sample also indicated they currently had a partner. When participants were asked how many of their friends share the same sexual orientation as themselves, 22% said “almost all of them,” while 40% said “some of them,” 33% said “a few of them,” and only 4% said “none of them.” We found no differences in household size and composition, relationships status, and peer network homogeneity across the three family typologies included in subsequent analyses.

Table 2 displays the correlations, means, and standard deviations for the primary study variables. Both depression and anxiety symptoms were significantly associated with Family-SS (depression: r=−.30, p<.01; anxiety: r=−.12, p<.05) and Family-SU (depression: r=.32, p<.01; anxiety: r=.35, p<.01) as well as with IH (depression: r=.22, p<.01; anxiety: r=.26, p<.01). Family-SS and Family-SU were negatively correlated (r=−.21, p<.01). Family-SS was positively associated with SOD (r=.13, p<.05). Family-SU was positively associated with IH (r=.20, p<.01). We observed a negative association between IH and SOD that was significant and moderate (r=−.33, p<.01). Age was not associated with our study variables.

Table 2.

Pearson Correlations Among Primary Study Variables

	M	SD	1	2	3	4	5	6
1. Depression	1.94	0.03
2. Anxiety	1.75	0.05	0.66 **
3. Family Social Support	3.18	0.04	−0.30 **	−0.12 *
4. Family Social Undermining	2.14	0.04	0.32 **	0.35 **	−0.21 **
5. Age	23.07	0.15	0.06	0.04	0.04	−0.01
6. Internalized Homophobia	1.61	0.04	0.22 **	0.26 **	−0.06	0.20 **	−0.05
7. Sexual Orientation Disclosure	0.75	0.02	−0.03	−0.03	0.13 *	−0.04	0.03	−0.33 **

^*p < .05,

^**p < .01

Associations of Support and Undermining with Psychological Distress

To address the second study aim, hierarchical multivariate linear regressions were modeled to examine the association of Family-SS and Family-SU with depression and anxiety symptom levels for the entire sample. We modeled each regression in three steps to test the main (Step 1 and 2) and stress buffering (Step 3) effects of support on undermining. Age and race/ethnicity were included as covariates in both models (Step 1) along with family type, internalized homophobia and sexual orientation disclosure. Family type was included as a categorical variable with family of origin as the referent. Support and undermining were subsequently entered (Step 2) followed by their interaction term (Step 3).

Table 3 includes the results from multivariate regression analyses modeling depression and anxiety symptoms. In all three steps, neither age nor race/ethnicity were significantly associated with depressive symptoms. In Step 1, IH (β = .20, p < .001) was associated with increased depressive symptom levels but no other associations were significant. With the addition of support and undermining in Step 2, IH (β = .17, p < .001) and undermining (β = .18, p < .001) were both positively associated with depressive symptoms. Support (β = −.22, p < .001) was negatively associated with depressive symptoms. Compared to those who defined family as family of origin, assigning friends as family (β = .22, p < .05) was also positively associated with depressive symptoms. The addition of support and undermining in Step 2 significantly increased the variance explained by the model (R²Δ = .14, p < .001). Addition of the support x undermining interaction term in Step 3 maintained the previous associations, with the interaction term being significant and negatively associated with depressive symptoms (β = −.13, p < .01). The full interaction model explained 20.2% of the variance in depressive symptoms (R² = .20, F(13, 336) = 7.20, P< .001). Using simple slope tests, we found that the association between undermining and depressive symptoms was attenuated at higher levels of support (Figure 1). The association between undermining and depressive symptoms was only statistically significant for participants who reported average support (β = .24; p<.001). The association between undermining and depressive symptoms was not observed for individuals in the lower or higher support groups. We also sought to examine whether the strength of the association between undermining and depressive symptoms differed across the three levels of support, yet we found no evidence that the beta coefficients were statistically significant from each other.

Table 3.

