From the Rainbow to the Sunset: Anticipated Availability of Informal Care for Older Sexual Minorities in the United States

Abstract

Anticipatory support—defined as the perceived availability of future care—functions as a psychological safety net that promotes mental well-being and is predictive of actual support received. Yet, little is known about how it varies by sexual identity. Using data of older adults without current ADL limitations from the 2022 wave of the Health and Retirement Study (n = 11,828), I conducted descriptive analyses, modified Poisson regressions, and pairwise comparisons of predicted probabilities of anticipated support by sexual identity. While there is no strong evidence of sexual identity disparities in the overall availability of future ADL support, the sources of such support varied by sexual identity. Among those with at least one child, the probability of lesbian/gay individuals counting on children/grandchildren were about 27 percentage points lower than that of heterosexual individuals. Additionally, the predicted probability of anticipating support from non-kin or someone else outside of a spouse, descendants, or other relatives was about 10 percentage points higher for lesbian/gay older adults compared to their heterosexual counterparts. These findings underscore both the vulnerability and resilience of sexual minority older adults—who may be less able to rely on kin-based care due to strained or absent family ties, yet more likely to turn to chosen family networks for future support. As formal care systems become increasingly strained, future research is needed to understand how chosen families function as caregiving structures and what barriers sexual minorities face in accessing both formal and informal long-term care.

Introduction

Rising longevity, coupled with declining fertility rates, has led to population aging in the United States (US). In 2022, approximately 1 in 6 Americans—about 58 million people—were aged 65 or older. By 2050, this number is projected to rise to 82 million, representing a 47% increase (Mather & Scommegna, 2024). These demographic trends bring to fore the rising demand for old-age care – be it informal or formal. And as the US population rapidly ages, so would its sexual minority members. One study estimates that 2.4% or approximately 2.7 million of older adults aged 50+ in the US identify as lesbian, gay, bisexual, and transgender (LGBT). Among them, around 1.1 million were 65 and older (Fredriksen-Goldsen & Kim, 2017). Despite their smaller numbers, sexual minority older adults are a crucial population to study, as many have faced a lifetime of stigma and discrimination – resulting in distinct disadvantages in health, healthcare access, family formation, and social support. In this study, I investigate the disparities in anticipatory support based on sexual identity.

Anticipatory or perceived support refers to the belief that someone will be available to provide help when the need arises (Krause, 1997; Wethington & Kessler, 1986). This form of support has been shown to promote well-being, particularly mental health as it functions as a social safety net that fosters a sense of security, belonging, and purpose. While anticipatory support is highly predictive of actual support received, it differs in important ways: unlike received support, it does not risk undermining personal autonomy and may instead promote a sense of mastery and encourage problem-solving from the vantage point of the recipient (Cheng, 2017; Krause, 1997; Wethington & Kessler, 1986).

In this study, I focus specifically on anticipatory support in the context of potential future limitations in activities of daily living (ADLs). These limitations affect about half of adults aged 65 and older and can significantly hinder their ability to age in place – especially when adequate support is unavailable (Freedman et al., 2024; H. Wu et al., 2024). Individuals who are minoritized due to their sexuality are at greater risk for functional limitations and chronic conditions, largely due to the cumulative effects of stigma and discrimination (Brooks, 1981; Fredriksen-Goldsen et al., 2017; Meyer, 2003). At the same time, they have historically faced barriers to family formation – particularly because same-sex marriage was only nationally legalized in 2015 – despite family often being the primary source of informal care (Gates, 2015; Westrick-Payne & Manning, 2025). As a result, sexual minority individuals are aging with a shrinking safety net – just as the cracks in that net are widening under the weight it is expected to bear. They face elevated health risks while also having fewer available support resources. This raises an important policy concern, as informal care may serve as either a substitute for or a complement to formal care systems that are increasingly overburdened by rapid population aging (Freedman et al., 2024). Understanding disparities in anticipatory support for future ADL limitations by sexual identity is therefore essential to identifying care gaps and promoting equity in later life.

Not Just Rainbows and Butterflies: The Lasting Health Impacts of Stigma and Discrimination on Individuals Identifying as Sexual Minorities

According to the minority stress model (Brooks, 1981; Frost & Meyer, 2023; Meyer, 2003; Rich et al., 2020) individuals identifying as sexual minorities would be at a higher risk of poor health outcomes compared to the sexual majority. Regardless of sexual orientation or gender identity, individuals who are under physical, mental, or emotional stress—whether stemming from acute events such as illness or the death of loved ones, or from everyday hassles—may experience detriments to their well-being. However, minoritized populations may face social stress, that is, stress brought about by their social environment in the form of norms, institutions, and social structures that propagate exclusion, prejudice, stigma, discrimination, or even violence against certain individuals based on their inherent characteristics. Additionally, the tension between the majority and minority groups may lead to the latter being excluded from participating in health-protective social institutions. One key example of such instance is before 2015, the institution of marriage was only reserved for heterosexual couples (Gates, 2015; Westrick-Payne & Manning, 2025), thereby restricting access to legal and social benefits. Social stress manifests in both distal forms (e.g., discrimination, structural stigma) and proximal forms (e.g., expectations of rejection, hypervigilance, or concealing one’s identity). Over time, these negative expectations, hypervigilance, and being closeted themselves become stressors that further erode health. For these reasons, persons who identify as sexual minorities experience excess stress beyond those faced by heterosexual individuals, which, in turn, contributes to health disparities across sexual identities.

