Association of Self-Reported Sexual Orientation Fluidity with Health and Behavior of Adults: Analysis of the Research Program on Genes,Environment,and Health Cohort

Abstract

Purpose:

Health disparities by sexual orientation fluidity are relatively unexplored in middle or late adulthood. We assessed differences in self-reported health and health behaviors according to fluidity status.

Methods:

We analyzed baseline (2002–2010) and follow-up (2015–2023) survey data of Research Program on Genes, Environment, and Health participants. We classified people based on responses to sexual orientation questions: consistently heterosexual, consistently sexual minority, and fluid (changed reported orientation). We then compared health behavior (smoking, binge drinking, and physical activity) and self-rated health (Patient-Reported Outcomes Measurement Information System and EuroQol EQ-5D-3L) across groups using linear or logistic regression with or without weights for attrition using complete case data and after multiple imputation.

Results:

Of the 32,058 adults (mean age: 56 years, male: 39.6%), 378 (1.2%) were fluid. Compared with consistently heterosexual adults, fluid individuals had significantly worse self-rated health scores after adjustment for demographics and attrition, although most differences were not clinically meaningful. No differences in health behavior were noted in adjusted analyses.

Conclusion:

Sexual orientation fluidity occurs in older adults, and this group may have worse health than their heterosexual counterparts.

Introduction

A growing literature has documented sexual orientation fluidity across the life course.^1–3 Shifting away from the conceptualization of sexual orientation as a stable trait, sexual fluidity acknowledges that some people experience changes in sexual identity, attraction, or behavior over time that are different from orientation concealment or disclosure (e.g., coming out of the closet).^4–6 This change occurs in any direction.^6–9 Initially believed to be limited to adolescence and young adulthood, studies have now demonstrated that people of older ages also exhibit fluidity.^2,7,8,10

Health researchers have become interested in examining the role of fluidity in health.⁵ While fluid identity should not inherently affect health, fluid individuals can experience stress that adversely affects health.^11,12 Some derive from stressors shared with nonfluid sexual minority (SM) individuals, such as internalized stigma or overt bi/homophobic encounters.¹³ There are also unique stressors like stress from identity change and management.^14,15 Society “forecloses fluidity,” leading to dismissive interactions or pressures to comply with binary norms.^12,16

Most research has focused on mental health. A review reported that fluidity is associated with adverse depression and substance use.⁵ A recent analysis found that sexually fluid adults have worse mental health, financial insecurity, and substance use outcomes compared to consistently straight individuals.⁷ However, this association may not be uniform, and the link between fluidity and mental health is likely complex and path-dependent.^12,14 A study found that those who changed to a plurisexual (e.g., bisexual) orientation had significantly higher distress scores than consistently heterosexual individuals.⁸ Meanwhile, those who changed to a heterosexual orientation had significantly lower distress scores than consistently plurisexual individuals. Although minority stress has been shown to affect physical health,¹⁷ this is less studied in fluidity. One study on adolescents showed path-dependent results: only men moving from homo- to heterosexual identity had lower physical activity and increased body mass index (BMI).¹⁸

There are knowledge gaps on the health of fluid individuals, especially those in middle or late adulthood. Addressing gaps can inform how to support fluid individuals in maintaining health. We hypothesize that individuals with evidence of fluid sexual identity have worse health outcomes and have a higher prevalence of health-harming behavior than people who have stable sexual identities.

Methods

Data source and population

We conducted a retrospective analysis by linking Research Program on Genes, Environment, and Health (RPGEH)¹⁹ survey data (baseline, 2002–2010) with a follow-up survey of the same participants through the Kaiser Permanente Research Bank (KPRB) intake survey (2015–2023).²⁰ The RPGEH cohort is a dataset of individuals recruited from Kaiser Permanente Northern California members in 2007 via a mailed survey. The KPRB is an expansion of the RPGEH effort to cover all the regions where Kaiser Permanente operates, that was started in 2015. All participants provided informed consent prior to participation in the survey and consented to secondary use of their data for research. This work is part of a project approved by the Kaiser Permanente—Mid-Atlantic States Region Institutional Review Board.