Multivariate Regressions by Psychological Distress Outcomes (N=350)

	Depression			Anxiety
	Step 1	Step 2	Step 3	Step 1	Step 2	Step 3
Age	0.01	0.01	0.02	0.01	0.01	0.01
Race/Ethnicity (White)
Black/A-A	0.08	0.00	0.02	−0.15	−0.22	−0.21
Latino	0.02	0.01	0.03	0.00	0.01	0.02
Other	0.04	−0.02	−0.05	0.00	−0.11	−0.13
Family Type (1 - Origin)
(2) Partner	0.06	0.02	−0.03	0.18	0.14	0.09
(3) Friends	0.18	0.22 *	0.42 *	0.21	0.22	0.27
(1) – (2)	−0.12	−0.11	−0.17	−0.03	−0.05	−0.04
(1) – (3)	−0.07	−0.02	−0.04	−0.13	−0.10	−0.11
(1) - (2) - (3)	0.07	0.11	0.20	0.17	0.19	0.21
IHa	0.20 ***	0.17 ***	0.29 ***	0.42 ***	0.33 ***	0.36 ***
SODb	0.12	0.15	0.24	0.16	0.15	0.14
Family-SSc		−0.22 ***	−0.28 ***		−0.11	−0.10
Family-SUd		0.18 ***	0.22 ***		0.38 ***	0.29 ***
Family-SS x Family SUe			−0.13 **			−0.12 ***
(Constant)	1.18 ***	1.54 ***	−1.22 **	0.71	0.42	−0.92 *
F	2.74 **	7.17 ***	7.17 ***	3.23 ***	6.17 ***	6.24 ***
Adjusted R2	0.05	0.19	0.20	0.07	0.16	0.17

^*p < .05,

^**p < .01,

^***p < .001

^ainternalized homophobia,

^bsexual orientation disclosure,

^cfamily social support,

^dfamily social undermining,

^einteraction term

Figure 1.

Relationship Between Depression and Family Social Undermining, as Conditioned by Family Social Support

The previous steps were repeated to model anxiety symptom (Table 3). In all three steps, neither age, race/ethnicity, nor family type were significantly associated with anxiety symptoms. In Step 1, IH (β = .42, p < .001) had the only significant relationship with anxiety symptoms and was associated with greater anxiety symptoms. This association persisted for IH (β = .33, p < .001) in Step 2 with the addition of support and undermining, however only undermining had a significant relationship with anxiety symptoms (β = .38, p < .001) and was associated with greater anxiety symptoms. Overall, addition of support and undermining significantly increased the model variance (R²Δ = .10, p < .001). These associations remained consistent with the addition of the support x undermining interaction term in Step 3. The interaction term had a negative relationship with anxiety symptoms and further increased the variance explained in anxiety symptoms (R²Δ = .01, p < .05). This interaction model explained 17.4% of the variance in anxiety (R² = .17, F(13, 336) = 6.24, P< .001). Using simple slope tests, we found that the association between anxiety symptoms and undermining was attenuated across lower (β = .45; p<.01), average (β = .27; p<.001), and higher (β = .25 p<.05); levels of support (Figure 2). We also sought to examine whether the strength of the association between undermining and anxiety symptoms differed across the three levels of support, yet we found no evidence that the beta coefficients were statistically significant from each other.

Figure 2.

Relationship Between Anxiety and Family Social Undermining, as Conditioned by Family Social Support

Stratified Analyses by Family Type

We performed a stratified analysis to examine these associations within the 6 family typologies. The omnibus F-tests for the regression models were significant for only three of the six family types: (1) Family of Origin (N=130); (2) Family of Origin and Friends (N=101), and (3) Family of Origin, Friends, and Partners (N=52). These three family type models were significant when depression (Table 4) and anxiety (Table 5) symptoms were modeled.

Table 4.