Sexual minority individuals’ greater susceptibility to poor health due to social stress may translate into higher care needs compared to their heterosexual counterparts. A large body of work provides evidence to support the prediction of the minority stress model. For example, compared to heterosexuals, sexual minority older adults were more likely to experience poor mental health (Nelson & Andel, 2020). Compared to heterosexual counterparts, lesbian, gay, and bisexual (LGB) older adults were more likely to have poorer general physical and mental health, weakened immune systems, greater levels of pain and disability, and higher prevalence of chronic conditions such as arthritis, asthma, heart attack, stroke, and cancer (Fredriksen-Goldsen et al., 2017). Bisexual 50–65 year-olds reported far worse health compared to straight individuals of the same age group (Liu et al., 2021). Bisexual older adults and older adults who identified as “something else” apart from being LGB had more activities of daily living and instrumental activities of daily living (ADL/ IADL) limitations which necessitated care compared to their straight counterparts (Travers et al., 2022). The chronicity of stress sexual minority individuals face may lead to increased risk for factors that are deleterious for cognitive health such as cardiovascular disease and depression (Correro II & Nielson, 2020). This, in turn, may imply that older adults who are sexual minorities are at a heightened risk for developing dementia, thereby further increasing their likelihood of requiring care.

Although health disadvantages among individuals who are sexual minorities are well-documented, it is worth mentioning that a few studies found instances when sexual minority older adults were healthier compared to their heterosexual counterparts. For example, although mental health among sexual minority individuals was found to be similar to or worse than that of their heterosexual counterparts, Brown & Grossman (2014) and Nelson & Andel (2020) found that the former reported better self-rated physical health compared to straight individuals. In a sample of older adults with HIV, Porter et al., (2019) demonstrated that older gay and bisexual men living with HIV reported better psychological well-being than their heterosexual counterparts. Sexual minority women hold more positive expectations with regards to their own aging process (Barrett & Barbee, 2017). These findings lend support to the crisis competence model by Kimmel (1978) which posits that overcoming adversity such as stigma and discrimination can build adaptive skills that enhance one’s competence to cope with future challenges.

Sexual Minorities' Social Support Networks in Later Life

Sexual minority older adults face greater risks for health conditions that may require care, making it essential to understand both their anticipated support for future ADLs and its sources. Because social support can buffer the deleterious health effects of minority stress (Breder & Bockting, 2025; Frost & Meyer, 2023; Meyer, 2003), understanding disparities by sexual identity is crucial to promoting health equity in older adulthood. Social support is best understood using the convoy model of social relations (Antonucci et al., 2011; Kahn & Antonucci, 1980) which posits that social networks can be conceptualized in terms of identity and relatedness. In this framework, relationships with spouses, family, friends, or acquaintances form a “convoy” that travels with the individual and can be activated when support is needed, though this is shaped by the life course and developmental context of both the individual and their social network. These relationships are dynamic, entering or exiting an individual's network over time. Social ties are typically described across four domains: relationship closeness, structural characteristics (e.g., size, composition, proximity, frequency of contact), function (e.g., tangible and intangible support), and quality. This current study is focused on anticipatory support, which is an intangible form of support.

Both situational or structural factors (such as family structure) and personal characteristics (such as age, gender) shape the availability, exchange, and quality of support (Antonucci et al., 2011; Breder & Bockting, 2025; Kahn & Antonucci, 1980). Recent scholarship by (Breder & Bockting, 2025) has “queered” the convoy model to better reflect the social position and lived experiences of individuals identifying as sexual and gender minorities by emphasizing the centrality of alternative and chosen families. In terms of structural factors, compared to their cisgender and heterosexual counterparts, older LGBT adults are less likely to have strong ties with their family of origin, more likely to live alone, less likely to have children, and more likely to have experienced network shrinkage due to the HIV/AIDS epidemic (Flatt et al., 2022; Hsieh & Wong, 2020). Stigma and discrimination have contributed to estrangement from biological families (Hsieh & Wong, 2020), while legal and social exclusion from marriage for most of history limited access to institutional support structures (Gates, 2015; Westrick-Payne & Manning, 2025). Same-sex marriage was only legalized nationwide in 2015, significantly limiting access to this institution for sexual minorities (Gates, 2015; Westrick-Payne & Manning, 2025). This is especially true for older adults who were of marrying age decades before Obergefell v. Hodges, the focus of this investigation. Not having a spouse or partner – the individual typically first in line to provide support – may lead older adults to rely more heavily on other family members or friends (H. Wu et al., 2024). After the spouse or partner, the next in line to provide this support having more living children may facilitate the availability of future support, as adult children, especially daughters, constitute the cornerstone of old-age informal support (H. Wu et al., 2024). Additionally, the biological and financial costs of childbearing are significantly higher for same-sex couples (Gates, 2015), helping to explain why LGBT individuals are substantially less likely to have children who are the main source of eldercare (H. Wu et al., 2024).