Fluidity

We focused on the fluidity of sexual identity assessed by the two surveys. The baseline (RPGEH) survey asked for sexual orientation with the following options: heterosexual/straight, bisexual, homosexual (gay/lesbian), other. The follow-up survey asked: “You consider yourself to be”: with the following options: heterosexual or straight, gay, lesbian, bisexual, other (specify), do not know, prefer not to answer. We then classified individuals into: “consistently heterosexual,” “consistently SM,” and fluid based on self-reported sexuality. The fluid (or “changed reported sexual orientation”) group is those who reported a change in orientation, which includes people who changed responses from heterosexual to SM or from SM to heterosexual. Those who changed identities but remained SM (e.g., gay to bisexual) were classified as consistently SM.

Those who reported “other” sexual orientation at baseline were classified as missing orientation. This study could not access the free-text responses in the baseline survey, and thus could not assign a category for individuals who refused to answer the sexual orientation question. “Other” responses in the follow-up survey were recoded by the first author (ASR) based on responses to the follow-up free-text question. We focused on the three-group comparison as the main analysis due to small group sizes if we subdivided the fluid group based on direction of change.

Outcomes

The outcomes were self-rated health (SRH) measures and health behaviors measured in follow-up. SRH used PROMIS® (Patient-Reported Outcomes Measurement Information System) physical and mental health T-scores²¹ and EuroQol EQ-5D-3L score.²² The physical and mental health T-scores were calculated from the responses to the Global Health Scale.²³ Physical and mental health sum scores were converted to T-scores using the HealthMeasures scoring service, producing T-scores with a mean of 50 and a standard deviation of 10. For example, a score of 40 means a score of one standard deviation from the reference population.

Physical health scores were based on items 3, 6, 7, and 8. The questions relate to general physical health (Item 3), carrying out everyday physical activities (Item 6), pain (Item 7), and fatigue (Item 8). Mental health scores were based on items 2, 4, 5, and 10. The questions relate to quality of life (Item 2), general mental health (Item 4), social relationships (Item 5), and emotional problems (Item 10).

The EQ-5D-3L score captures quality of life related to five domains: mobility, self-care, usual activity, pain/discomfort, and anxiety/depression. Usually, responses to the EQ-5D-3L instrument are converted to a single score based on population preferences.²² In the RPGEH dataset, the scores were calculated from items 2, 3, 4, 6, 7, 8, 9, and 10 of the PROMIS® Global Health Scale.²² Item 9 is related to usual social activities. The other items were reported in prior paragraphs.

Health behaviors include (1) current smoking, (2) current binge drinking, and (3) adequate physical activity. Current smoking was assessed by asking, “Do you currently smoke cigarettes every day, some days, or not at all? [every day, some days, not at all, not applicable].” Current smokers are those who reported smoking every day or on some days. Binge drinking was assessed by asking, “During the past year, how often did you have six or more drinks at one occasion? [never, less than once a month, monthly, weekly, daily/almost daily, not applicable]).” Current binge drinkers were those who reported drinking 6 or more drinks at least once a month.

Adequate physical activity was defined as reporting at least 3 days of moderate physical activity with ≥30 minutes per day or at least 3 days of vigorous activity with ≥20 minutes per day. Moderate activities are exercise, sports, or other physical activity that caused your heart rate to increase somewhat (other than walking), while vigorous activities are exercise, sports, or other physical activity that caused you to work up a sweat or causes your heart rate to increase greatly. The question used was: “During the past 7 days, please record the number of days that you did each of the (moderate/vigorous) activities [none, 1–2 days, 3–4 days, 5–6 days, everyday]. Also, record the average minutes per day that you did each activity on the days that you did that activity [0–9 minutes, 10–19 minutes, 20–29 minutes, 30–59 minutes, 60 minutes or more, not applicable].”