Multivariate Regressions by Family Type, Predicting Depression

	Origin (n=130)			Origin and Friends (n=101)			Origin, Friends, & Partners (n=52)
	B	SE B	β	B	SE B	β	B	SE B	β
Age	0.02	0.44	0.12	0.04 *	0.02	0.21	−0.01	0.03	−0.05
Race/Ethnicity (White)
Black/A-A	0.03	0.11	0.02	−0.14	0.12	−0.13	0.09	0.22	0.07
Latino	0.13	0.14	0.09	−0.17	0.19	−0.09	−0.44	0.28	−0.25
Other	0.30	0.20	0.13	−0.12	0.19	−0.06	−0.49	0.34	−0.22
IHa	0.22 **	0.07	0.29	0.15	0.10	−0.16	0.22	0.14	0.21
SODb	0.32 *	0.14	0.18	−0.13	0.19	−0.07	0.43	0.30	0.21
Family-SSc	−0.27 ***	0.06	−0.36	−0.17	0.09	−0.20	−0.38 **	0.14	−0.38
Family-SUd	0.12 *	0.06	0.18	0.25 **	0.09	0.29	−0.27	0.13	0.03
(Constant)	1.36 **	0.45		0.88	0.66		2.88 **	0.95
F	5.60 ***			3.90 **			2.47 *
Adjusted R2	0.22			0.19			0.19

^*p < .05,

^**p < .01,

^***p < .001

^ainternalized homophobia,

^bsexual orientation disclosure,

^cfamily social support,

^dfamily social undermining

Table 5.

Multivariate Regressions by Family Type, Predicting Anxiety

	Origin (n=130)			Origin and Friends (n=101)			Origin, Friends, & Partners (n=52)
	B	SE B	β	B	SE B	β	B	SE B	β
Age	0.00	0.03	0.00	0.05	0.03	0.17	−0.02	0.04	−0.05
Race/Ethnicity (White)
Black/A-A	−0.17	0.20	−0.09	−0.36 *	0.17	−0.24	−0.18	0.28	−0.10
Latino	0.11	0.25	0.04	−0.38	0.27	−0.14	−0.54	0.35	−0.21
Other	0.13	0.36	0.03	−0.30	0.28	−0.11	−0.73	0.42	−0.22
IHa	0.28 *	0.12	0.20	0.08	0.15	0.06	0.79 ***	0.17	0.51
SODb	0.29	0.26	0.10	−0.44	0.27	−0.16	1.48 ***	0.37	0.50
Family-SSc	−0.14	0.11	−0.10	−0.10	0.13	−0.08	−0.60 ***	0.17	−0.40
Family-SUd	0.42 ***	0.10	0.35	0.48 ***	0.13	0.38	0.16	0.16	0.11
(Constant)	0.65	0.82		0.22	0.96		1.76	1.17
F	4.00 ***			3.40 **			5.63 ***
Adjusted R2	0.15			0.16			0.42

^*p < .05,

^**p < .01,

^***p < .001

^ainternalized homophobia,

^bsexual orientation disclosure,

^cfamily social support,

^dfamily social undermining

Table 4 includes the results from multivariate regressions modeling depressive symptoms for each of three family types: family of origin, F(8,121) = 5.60, p < .001; family of origin and friends, F(8,92) = 3.90, p < .01; and family of origin, friends, and partners, F(8,43) = 2.47, p < .05. Within the family of origin group, both IH (β = .22, p < .01) and SOD (β = .32, p < .05) were positively associated with depressive symptoms as was undermining (β = .12, p < .05). In contrast, support (β = −.27, p < .001) was strongly associated with fewer depressive symptoms. This model accounted for 22.2% of the variance in depressive symptoms among YBGM with families of origin (R² = .22, F(8, 121) = 5.60, P<.001). Among participants who indicated family of origin and friends as their family, age (β = .04, p < .05) and undermining (β = .25, p < .01) were both associated with greater depressive symptoms. No other associations in this group were significant. The amount of variance explained in this group was 18.8% (R² = .19, F(8, 92) = 3.90, P<.01) Finally, results for the family type defined by family of origin, friends, and partners indicated that support (β = −.38, p < .01) was negatively associated with depressive symptoms. No other associations were significant in this group. The model explained 18.7% of the variance in depressive symptoms for this family type (R² = .19, F(8, 43) = 2.47, P<.05). When we compared the beta coefficients between our three models, we found no statistical differences in the magnitude of the observed associations.