But despite the emergence of alternative support systems such as chosen family, concerns about care availability in later life remain. A recent study by (Singleton et al., 2022) found that LGB older adults reported a higher perceived likelihood of needing nursing home care in the future – an average of 34% compared to 27% among heterosexual adults. Since formal care, including nursing home services, often serves as a substitute when informal care is unavailable or insufficient to meet complex needs (Freedman et al., 2024; H. Wu et al., 2024), this finding raises important questions. Does this anticipation of nursing home use reflect a heightened expectation of complex care needs shaped by minority stress? Or does it stem from an awareness that informal care may not be available due to strained or absent family ties rooted in discrimination? Importantly, prior research has also documented that individuals who are sexual and gender minorities often face stigma and poor treatment in formal care settings (Flatt et al., 2022; Singleton & Enguidanos, 2022), creating an additional barrier to nursing home use and raising concerns about the adequacy and inclusivity of institutional care. Given this significant barrier, it becomes even more critical to understand what informal care sources are available—and perceived as available—among sexual minority older adults. Thus, in this current study, I examined differences by sexual identity in anticipatory support for future ADL limitations among older adults in the US. I also assessed how the expected source of support – whether they come from descendants, relatives, or non-relatives – varies by sexual identity.

For the reasons cited above, I hypothesize that sexual minority older adults will be less likely than their heterosexual counterparts to anticipate the availability of future care from their social ties outside of their partner/ spouse (H1), anticipate the availability of future care from their descendants (H2), and anticipate the availability of future care from their relatives (H3). And because of the systemic exclusion from typical sources of support, individuals who are sexual minorities have formed of chosen families, which are expected to strengthen over time in tandem with sociohistorical shifts in LGBT rights as sexual minority individuals and their convoys age (Breder & Bockting, 2025). Thus, I additionally hypothesize that sexual minority older adults will be more likely to anticipate that non-relatives will provide care in the future (H4).

To test these hypotheses, I also account for personal characteristics such as age, gender, race-ethnicity, socioeconomic status, and health, which, according to the convoy of social support model, may shape intangible support (Antonucci et al., 2011; Breder & Bockting, 2025; Kahn & Antonucci, 1980). As individuals age, their social networks—and consequently, their sources of anticipatory support—often shrink due to factors such as retirement, voluntary social withdrawal, or the loss of network members through death (Carstensen et al., 2003). Women are also more likely to engage in familial exchanges of support compared to men, in part due to gendered expectations rooted in a patriarchal societal structure, which often assigns women the role of kin keepers (Kim et al., 2020). People from racially and ethnically minoritized backgrounds may also report stronger expectations of support from their descendants and relatives, as cultural norms often emphasize intergenerational and familial responsibility (Taylor et al., 2022). Additionally, due to assortative mating and homophily in social networks, individuals are more likely to share similar socioeconomic status with those in their support systems. As a result, older adults with higher socioeconomic status may be more likely to anticipate future support, given that their social connections are also more likely to have the resources and capacity to provide such support when needs arise (Schafer & Vargas, 2016). Finally, poor self-rated health may reflect chronic conditions that create greater dependence on social support, which over time can strain and deplete available support (Z. Wu & Hart, 2002).

Methods

Data came from the Health and Retirement Study (HRS), a biennial longitudinal survey representative of US adults over age 50 (data is described in greater detail elsewhere, see for example Fisher & Ryan, 2018). The HRS is sponsored by the National Institute on Aging (NIA U01AG009740) and is conducted by the University of Michigan. The survey collects information on topics involving health, finances, labor force participation, social relationships, and psychosocial factors. The HRS employes a multistage probability design, oversamples African American and Hispanic populations. The HRS replenishes its sample every six years by adding new birth cohorts.

The HRS began asking new respondents about their sexual identity in 2016, expanding the question to all remaining respondents by 2022. To maximize the sample size, I focus on the 2022 wave and retrospectively link sexual identity responses to prior waves where applicable. 15,856 individuals were interviewed in 2022¹, and after removing those younger than 51 (n=458), those with at least one functional limitations (i.e., those who are ineligible to answer the anticipatory support question, n=2,500), and those with missing responses to the variables included in the study (n=1,070²), I arrived at an analytic sample of n=11,828. For the models that predict the source of future support, I further confined the analytic sample to (1) those who said help was available in the future (n=7,982) and (2) those who said help was available in the future and had at least one living child (n=7,270).

Measures

The four outcomes of interest were anticipatory support or the perceived availability of relatives (excluding spouses/ partners) or friends to help when future ADL limitations like personal care arise and the source of future support. The HRS asked respondents without functional limitations the following: ‘Suppose in the future, you needed help with basic personal care activities like eating or dressing. Do you have relatives or friends [besides your [husband/wife/partner]] who would be willing and able to help you over a long period of time?’ ‘Yes’ responses were coded as 1, and ‘no’ and ‘don’t know’ responses were coded as 0. Subjects who answered affirmatively were further asked about the source of this future support; options identified were descendants (i.e., child/ child-in-law and grandchild), other relatives, or someone else. For the model predicting the availability of descendants for future ADL care provision, I further excluded the sample to those with at least one living child.

The key independent variable was sexual identity. Respondents were asked: ‘Do you consider yourself to be lesbian/ gay, straight, bisexual or something else?’ Responses were categorized into four groups: straight, lesbian/gay, bisexual, and something else/ other. In 2022, the ‘something else’ category was renamed ‘other’, and these were collapsed for consistency. To maximize the sample size and analytic power, I pooled responses from previous waves. Eighty-one percent of sexual identity responses came from the 2022 wave, while 18% came from 2016, and the remaining responses were from 2018 and 2020. Note that this approach inherently assumes sexual identity to be time-invariant, despite broad recognition that sexual identity is fluid (Veldhuis et al., 2024). I therefore ran a battery of sensitivity analyses (described in the next subsection) to ascertain the robustness of the results.