Covariates

Covariates were obtained from the baseline survey. Demographics included year of survey, baseline age (from “What is your date of birth”?), gender (“What is your gender?” [male, female]), race and ethnicity (“What best describes your race or ethnicity?” (select all that apply: Black, Hispanic, Asian, White, Pacific Islanders, Native Hawaiian or Other Pacific Islander, American Indian/Alaska Native), and nativity (“Were you born in the US?” [Yes, No]). For race and ethnicity, we only had access to the constructed variable where responses were hierarchically rolled up in this order: Black, Hispanic, Asian, White, Other. Other included Pacific Islanders, Native Hawaiian or Other Pacific Islander, and American Indian/Alaska Native. For example, if a person selects Black and Asian, they are classified as Black.

Socioeconomic indicators included: marital status (“What is your current marital status?” collapsed to [currently married, never married, divorced/separated/widowed]), employment (“What is your employment or work status?” collapsed to [full time/part-time/student/homemaker vs. unemployed/retired]), income (“What best describes your household income [before taxes]?” collapsed to [<40,000, 40,000 to <60,000, 60,000 to <1,00,000, ≥1,00,000]), educational attainment (“What is the highest level of school that you have completed?” collapsed to [elementary/some high school/high school, some college/trade/others, college/graduate school]), and household size (“How many other people live in your household (include spouse, partner, children, and other relatives)?” categorized to Living alone, Living with others). We also included baseline values of health behaviors (see outcomes section for definitions).

Gender at the baseline only allowed binary options, but the follow-up survey allowed more expansive options (male, female, intersex, male to female transgender, female to male transgender, other [specify]). Notably, gender identity questions are not reflective of current standards and recommendations.²⁴

Statistical analysis

We first calculated descriptive statistics of covariates by fluidity status. We then calculated unadjusted differences in SRH plus linear models to assess differences after adjusting for potential confounders (age, survey year, gender, race and ethnicity, and nativity) and potential mediators (remaining covariates). The mediators were selected based on a directed acyclic graph informed by ecosocial theory.²⁵ (Supplementary Fig. S1).

We extracted the coefficients from the fluidity variable as the estimate of disparity with the consistently heterosexual group as the reference group. Similarly, we calculated unadjusted and adjusted disparities in the prevalence of health behaviors using robust Poisson regression models. The third model included the baseline value for health behavior. After estimating the model, we used the avg_comparisons function from the marginaleffects package to obtain the marginal odds ratio.²⁶ These odds ratios were used as the estimate of the disparity with the consistently heterosexual as the reference group. All statistical analyses were run in R4.3.0 (see Supplementary Data for sample R code).

To address potential bias from differential loss to follow-up and missing data (exposure, covariate, outcome), we repeated the analysis with combinations of inverse propensity attrition weights and multiple imputation.²⁷ Stabilized attrition weights were calculated to account for loss between survey waves using the WeightIt package.²⁸ The attrition weighting model used the following covariates: age, gender, race and ethnicity, nativity, income, BMI, education, SM in RPGEH, year of RPGEH, and general health rating at RPGEH. Weighted regression was done using the survey package.

Multiple imputation was implemented using mice.²⁹ We imputed 10 datasets with 5 maximum iterations at default settings. Covariates of the imputation model include fluidity status, survey year, demographics (age, gender, race and ethnicity, nativity, and sexual orientation), socioeconomics (living alone, marital status, employment, educational attainment, and income), stabilized attrition weights, SRH, and health behavior (baseline and follow-up). Results from multiple imputation analysis were pooled following Rubin’s rules.³⁰ We also assessed if health behavior disparities were modified by (binary) gender, like previous work.¹⁸

We did three additional exploration analyses. First, we conducted an intra-categorical analysis of consistently SM adults to assess differences in sociodemographics and outcomes of people who report the same SM category in both surveys versus those who reported a different SM category on follow-up (e.g., bisexual to lesbian). This explored how fluidity influences health while partly accounting for minority stress from SM identity. Second, we compared outcomes by fluidity within heterosexual and SM individuals at baseline. Third, we compared outcomes by baseline and follow-up sexual identity. The last two analyses explored the path dependence of health impacts and were limited to descriptive statistics.