Table 5 includes the results from multivariate regressions modeling anxiety symptoms, repeated in the same manner as they were for depressive symptoms, among three family types: family of origin, F(8,121) = 4.00, p < .001; family of origin and friends, F(8,92) = 3.40, p < .01; and family of origin, friends, and partners, F(8,43) = 5.63, p < .001. Among YGBM with family of origin only, IH (β = .28, p < .05) and undermining (β = .42, p < .001) were both associated with anxiety symptoms. With no other significant associations, this model explained 15.4% of the variance in anxiety symptoms (R² = .15, F(8, 121) = 3.94, P<.001). Similarly, undermining was negatively associated with anxiety symptoms (β = .48, p < .001) in the family of origin and friends group. In this group, identifying as Black (β = −.36, p < .05) was negatively associated with anxiety symptoms as compared to White individuals. This model explained 16.1% of the variance in anxiety symptoms (R² = .16, F(8, 92) = 3.40, P<.01). In the last group, family of origin, partners, and friends, both IH (β = .79, p < .001) and SOD (β = 1.48, p < .001) were associated with greater anxiety symptoms. Support (β = −.60, p < .001) was negatively associated with anxiety symptoms while there was no significant association between undermining and anxiety symptoms. Lastly, this model accounted for 42.1% of the variance in anxiety symptoms (R² = .42, F(8, 43) = 5.63, P<.001). When we compared the beta coefficients between our three models, we found no statistical differences in the magnitude of the observed associations.

Discussion

This study is among the first quantitative studies to examine how family typologies and processes (family social support and family social undermining) are associated with YGBM mental health. Based on previous ethnographic studies of gay and lesbian familial and friendship networks, we asked YGBM in the current study to define their family as family of origin (e.g., parents and siblings), family of friends or roommates, or family formed by marriage or partnering, including any combination of these groups (Nardi, 1992; Weston, 1990). Our results support Weston’s (1990) observations that gays and lesbians form chosen families in addition to or in place of families of origin, and provide a more nuanced and current understanding of the typologies and experiences of support and conflict within YGBM’s families.

We constructed a typology consisting of six different family types among YGBM. The most common types of family were family of origin, family of origin and friends, and family of origin with friends and partners, which all accounted for 81% of YBGM’s definitions of family. Family as friends, family of origin with partners, and family as partners constituted the remaining family types. Overall, 53% of YGBM included friends in their family network while 25% included partners, and just over half of those who included partners were currently partnered. These inclusions of non-biolegal ties supports the theory that LGB populations form chosen family networks (Dewaele et al., 2011; Muraco, 2006; Weston, 1990). Nonetheless, the vast majority (88%) of YGBM still incorporated family of origin into their definitions of family whether it was family of origin alone or in combination with other sub-networks. Therefore, family of origin, which includes parents, remains an important network for YGBM (Goldfried & Goldfried, 2001; Needham & Austin, 2010).

Family social support was protective against anxiety symptoms in our unstratified analyses, highlighting the importance of family support as a protective factor for YGBM (Needham & Austin, 2010; Sheets & Mohr, 2009; Shilo & Savaya, 2011). Family social undermining, on the other hand, was associated with greater symptoms of both depression and anxiety. These findings are consistent with prior research documenting a robust relationship between undermining and psychological distress (Abbey et al., 1985; Taylor, 1991; Vinokur & van Ryn, 1993). Taken together, these findings highlight the dual nature of close relationships where support and conflict can occur through the same social ties yet have independent effects on health and well-being (Abbey et al., 1985; Cranford, 2004; Walen & Lachman, 2000; Newsom et al., 2005). In some cases, however, we also found that the significance and strength of the associations between undermining and distress depended on the existing level of support. While most family studies of sexual minority youth have focused on supportive and non-supportive behaviors separately, we are among the first to examine family-related social support and conflict concurrently, highlighting the importance of examining both positive and negative dimensions of YGBM’s familial relationships (Bouris et al., 2010; Bregman et al., 2012).