Based on the convoy model of social support, minority stress theory, and the extant literature described in the previous section, I controlled for the following that could confound the disparity of anticipatory support by sexual identity: age, sex (female, male), race-ethnicity (non-Hispanic Black, Hispanic, non-Hispanic others, non-Hispanic white), educational attainment (high school or less, any college, college and above), marital status (married/ partnered; previously married which includes divorced, separated, and widowed; never married), self-rated health (poor/ fair, excellent/ very good/ good), and the number of living children (0, 1, 2, 3+). Time invariant control variables came from RAND Corporation Center for the Study of Aging’s user-friendly HRS Longitudinal File (version 1, 2022). Given the limited sample sizes among sexual minority respondents, I kept the number of control variables to a minimum and collapsed the categories for certain variables, including marital status, self-rated health, and number of children. All measures are observed in the 2022 wave.

Analytic Plan

Descriptive statistics, including means, frequencies, and prevalences, were generated to summarize key variables. χ² and ANOVA tests were used to assess bivariate differences in selected variables by sexual identity. Given that the outcomes of interest are binary and relatively common, a log-binomial regression would be the preferred approach, as logistic regression tends to overestimate risk when the outcome is frequent. However, due to small sample sizes and sparse cells for the key predictor, the log-binomial models failed to converge. As an alternative, I used a modified Poisson regression with robust standard errors—a widely accepted workaround proposed by Zou (2004) —which provides consistent estimates of relative risk and performs well under conditions of non-convergence in log-binomial models. Additionally, the modified Poisson performs better under model misspecification as opposed to the log-binomial models (Chen et al., 2018). Regression results were presented as incidence risk ratios (IRRs) with corresponding 95% confidence intervals using two-tailed tests. Additionally, I ran pairwise comparisons of the predicted probabilities of the outcomes across sexual identity groups, using a 5% significance level and Bonferroni’s adjustment which provides more conservative comparisons and is not influenced by group sample size (Bland & Altman, 1995). Stata 19 was utilized to conduct these analyses. Note that all analyses were unweighted, given that the 2022 survey weights for Early Gen-Xers were not available at the time this research was being conducted.

Assessing the sensitivity of results based on when and how sexual identity is measured is particularly important, given the lack of consensus in both the literature and the broader community regarding its classification. As mentioned, the limited number of sexual minority respondents in the HRS required me to backfill the sexual identity responses using pre-2022 waves, inadvertently assuming that sexual identity is time invariant. However, because sexual identity is generally considered fluid rather than static (Veldhuis et al., 2024), I conducted sensitivity analyses using only the 2016 cohort to assess the robustness of the findings. Additionally, I assessed the robustness of the results based on how sexual identity was coded. Past work using the HRS have taken different approaches: for example, Singleton et al., (2022) dichotomized it as sexual minority versus heterosexual and excluded respondents who identified as ‘something else’ to avoid potential misrepresentation, whereas Liu et al. (2021) included them for completeness. Thus, I ran the following sensitivity analyses by replicating the main analyses with the following modifications: (1) using the 2016 data on new respondents, (2) aggregating LGB to a single sexual minority category to overcome the small cell size problem, and (3) excluding those who considered themselves to be ‘something else’ or ‘other’ from the dichotomized sexual minority indicator.

Finally, I estimated models that also included respondents with complete data on control variables who answered ‘don’t know’ (n=260), refused to answer (n=58), or had missing responses (n=8) to the sexual identity question (total n=326). This approach acknowledges that both self-identification and non-response may carry meaningful information about a person’s identity, particularly given that the HRS question structure – limited to lesbian/gay, straight, bisexual, or something else/other – may still be construed by some respondents as exclusionary (Veldhuis et al., 2024).

Results

Table 1 presents unweighted descriptive statistics broken down by sexual identity; I only highlight the most notable findings in the text. A majority of the sample identified as straight (94.6%), while individuals considering themselves as sexual minorities collectively constitute 5.4%. Among those who identified as sexual minorities, something else was the most endorsed identity and bisexual was the least endorsed. Nearly 70% of respondents reported that they would have help available if they experienced functional limitations. This prevalence was driven by those identifying as heterosexual and lesbian/gay. In contrast, bisexual respondents and those identifying as something else/ other were a little over 10 percentage points less likely to anticipate support compared to the full sample and heterosexual and lesbians/ gay respondents. Lesbian/gay respondents were significantly more likely to be non-Hispanic white, college-educated, never married, and childless compared to both the full sample and those who identified as straight. Respondents who identified as something else or other were more likely to be female, Hispanic, have less than some college education, report poor health, and have three or more children compared to the complete sample and those who identified as straight. Quantitatively similar patterns were observed in the subsamples restricted to (1) respondents who reported that help would be available in the future (Table S2), and (2) those who reported future help availability and had at least one living child (Table S1).