Results

We included 32,058 adults (baseline mean age: 56 years, male: 39.6%) with an average 14 (standard deviation [SD]: 3) years’ time interval between surveys. Among those who were heterosexual at baseline (n = 30,492), 274 (0.9%) reported a change to an SM orientation, with most (83%) reporting bisexual orientation on follow-up. Meanwhile, among SM adults (22.7% bisexual) at baseline, 104 (7%) reported a change to heterosexual orientation (Supplementary Table S1). Overall, the sample was 94.3% consistently heterosexual, 4.6% consistently SM, and 1.2% fluid. The fluid group tended to be younger at baseline than consistently heterosexual adults on average (Table 1). All age deciles showed the occurrence of orientation changes, but these proportions decreased as the age increased (from 5.3% in 30 or younger to 0.7% in 70 or older). There were also differences in educational attainment, income, and marital status.

Table 1.

Summary of Sociodemographic Characteristics by Reported Change in Sexual Orientation

	Fluid (n = 378)	Consistently heterosexual (n = 30,218)	Consistently sexual minority (n = 1462)
Age (mean [SD])	49.67 (13.21)	56.21 (10.45)	52.53 (9.71)
Age intervals, n (%)
18 to <30 years	31 (8.2)	524 (1.7)	28 (1.9)
30 to <40 years	52 (13.8)	1626 (5.4)	105 (7.2)
40 to <50 years	88 (23.3)	4880 (16.1)	378 (25.9)
50 to <60 years	114 (30.2)	10,898 (36.1)	619 (42.3)
60 to <70 years	77 (20.4)	9920 (32.8)	285 (19.5)
70 years or older	16 (4.2)	2370 (7.8)	47 (3.2)
Race and ethnicity, n (%)
Asian	28 (7.4)	2116 (7.0)	64 (4.4)
Black	7 (1.9)	767 (2.5)	34 (2.3)
Hispanic	25 (6.6)	1724 (5.7)	88 (6.0)
White	317 (84.1)	25,505 (84.7)	1272 (87.2)
Other^a	0 (0.0)	<5	0 (0.0)
Gender, n (%)
Female	236 (62.4)	18,517 (61.3)	625 (42.7)
Male	142 (37.6)	11,701 (38.7)	837 (57.3)
US born (nativity), n (%)
US born	330 (88.9)	26,431 (89.2)	1318 (91.7)
Not US born	41 (11.1)	3200 (10.8)	119 (8.3)
Sexual orientation at baseline, n (%)
Bisexual	93 (24.6)	0 (0.0)	263 (18.0)
Heterosexual	274 (72.5)	30,218 (100.0)	0 (0.0)
Gay/lesbian	11 (2.9)	0 (0.0)	1199 (82.0)
Sexual orientation on follow-up, n (%)
Bisexual	227 (60.1)	0 (0.0)	263 (18.0)
Heterosexual	104 (27.5)	30,218 (100.0)	0 (0.0)
Gay/lesbian	29 (7.7)	0 (0.0)	1180 (80.7)
Other sexual minority	18 (4.8)	0 (0.0)	19 (1.3)
Time between surveys, years, mean (SD)	14.29 (2.76)	14.40 (2.72)	14.82 (2.77)
Highest educational attainment, n (%)
Elementary to high school	15 (4.0)	1824 (6.2)	39 (2.7)
Some college, trade school, others	97 (26.1)	7598 (25.8)	263 (18.3)
Completed college or higher	259 (69.8)	19,988 (68.0)	1136 (79.0)
Household annual income, US dollars, n (%)
<40,000	45 (12.6)	2262 (7.9)	159 (11.2)
40,000 to <60,000	63 (17.6)	3702 (13.0)	179 (12.6)
60,000 to <100,000	113 (31.6)	8906 (31.2)	396 (27.9)
≥100,000	137 (38.3)	13,642 (47.8)	684 (48.2)
Employed, n (%)
Employed (full- or part-time employment/student/homemaker)	302 (79.9)	21,379 (70.7)	1164 (79.6)
Not employed (Unemployed/retired)	76 (20.1)	10,954 (29.3)	338 (20.4)
Marital status, n (%)
Currently married	234 (62.7)	22,977 (76.5)	704 (48.6)
Divorced/separated/widowed	62 (16.6)	4915 (16.4)	177 (12.2)
Never married	77 (20.6)	2146 (7.1)	568 (39.2)
Living alone, n (%)	67 (18.3)	4257 (14.4)	422 (29.5)
Baseline health,^b mean (SD)	2.39 (0.88)	2.22 (0.87)	2.21 (0.89)

Other included Pacific Islanders, Native Hawaiian or Other Pacific Islander, and American Indian/Alaska Native.