YGBM in our sample were embedded in rich social networks. The majority of YGBM lived with multiple people who they mostly identified as friends/roommates and mothers but also siblings and fathers. YGBM had disclosed their sexual orientation to most of their social network; the vast majority had disclosed to their friends and mothers yet only half had disclosed to their fathers. This finding is consistent with previous reports of LGB youth disclosing sexual identity to mothers first, however; LGB youth are increasingly disclosing to fathers and experiencing support from them as well (D’Augelli et al., 1998, 2005; Grafsky, 2014; Grov et al., 2006; Savin-Williams and Dube, 1998; Savin-Williams and Ream, 2003). With respect to peers, most YGBM had a least a few or some friends who shared the same sexual orientation. Nearly half of YGBM who had partners said they were in a “serious relationship” that had been ongoing for an average of about two years. Consequently, YGBM in the current study are likely not experiencing the type of extreme social and physical isolation stereotypically ascribed to sexual minority populations (Gonzalez et al., 2013; Goodrich, 2009; Grossman, 1997; Meyer, 1995).

The magnitude and significance of the associations between family social processes and psychological distress varied when we stratified the sample into the three most prevalent family types. Consistent with prior studies, family support was associated with decreased psychological distress symptoms whereas family undermining was associated with increased symptoms (Newsom et al., 2005; Rook, 1984; Schuster et al., 1990). Interestingly, family support was at times an equally or more important predictor of distress within certain family types (Newsom et al., 2003, 2005; Rook, 1984). Several important patterns emerged by family type regarding the association between psychological distress and family support and undermining. Among YGBM who defined their family as solely their family of origin, both support and undermining were associated with psychological distress. For YGBM whose families included their family of origin and their friends, only family social undermining was associated with psychological distress. Conversely, only family support was associated with psychological distress among YGBM who included family of origin, partner, and friends in their definition of family. Taken together, these findings suggest that family support and undermining may be differentially associated with distress based on the composition of a family. From a social network perspective, Israel (1982) has argued that some actors in a network might share social ties that cross more than one content area (i.e., meaning assigned to a relationship), referred to as multiplex ties. Without taking these multiplex ties into account, examination of family social processes may mask underlying patterns of support and undermining if family is defined only as family of origin (i.e., biological relationships). Future work using sociometric network analyses may be warranted to further examine structural and functional characteristics of YGBM’s origin and chosen family networks (Valente, 2010).

Beyond family typologies, YGBM’s sources of conflict and support might also be tied to their comfort with their same-sex attractions, their sexual orientation disclosure to their social network, and their family of origin’s reactions and understanding of their children’s sexuality(Doty et al., 2010; Needham & Austin, 2010; Ryan et al., 2009, 2010; Shilo & Savaya, 2011; Willoughby et al., 2010). For example, although our measure of family social undermining did not include language specific to sexual identity, undermining and IH were positively correlated in bivariate analyses suggesting a link between YGBM’s discomfort with their own sexual identity and the undermining they perceive from family. In multivariate models, we noted associations between psychological distress and internalized homophobia and sexuality-related disclosure, respectively, for two of the family types (i.e., family of origin and family of origin, friends, and partners). These findings are consistent with prior research suggesting that sexual minority youth’s mental health might be compromised if they experience conflict within the home and/or lack supportive friendships (Bregman et al, 2012; Meyer, 2003; Szymanski et al., 2008). Other social contexts might explain these differences across family types. For example, YGBM who defined their family as comprised solely by their family of origin may be dealing with conflict arising from recent SOD or feel continued pressured to hide or suppress parts of their identity. Conversely, among YGBM defining their families as comprised by families of origin, friends, and partners, the associations we found might be linked to the inclusion of “partner” in this family type. The ability to include or imagine a partner in one’s family network suggests greater attainment or perceptions of support, but this may also come at the psychological cost associated with SOD, introduction of a partner to family of origin, and/or experiences of sexuality-based stigma and discrimination (Frost & Meyer, 2009; Igartua et al., 2003; Meyer, 2003; Otis, 2006). Consequently, although sexual orientation disclosure may have benefits, including garnering sexuality-specific support from accepting social network members, the benefits of disclosure may depend on whether their family’s social environments are welcoming and foster authentic self-expression (Beals and Peplau, 2006; Cox et al., 2011; Legate et al., 2011; Rosario et al., 2009).