Table 1:

Unweighted Descriptive Statistics for the Main Analytic Sample

	All (100%)		Straight (94.7%)		Lesbian/ gay (1.7%)		Bisexual (0.8%)		Something else/ other (2.9%)		χ2 / ANOVA
	Freq.	Mean (SD)/ %	Freq.	Mean (SD)/ %	Freq.	Mean (SD)/ %	Freq.	Mean (SD)/ %	Freq.	Mean (SD)/ %
Observations	11,828	100	11,200	100	196	100	92	100	340	100
Future Help Available (Yes)	7,951	67	7,590	68	127	65	51	55	183	54	***
Age	NA	67.99 (9.76)	NA	68.07 (9.77)	NA	63.54 (8.89)	NA	65.49 (9.12)	NA	68.46 (9.61)	***
Female	6,816	58	6,463	58	85	43	51	55	217	64	***
Race-ethnicity											***
Non-Hispanic white	6,380	54	6,146	55	127	65	53	58	54	16
Non-Hispanic Black	2,902	25	2,812	25	32	16	21	23	37	11
Hispanic	1,966	17	1,690	15	32	16	13	14	231	68
Non-Hispanic others	580	5	552	5	5	3	5	4	18	5
Education											***
High school or less	4,837	41	4,515	40	37	19	32	35	253	74
Some college	3,371	29	3,244	29	52	27	29	32	46	14
College	3,620	31	3,441	31	107	55	32	34	41	12
Marital status											***
Married/ partnered	7,088	60	6,742	60	112	57	49	53	185	54
Previously married	3,849	33	3,675	33	27	14	26	28	121	36
Never married	891	8	783	7	57	29	17	18	34	10
Poor/ fair health	2,696	23	2,493	22	34	17	23	25	79	43
Number of children											***
0	1,217	10	1,041	9	125	64	19	21	32	9
1	1,465	12	1,388	12	27	14	19	21	31	9
2	3,399	29	3,266	29	28	14	31	34	74	22
3+	5,747	49	5,505	49	16	8	23	25	203	60

Notes: χ² tests were used to assess differences in the distribution of categorical variables by sexual identity, and ANOVA was used for age.

⁺p<0.10

^*p<0.05

^**p<0.01

^***p<0.001

n.s.= not significant.

Figure 1 presents bivariate differences in the outcomes of interest by sexual identity, based on χ² tests of association. Among heterosexual respondents who anticipated future support, children or grandchildren were the most commonly expected source of help, while non-kin were the least common. The proportion of lesbian/ gay respondents who expected help from children or grandchildren was almost 60% that of heterosexual respondents. However, lesbian/ gay individuals were significantly more likely than heterosexuals to anticipate support from other relatives or from someone else apart from kin. Finally, the pattern of anticipated sources of support among bisexual respondents and those identifying as something else/other was somewhat similar to that of heterosexual respondents, with children/grandchildren and relatives being the main expected sources of help.

Fig. 1.

Percentage of respomdents anticipating help with future ADLs by sexual identity, overall and from specific sources of support. Note: χ2 tests are for bivariate association between sexual identity and each outcome (***p<0.001). Values shown in bars are rounded unweighted percentages

Table 2 Panel A provides the incidence risk ratios (IRRs) for the modified Poisson regression results while Panel B provides pairwise comparisons of the predicted probabilities of the outcomes across sexual identity groups. In Model 1, the rate at which lesbian/ gay older adults perceive the availability of future help with ADLs was 12% higher compared to straight counterparts. Compared to heterosexual respondents, those identifying as bisexual and something else/other had 14–15% lower rates of anticipating support for future functional limitations. In terms of predicted probabilities (Panel B), lesbian and gay individuals did not differ significantly from heterosexual individuals. However, their probability of perceiving future ADL help being available was about 18 percentage points higher than those identifying as bisexuals and as something else or other. Those who identified as female, non-Hispanic Black, previously married, never married, and those who have least 1 child had about 5–32% higher rates of anticipating future assistance. In contrast, Hispanic and non-Hispanic other respondents, college graduates, as well as those reporting poor or fair self-rated health, had about 9–17% lower rates of perceiving the availability of future support.