Self-rated general health from 1 = excellent to 5 = poor. Demographic variables are from the baseline survey unless stated otherwise.

SD, standard deviation; US, United States.

Self-rated health

Fluid adults had slightly lower mean SRH (physical: 49.9 [SD: 8.0], mental: 49.6 [SD: 8.6], EQ-5D-3L: 0.74 [SD: 0.1]) than consistently heterosexual (physical: 50.6 (SD: 7.9), mental: 52.6 (SD: 8.2), EQ-5D-3L: 0.75 [SD: 0.1]) and consistently SM (physical: 50.9 [SD: 7.9], mental: 51.5 [SD: 8.6], EQ-5D-3L: 0.75 [SD: 0.1]) adults. Adjusted analysis (Table 2) showed consistently SM adults had significantly lower mental health compared to consistently heterosexual adults, but this did not remain significant after adjusting for sociodemographics. Meanwhile, fluid adults had significantly lower SRH for all measures compared to consistently heterosexual adults in all unweighted models except Model 2 for physical health (Table 2).

Table 2.

Differences in Self-Rated Health Ratings between Fluid or Consistently Sexual Minority Adults and Consistently Heterosexual Adults with Complete Case Analysis

	Unweighted for attrition				Weighted for attrition
	Fluid		Consistently SM		Fluid		Consistently SM
	Difference (95% CI)	p Value	Difference (95% CI)	p Value	Difference (95% CI)	p Value	Difference (95% CI)	p Value
Physical health
0	−0.97 (−1.82 to −0.12)^a	0.025	0.24 (−0.19 to 0.68)	0.274	−0.64 (−1.58 to 0.29)	0.177	0.45 (−0.05 to 0.95)	0.076
1	−1.6 (−2.57 to −0.63)^a	0.001	−0.98 (−2.93 to 0.96)	0.32	−1.48 (−2.56 to −0.4)^a	0.007	−0.99 (−3.18 to 1.19)	0.373
2	−0.63 (−1.46 to 0.21)	0.142	−0.98 (−2.65 to 0.69)	0.248	−0.62 (−1.54 to 0.29)	0.182	−0.98 (−2.91 to 0.94)	0.317
Mental health
0	−3.16 (−4.05 to −2.28)^a	<0.001	−1.02 (−1.48 to −0.57)^a	<0.001	−3.01 (−4 to −2.03)^a	<0.001	−0.99 (−1.52 to −0.46)^a	<0.001
1	−2.87 (−3.9 to −1.84)^a	<0.001	−1.65 (−3.7 to 0.41)	0.116	−2.72 (−3.83 to −1.61)^a	<0.001	−1.59 (−3.86 to 0.68)	0.171
2	−1.95 (−2.89 to −1.00)^a	<0.001	−1.4 (−3.28 to 0.47)	0.143	−1.86 (−2.87 to −0.85)^a	<0.001	−1.19 (−3.32 to 0.95)	0.276
EQ-5D-3L
0	−0.02 (−0.03 to −0.01)^a	<0.001	0 (−0.01 to 0.002)	0.213	−0.01 (−0.02 to −0.005)^a	0.004	0 (−0.01 to 0.005)	0.839
1	−0.02 (−0.03 to −0.01)^a	<0.001	−0.02 (−0.04 to −0.003)	0.106	−0.02 (−0.03 to −0.01)^a	<0.001	−0.02 (−0.04 to 0.01)	0.198
2	−0.01 (−0.02 to −0.005)^a	0.003	−0.02 (−0.03 to 0.002)	0.074	−0.01 (−0.02 to −0.003)^a	0.015	−0.01 (−0.04 to 0.01)	0.182

Statistically significant at alpha = 0.05. Models are increasingly adjusted: 0 has no other covariates, 1 adjusts for age, year of survey, gender, race and ethnicity, nativity, 2 additionally adjusts for marital status, living alone, employment, income, and education. Weighting for attrition was based on age, gender, race and ethnicity, nativity, income, body mass index, education, baseline sexual minority status, year of baseline survey, and general health rating at baseline.