We acknowledge several limitations and strengths relevant to the study design and to the interpretation of our results. First, our findings are limited to a non-random community sample of YGBM residing in the DMA. Second, given the cross-sectional nature of our study, we cannot ascertain causality. While it is possible that family-based support/undermining may influence depressive and anxiety symptoms among YGBM, it is plausible that these symptoms also affect support/undermining either directly through YGBM’s perceptions or indirectly by isolating YGBM from their families and other relationships. Third, the study sample itself is diverse with respect to inclusion of participants from racial and ethnic minority groups (71% identified as Black/African-American, Latino/Hispanic, or Multi-racial/ethnic); however, the study’s sample size was a limitation to conducting any subgroup analyses by race/ethnicity. In general, the current findings are a unique contribution to the literature given that many studies of LGBT health in the United States include samples of LGBT populations primarily drawn from large coastal cities.

To our knowledge, we are among the first to develop a family typology of origin and chosen family networks, based on answers provided by YGBM, and to report the prevalence of chosen family networks in this population. To conduct a more thorough family network analysis, however, we would need additional information such as the number of network members and frequency of contact between them, content areas and duration of each social tie, and different time frames (e.g., previous week, previous month, lifetime) over which YGBM had been experiencing the support and undermining asked about. This might also allow for comparisons between the whole network and sub-networks in terms of support and undermining experienced in relation to psychological distress. Due to sample size, we were also unable to conduct further analyses with the three smaller chosen family types we found and the small number of YGBM that indicated having no family at all. An important task for future research in this area is determining the precise role of family in the lives of young adults. Much of the existing literature on LGB populations and their families is specific to the experiences of adolescents and the impact of homophobia and rejection within the family. As a result, current measures of support may be inadequate for examining the role of family support in the lives of LGB young adults.

Implications

Few family and sexual minority health scholars have considered the role of chosen families among young adults in their quantitative research. To date, much of this work is based on heteronormative assumptions used to construct a family binary that “privileges biological and legal ties as ‘genuine’ family and designates other forms of relations as ‘pseudo’” (Oswald,2005, pg.146). This begs a basic question: Who counts as family? When we asked YGBM to define their families based on single or combinations of networks, only 37% of them indicated their family of origin alone, and 19% didn’t include family of origin at all. The remaining participants described families that also consisted of friends and/or romantic partners. Weston (1991) captured these overlapping dimensions of family and friendship relationships in her ethnography of LGB individuals living in San Francisco. Nardi (1992) also observed and chronicled the important family roles of gay men’s current and former lovers. Therefore, by employing a method in which the researcher and participant co-construct definitions of family, researchers more accurately represent the structural and functional qualities of sexual minorities’ family networks. In addition to Weston’s (1991) and Nardi’s (1992) work, and building off of Kahn and Antonucci’s (1980) social convoy model, Pahl and Spencer’s (2010) concept of personal community, defined as “the set of personal relationships that a person considers important” (pg. 205), offers a similar conceptual approach to address these limitations. To this end, we stress the importance of examining both supportive and non-supportive elements of relationship ties in origin and chosen family networks.

The lack of family-based interventions that focus on the health and well-being of YGBM reflects greater gaps in knowledge among researchers about the role of families in these populations. Given the alarming rates of HIV infection disproportionately impacting young adult Black and Latino MSM, for example, the lag in development and evaluation of family-based HIV prevention interventions for YGBM is problematic (Bouris et al., 2010; CDC, 2013; Mustanski et al., 2011). Some evidence suggests that family-related factors such as family acceptance and support of a child’s sexuality, family connectedness, and parent–child communication about sex influences HIV-related sexual risk and HIV testing behaviors among YGBM (Bouris et al., 2015; Glick and Golden, 2013; Resnick et al., 1997; Ryan et al., 2010). Even fewer researchers have included peer and sexual networks, including romantic partnerships, alongside family networks (Kapadia et al., 2013; Latkin et al., 2011; Schneider et al., 2012).

We believe that by asking YGBM about their personally defined family, we might be better equipped to define and understand their families without excluding non-biolegal ties. This approach might also help to clarify the measurement of family support and undermining, respectively, rather than assuming that participants understand “family” to mean parental or caregiver support. Future research acknowledging a broader range of family ties might advance our understanding of family-based influences on health, as well as strengthen the development of family-based interventions for YGBM.