Table 2: Modified Poisson Regression Results

Panel A: Incidence Risk Ratios
VARIABLES	Predicting the Source of Future ADL Help
	Model 1: Future ADL help available	Model 2: From Children/ grandchildren	Model 3: From Relatives	Model 4: From Someone else
Sexual identity (ref: heterosexual)
Lesbian/ gay	1.119*	0.646**	0.918	1.345***
	(1.007 – 1.243)	(0.482 – 0.865)	(0.795 – 1.061)	(1.148 – 1.576)
Bisexual	0.854+	0.902	1.072	1.334+
	(0.715 – 1.020)	(0.733 – 1.109)	(0.842 – 1.364)	(0.981 – 1.815)
Something else/ other	0.850**	0.958	1.090	1.076
	(0.770 – 0.939)	(0.873 – 1.051)	(0.930 – 1.278)	(0.858 – 1.350)
Age	0.996***	1.007***	0.978***	0.989***
	(0.995 – 0.998)	(1.006 – 1.008)	(0.975 – 0.981)	(0.986 – 0.993)
Female (ref: male)	1.090***	1.109***	1.030	1.095*
	(1.062 – 1.120)	(1.078 – 1.141)	(0.980 – 1.083)	(1.022 – 1.174)
Race-ethnicity (ref: non-Hispanic white)
Non-Hispanic Black	1.127***	0.961*	1.436***	1.113**
	(1.096 – 1.159)	(0.931 – 0.992)	(1.357 – 1.518)	(1.031 – 1.201)
Hispanic	0.909***	0.982	1.182***	0.855**
	(0.872 – 0.947)	(0.942 – 1.023)	(1.097 – 1.274)	(0.762 – 0.958)
non-Hispanic others	0.826***	0.983	1.157*	1.031
	(0.766 – 0.891)	(0.915 – 1.057)	(1.026 – 1.306)	(0.880 – 1.209)
Education (ref: high school or less)
Some college	1.000	1.034**	0.994	1.199***
	(0.971 – 1.029)	(1.003 – 1.065)	(0.937 – 1.054)	(1.103 – 1.303)
College	0.913***	1.018	0.969	1.353***
	(0.884 – 0.943)	(0.985 – 1.052)	(0.911 – 1.030)	(1.245 – 1.470)
Marital status (ref: married/ partnered)
Previously married	1.059***	1.024+	0.910**	1.213***
	(1.030 – 1.089)	(0.996 – 1.054)	(0.859 – 0.964)	(1.126 – 1.306)
Never married	1.053+	1.078*	0.889**	1.026
	(0.997 – 1.114)	(1.005 – 1.156)	(0.819 – 0.964)	(0.909 – 1.158)
Poor/ fair health (ref: excellent, very good, good)	0.860***	0.996	0.928*	0.828***
	(0.832 – 0.890)	(0.965 – 1.029)	(0.871 – 0.988)	(0.755 – 0.907)
Number of children
0
1	1.134***		0.763***	0.770***
	(1.062 – 1.212)		(0.704 – 0.825)	(0.687 – 0.864)
2	1.238***	1.255***	0.607***	0.612***
	(1.166 – 1.314)	(1.180 – 1.334)	(0.564 – 0.654)	(0.551 – 0.681)
3+	1.320***	1.426***	0.481***	0.534***
	(1.245 – 1.399)	(1.346 – 1.510)	(0.447 – 0.517)	(0.481 – 0.593)
Constant	0.688***	0.329***	2.928***	0.776+
	(0.616 – 0.769)	(0.293 – 0.369)	(2.379 – 3.603)	(0.590 – 1.019)
Observations	11,828	7,280	7,951	7,951
Panel B: Comparing Predicted Probabilities of the Outcome by Sexual Identity
Sexual Identity	Model 1: Future ADL help available	Model 2: From Children/ grandchildren	Model 3: From Relatives	Model 4: From Someone else
Heterosexual	0.675 b, c	0.762 b	0.431 a	0.302 a
Lesbian/ Gay	0.755 c	0.492 a	0.396 a	0.406 b
Bisexual	0.576 a, b	0.687 a, b	0.462 a	0.403 a, b
Something else/ other	0.573 a	0.730 b	0.470 a	0.325 a, b

Note for Panel A: 95% confidence intervals in parentheses.

⁺p<0.10

^*p<0.05

^**p<0.01

^***p<0.001.

Note for Panel B: Margins sharing a letter are not significantly different at the 5% level based on Bonferroni’s method for conducting multiple comparisons.

Models 2–4 predict the source of future assistance, with Model 2 limiting the sample to those who have at least one child. Among older adults with at least one child, lesbian/gay respondents anticipated support from children or grandchildren at about 40% the rate of their heterosexual counterparts, whereas bisexual and something else/other respondents did not differ significantly from heterosexuals. These sexual identity differences remained statistically significant when pairwise comparisons were conducted. Specifically, lesbian/gay respondents’ predicted probability of counting on descendants was almost 30 percentage points lower than heterosexual respondents and those reporting as something else or other. Anticipated support from descendants was modestly higher (2–11%) among those identifying as female, those with some college education, and those previously or never married whereas having 2 or more children significantly increased the rate of counting on their children/ grandchildren for future support (26–43%). Being non-Hispanic Black was associated with about 4% lower rate of perceiving that children would be available for future support compared to non-Hispanic white individuals.

Model 3 predicts the perceived availability of relatives to support when future functional limitations arise. Lesbian/gay, bisexual, and individuals identifying as something else/ other reported rates of anticipating support from relatives that were not statistically different from heterosexuals. This absence of sexual identity differences held for pairwise comparisons. The rate was about 2–9% lower for respondents who were older, previously married, and had poor/ fair self-rated health. Notably, having more children was associated with a significantly lower rate of relying on relatives for future ADL assistance – the rate decreased progressively with greater parity (IRR=0.48 to 0.76).

Model 4 demonstrates that lesbians/ gay and bisexual individuals had about 33–35% higher rate of counting on someone else other than children/ grandchildren and relatives compared to straight older adults. Only the difference between lesbian/ gay and heterosexual respondents held when pairwise comparisons of predicted probabilities (0.406 vs. 0.302, respectively) were considered. Although being older modestly lowered the rate of perceiving future help from non-kin, being Hispanic and reporting poor/ fair health reduced the rate by around 15–17%. Notably, having more children was associated with a significantly lower likelihood of relying on someone else other than kin for future ADL assistance – the rate decreased progressively as the number of children increased (IRR=0.53 to 0.77). In contrast, being female, having more than a high school education, being previously married, and identifying as Hispanic, were positively associated with perceiving non-relatives as a source of anticipated ADL support (IRR ranges from 1.09 to 1.36).