CI, confidence interval; SM, sexual minority.

Sensitivity analyses with multiply imputed data had similar results, except that differences in physical health between fluid and consistently heterosexual adults were not significant in the unadjusted unweighted model for physical health and the weighted model 2 for the EQ-5D-3L (Supplementary Table S2).

Health behaviors

Fluid individuals had significantly higher unadjusted odds of current smoking on follow-up than consistently heterosexual and consistently SM individuals (Table 3). In adjusted analyses, however, there were no differences in smoking odds after covariate adjustment. The odds of binge drinking and meeting physical activity recommendations were comparable across three groups in unadjusted and adjusted models. Sensitivity analyses showed similar results (Supplementary Table S3).

Table 3.

Odds Ratio of Health Behaviors Comparing Fluid or Consistently Sexual Minority Adults to Consistently Heterosexual Adults with Complete Case Analysis

	Unweighted for attrition				Weighted for attrition
	Fluid		Consistently SM		Fluid		Consistently SM
	Odds ratio (95% CI)	p Value	Odds ratio (95% CI)	p Value	Odds ratio (95% CI)	p Value	Odds ratio (95% CI)	p Value
Current smoking
0	1.85 (1.06–3.24)^a	0.031	1.58 (1.16–2.17)^a	0.004	1.89 (0.995–3.6)	0.052	1.62 (1.15–2.27)^a	0.006
1	0.98 (0.44–2.2)	0.961	0.42 (0.13–1.32)	0.135	0.89 (0.38–2.1)	0.796	0.33 (0.09–1.22)	0.098
2	0.91 (0.44–1.9)	0.803	0.38 (0.12–1.21)	0.101	0.87 (0.4–1.88)	0.716	0.37 (0.11–1.28)	0.117
Binge drinking
0	1.27 (0.99–1.62)	0.058	1.18 (1.03–1.34)	0.014	1.23 (0.94–1.61)	0.136	1.14 (0.99–1.31)	0.079
1	1.27 (0.95–1.69)	0.103	1.08 (0.62–1.88)	0.776	1.19 (0.87–1.65)	0.279	0.96 (0.51–1.79)	0.895
2	1.27 (0.96–1.7)	0.099	1.09 (0.63–1.9)	0.753	1.2 (0.87–1.65)	0.276	0.96 (0.52–1.79)	0.899
Adequate physical activity
0	1.1 (0.83–1.45)	0.507	1.05 (0.9–1.21)	0.544	1.17 (0.87–1.57)	0.302	1.07 (0.91–1.25)	0.399
1	0.91 (0.67–1.25)	0.574	1.03 (0.54–1.98)	0.928	0.98 (0.7–1.36)	0.882	0.93 (0.45–1.96)	0.858
2	0.97 (0.71–1.31)	0.823	1.15 (0.61–2.15)	0.67	1.05 (0.76–1.45)	0.77	1.11 (0.52–2.34)	0.790

Statistically significant at alpha = 0.05. Models are increasingly adjusted: 0 has no other covariates, 1 adjusts for age, year of survey, gender, race and ethnicity, nativity, 2 additionally adjusts for marital status, living alone, employment, income, education, and baseline behavior. Weighting for attrition was based on age, gender, race and ethnicity, nativity, income, body mass index, education, baseline sexual minority status, year of baseline survey, and general health rating at baseline.

Exploratory analysis stratified by gender (Supplementary Table S4) also showed disparities in smoking. Among female participants, consistently SM but not fluid individuals had higher odds of smoking than consistently heterosexual females. Meanwhile, fluid males but not consistently SM males had higher odds of smoking than consistently heterosexual males. However, these differences were not significant after covariate adjustment.