The results were generally robust across an alternative data source and most model specifications (Table 3). When limiting the analyses to new respondents in 2016—those to whom the HRS first administered the sexual identity question (Panel A)—lesbian/gay respondents continued to show a significantly lower (higher) predicted probability of anticipating support from descendants (someone else) compared to heterosexuals. Lesbian and gay individuals continued to differ from bisexual respondents in their predicted probability of relying on non-kin for support, largely because bisexual older adults in the HRS tended to exhibit patterns of anticipated support similar to those of heterosexuals. When lesbian/ gay, and bisexual identity were aggregated as sexual minority identity– either including (Panel B) or excluding the ‘something else/other’ category (Panel C), sexual minority respondents continued to exhibit a disadvantage in relying on descendants, but an advantage in being able to rely on non-kin compared with their heterosexual counterparts. However, the differences between sexual majority vs. sexual minority older adults narrowed compared to the main analyses, probably due to those identifying as bisexual and something else/other generally following the pattern of anticipatory support of straight individuals. In sum, the pattern in support sources remained unchanged: sexual minority respondents continued to anticipate relying more on individuals outside their immediate family (i.e., beyond children, grandchildren, or other relatives). When sexual identity is coded to include the don’t know, refused, and missing responses of the sexual identity variable alone (Panel D), the lesbian/gay disadvantage relative to straight older adults in terms of relying on descendants remained consistent. However, the disparities in relying on someone else other than kin disappeared.

Table 3:

Sensitivity Analyses; Modified Poisson Regression Results in Predicted Probabilities

Panel A: Using Data on 2016 New Respondents
KEY PREDICTOR VARIABLE	Model 5: Future ADL help available	6: From Children/ grandchildren	7: From Relatives	8: From Someone else
Heterosexual	0.685 a	0.633 b	0.562 a	0.338 a
Lesbian/ gay	0.805 a	0.346 a	0.545 a	0.511 b
Bisexual	0.616 a	0.583 a, b	0.424 a	0.232 a
Something else/ other	0.620 a	0.599 a, b	0.601 a	0.427 a, b
Observations	3,173 a	1,887	2,179	2,179
Panel B: Using Dichotomized Sexual Identity
KEY PREDICTOR VARIABLE	9: Future ADL help available	10: From Children/ grandchildren	11: From Relatives	12: From Someone else
Heterosexual	0.675	0.762	0.431 a	0.302
Sexual Minority (lesbian/ gay,	0.627	0.680	0.436 a	0.374
bisexual, something else/ other)
Observations	11,828	7,280	7,951	7,951
Panel C: Using Dichotomized Sexual Identity and Excluding Individuals Identifying as ‘Something else/ other’
KEY PREDICTOR VARIABLE	13: Future ADL help available	14: From Children/ grandchildren	15: From Relatives	16: From Someone else
Heterosexual	0.676 a	0.762	0.431 a	0.303
Sexual Minority (lesbian/ gay, bisexual)	0.692 a	0.585	0.411 a	0.409
Observations	11,488	7,111	7,768	7,768
Panel D: Using Sexual Identity that Includes ‘Don’t know,’ ‘Refused,’ and Missing Responses
KEY PREDICTOR VARIABLE	17: Future ADL help available	18: From Children/ grandchildren	19: From Relatives	20: From Someone else
Heterosexual	0.673 b, c	0.762 a	0.430 a	0.300 a
Lesbian/ gay	0.752 c	0.492 b	0.394 a	0.405 a
Bisexual	0.573 a	0.730 a, b	0.467 a	0.322 a
Something else/ other	0.510 a, b	0.814 b	0.409 a	0.213 a
Don’t Know	0.673 b, c	0.762 a	0.430 a	0.300 a
Refused	0.575 a, b, c	0.688 a, b	0.460 a	0.402 a
Missing	0.527 a	0.739 a, b	0.448 a	0.313 a
Observations	12,154	7,432	8,113	8,113

Note: All models controlled for age, gender, race-ethnicity, educational attainment, marital status, self-rated health, and number of children. Incident risk ratios (IRRs) from the full regression models shown in Tables S3 to S5.

Discussion and Conclusions

This study draws on the minority stress model and the convoy model of support to examine sexual identity disparities in anticipated support for future activities of daily living from children or grandchildren, other relatives, and non-kin, excluding spouses and partners, among older adults in the United States. Anticipatory support – the belief that someone will be available to provide care if needed – acts as a psychological safety net that enhances emotional well-being, mastery, and sense of self. Because anticipatory support is highly predictive of informal care received, which is a cornerstone of eldercare arrangements, understanding disparities in its availability is essential for advancing health equity.

Using the data from the 2022 wave of the Health and Retirement Study and controlling for personal and situational factors, I found that notable disparities in anticipated support by sexual identities exist. Main results show that lesbian/ gay respondents’ rate of perceiving available support was comparable to heterosexuals, undermining H1. Even though the sexual minority community continually faces stigma and discrimination, findings from this study suggest that some sexual minority older adults – specifically lesbian/gay ones – perceive that they have access to future support from their network excluding their partner/ spouse. However, in partial support of H1, those identifying as something else/other had 10 percentage points lower rates of anticipating support compared to straight-identifying individuals. Note however, that this disparity did not hold when a smaller dataset was used (i.e., new HRS respondents in 2016, Table 3, Panel A), so readers should interpret this finding with caution.

In terms of within-sexual-minority differences, those identifying bisexual and as something else/other had about 18 percentage points lower predicted probability of perceiving that future ADL help is available compared to lesbian/gay older adults. Several mechanisms may help explain the observed disadvantage among those identifying bisexual and as something else/ other in anticipated support. Bisexual individuals are often stigmatized as sexually promiscuous and incapable of forming committed relationships by both heterosexual and sexual and gender minority communities (e.g., Liu et al., 2021). Such stigma can limit opportunities for sustained social connection. Bisexual older adults are more likely to have predominantly heterosexual social networks (e.g., Grossman et al., 2000), which may include individuals who hold negative attitudes toward bisexuality. For these reasons, bisexual older adults may be less likely to identify others they can rely on for future support. While we do not truly know the sexual identities of those identifying as something else/ other from the HRS, we do know from the descriptive statistics (Table 1) that they are more likely to report being female, Hispanic, have the lowest educational attainment, were previously married, and have the highest number of children. These demographic characteristics may suggest that this group experiences heightened social or economic disadvantage, which, in turn, may translate to disadvantages in anticipatory support (Antonucci et al., 2011; Kahn & Antonucci, 1980). Future research should increase representation of those identifying as bisexual and as something else within sexual minority samples to clarify within-group disparities in anticipated old-age support and related outcomes.

Sources of anticipatory support generally varied notably by sexual identity, with the exception that no differences were observed in anticipating support from relatives, contrary to H3. These findings are robust across different datasets and ways of coding sexual identity. Among those with children, the probability of lesbian/gay adults anticipating support from children or grandchildren was about 30 percentage points lower than the rate of their heterosexual counterparts in support of H2. However, those reporting as bisexual and as something else/other did not differ significantly from heterosexuals, undermining H2. Several mechanisms could help explain this disparity. First, even among those with children, sexual minority individuals tend to have fewer children on average than heterosexuals (Hsieh & Wong, 2020), reducing the potential supply of adult-child caregivers. Second, some adult children may have estranged their parent due to their parent’s sexual identity. These older adult respondents came of age in a sociohistorical context in which same-sex relationships were heavily stigmatized, homosexuality was classified as a mental disorder in the DSM, and many entered so-called “lavender marriages” to conceal their sexual orientation (Flatt et al., 2022; Gates, 2015). Dissolving such marriages may have led to strained parent–child relationships, increasing the risk of estrangement later in life. To date, little research has examined parent–child relationships in later life in relation to parental marital history. However, one qualitative study from Australia found that children of sexual minority parents in former lavender marriages attributed family conflict primarily to the ex-spouse, rather than to the parent–child relationship itself (Titlestad & Pooley, 2014). Third, as mentioned, bisexual older adults are more likely to have predominantly heterosexual social networks (e.g., Grossman et al., 2000), which may mean that they either adopt social patterns similar to their heterosexual peers or are better able to conceal their sexual identity, thereby maintaining access to support from descendants that may be less available to lesbian and gay individuals.

Finally, lesbian/gay individuals were significantly more likely to anticipate receiving help from non-kin or “someone else” outside of a spouse or partner, descendants, or other relatives in support of H4. Due to the way the HRS phrased this question, it is unclear who exactly these individuals are. However, they may be understood as part of the chosen family (Breder & Bockting, 2025). Structural barriers that historically excluded persons who are sexual minorities from marriage and childbearing, along with stigma and discrimination that strained ties with families of origin, have compelled many to form alternative support networks (Gates, 2015; Hsieh & Wong, 2020; Westrick-Payne & Manning, 2025). Forming these chosen families serves as a resilience strategy for coping with minority stress and plays a vital role in filling the social support gaps created by structural and interpersonal exclusion (Breder & Bockting, 2025).

Even though this current study is probably the first to document sexual identity disparities in anticipated support for future ADL limitations, it is not without shortcomings. First, the analysis relies on cross-sectional data with a relatively small sample size of sexual minority respondents. As such, the findings should not be interpreted as causal or fully generalizable. To increase statistical power and improve representation of sexual minority older adults, I backfilled responses to the sexual identity question from past HRS waves. This approach assumes that sexual identity is static over time, despite its fluidity (Veldhuis et al., 2024). Importantly, the overall patterns of association remained consistent even when the analyses were restricted to a significantly smaller sample without backfilled data (i.e., the new respondents in 2016 wave, Table 3, Panel A). Second, while it would have been ideal to explicitly test for the association by intersecting identities (Veldhuis et al., 2024), the small n of new HRS respondents identifying as sexual minority rendered intersectional analyses infeasible. Additionally, the way sexual identity was measured in the HRS limited distinctions to lesbian, gay, bisexual, and something else/ other, preventing the inclusion of more diverse identities such as transgender, two-spirit, asexual, intersex individuals, and others. Nevertheless, the inclusion of sexual identity in the HRS provides good vantage point, especially considering that older persons identifying as sexual minority were largely invisible in the gerontological literature (Brown, 2009).

Future research is needed to understand individuals’ who identify as sexual minority’ preferred, planned, and actual long term care arrangements and barriers in accessing both formal and informal care (Flatt et al., 2022). Additionally, greater attention is needed to understand the caregiving experiences of LGBT older adults, who – due to their generally homogenous and shrinking social networks – may face unique challenges as they navigate caregiving roles while also managing their own aging process and care needs (Breder & Bockting, 2025; Hsieh & Wong, 2020; Singleton & Enguidanos, 2022).