Exploratory analyses

In the exploratory analysis of the consistent SM group (n = 1462), there were 99 (6.8%) who changed their reported SM category (Supplementary Table S1). The most common changes were gay/lesbian to bisexual (47), bisexual to gay/lesbian (33), and bisexual to other (14). People who reported the same SM category (n = 1363) had higher mean age and had higher proportions of male, never married, and gay/lesbian at baseline compared to those who reported a change in SM orientation (n = 99) (Supplementary Table S5).

While the means of the SRH scores were lower (Supplementary Table S6), there were no significant differences in outcomes between those who reported the same SM category on both surveys versus those who reported a different SM category (physical: 50.9 [SD: 7.9] vs. 50.0 [8.9], mental: 51.6 [8.5] vs. 50.6 [9.2], EQ-5D-3L: 0.75 [0.08] vs. 0.74 [0.09]). Those who changed the SM category had worse prevalence of behavioral outcomes (smoking: 3% vs. 5%, binge drinking: 24% vs. 28%, physical activity met: 21% vs. 12%), but these differences were not statistically significant. Complete case analysis, adjusting for demographics, also did not show significant disparities for SRH and behavioral outcomes.

In another exploratory analysis, we found that compared to consistently heterosexual people, those who changed to SM had significantly lower mental health and EQ-5D-3L scores even after weighting for attrition, and in unweighted analysis by gender (Supplementary Table S7). Consistently, SM female individuals had significantly higher mental health scores than those who changed from SM to heterosexual. In heterosexual males at baseline, heterosexual to SM males had significantly lower physical health scores than consistently heterosexual males. In SM females at baseline, SM to heterosexual females had significantly higher smoking prevalence. No other significant differences were noted.

We also conducted additional analysis looking at outcomes by baseline and follow-up sexual orientation (e.g., bisexual to heterosexual) (Supplementary Tables S8 and S9). There were no significant differences in outcomes across fluidity groups among bisexual and gay/lesbian individuals at baseline. In the heterosexual at baseline group, we found that consistently heterosexual adults had higher mental health scores than those who changed to any SM orientation.

Discussion

We provide additional evidence on the occurrence of sexual identity fluidity in a large cohort of insured individuals, mostly in middle or late adulthood, with relatively high socioeconomic positions. Similar to prior work, we found that the most common changes were from bisexual to heterosexual or heterosexual to bisexual, changes were more likely to occur in younger participants (30 or younger), and fluid individuals had worse mental health.^7,11,15

As new contributions, we found statistically significant but not meaningful differences in physical health and EQ-5D-3L scores (PROMIS T-scores: 2–6 points,³¹ EQ-5D-5L: 0.04–0.07³²). The disparities can be from minority stress shared with other SM individuals or those unique to fluid individuals. The EQ-5D-3L disparities may be related to mental health, as it includes anxiety and depression in its score. Physical health disparities could arise from mental health issues manifesting into physical limitations or could be unexplored direct embodiment pathways (e.g., stress causing inflammation).¹⁷ For behaviors, we found no significant differences across groups after adjustment. These null findings contrast with studies that often show smoking and drinking disparities between SM and heterosexual populations.^33,34 The older ages of this group could be a factor as health-harming behaviors tend to decline with age.^35,36

With our current analysis, it is difficult to parse sources of stress. Our result of no difference in mental health between consistently SM adults and consistently heterosexual adults suggests that fluid individuals experienced worse stressors than consistently SM individuals. Prior work showed that fluid individuals can experience stress from identity management, and that stress occurrence is path-dependent.^10,14,15 For example, in women, stress increases when moving away from heterosexual identity but lowers when moving toward it.¹⁰ We saw a similar trend in our exploratory analyses (Supplementary Tables S7, S8 and S9). However, the null finding in the SM at baseline group could also be a power issue. Path dependence also implies that baseline orientation matters, as there are likely baseline disparities by sexual orientation.^8,37 We are unable to explore this fully in this work.

Limitations

Our operationalization of fluidity used two measurements of sexual orientation with changes in the question stem and options. At best, we capture only fluid sexual identity. Fluid attraction and behavior likely have different impacts on health and need to be studied.⁵ For example, fluid sexual behavior needs to be considered when providing sexual health advice. We may have misclassified people who changed their sexual identity and returned to their baseline identity within the intervening years. Our operationalization makes it difficult to differentiate between fluidity and willingness to divulge information. Some people who initially reported being heterosexual and then being SM on follow-up might have always identified as SM but only decided to disclose this on follow-up. We cannot tease out individuals who changed from SM to heterosexual due to re-concealing their SM status. Being in the closet (concealing orientation) causes minority stress, leading to adverse mental health and the disparities we observed.³⁸

Aside from measurement issues, this work has other limitations. First, the nonrepresentative sampling limits generalizability, especially to other generations and other geocultural contexts. Secondly, questions on orientation changed between surveys, which may have affected fluidity measurement, especially for those changing within SM categories. Third, we are unable to adjust for certain covariates, like comorbidities that could drive disparities.

Fourth, despite the large overall sample, subgroup sizes were small and likely affected the power to detect changes in behavioral prevalence. The small groups and limited demographic/socioeconomic diversity also hindered examining fluidity within other categories, such as race or income. From intersectionality, we expect fluidity’s occurrence and impact on health would be modified by other social identities/positions.^39–41 Also, the multiple marginalizations concept would suggest that fluid individuals with other marginalized identities (e.g., Hispanic transgender woman) would fare worse than fluid individuals with historically privileged groups (e.g., White cisgender man).^39,42 However, intersectionality also tells us that oppressions (and/or strengths) are not necessarily just sums or products of multiple identities.^41,43 For example, a study showed generally that multiply marginalized individuals have higher suicidal ideation, but certain groups go against the expected trends.⁴⁴ They show that White bisexual women in large and small metropolitan counties had higher suicidal ideation than Black bisexual women. Empirical studies looking at fluidity and its intersections are thus necessary. Fifth, while we accounted for attrition, the model only adjusted for selected observed covariates. Finally, we focused on SRH, which is an integrative health measure, but does not point to disparities in specific conditions or functional limitations.⁴⁵

Additional studies, including qualitative ones, that comprehensively measure fluidity (identity, attraction, and behavior) over time and parse out disclosure decisions are needed.^5,46 Our study demonstrates that studying fluidity, especially path dependence, likely requires large samples. The current political climate will be challenging, but taking advantage of administrative sources that record sexual orientations could help. However, these sources require augmentation with surveys that comprehensively measure sexual identity, attraction, and behavior. Future work that pays attention to fluidity through an intersectional lens is also needed.

Conclusions

In this study, we provided additional evidence that sexual fluidity occurs in adults and that this population is likely to experience mental health disparities. Future studies on different fluidity types and mechanisms, and how to develop fluidity-inclusive health systems, are warranted.

Footnotes

Authors’ Contributions

All authors contributed significantly to the development of the article. Detailed contributions are as follows: A.S.R.—Conceptualization, methodology, formal analysis, data curation, and writing—original draft. C.R.C.—Conceptualization, methodology, and writing—review and editing. R.C.H.—Conceptualization, methodology, writing—review and editing, and funding acquisition.

Data Availability

Full details about the KPRB are available at their website: https://researchbank.kaiserpermanente.org/for-researchers/data-resource/. Access to the RPGEH and KPRB data is possible through the KPRB website (). All applications are reviewed by the KPRB prior to data release. Researchers without an official Kaiser Permanente affiliation are required to collaborate with Kaiser Permanente researchers. The KPRB can facilitate these collaborations.

Author Disclosure Statement

No competing financial interests exist.

Funding Information

Authors did not receive funding specifically for this work. The authors received funding from the following agencies in the past 36 months that are unrelated to the study and its findings: American Heart Association (A.S.R.), Gilead Sciences (R.C.H.), Kaiser Permanente Garfield Memorial Fund (C.R.C., R.C.H.), Merck (C.R.C.), National Cancer Institute (C.R.C., R.C.H.), and National Institute on Drug Abuse (R.C.H.), and Opioid Post Marketing Requirements Consortium (R.C.H.). The Kaiser Permanente Research Program on Genes, Environment, and Health (RPGEH) was funded by the Robert Wood Johnson Foundation, the Wayne and Gladys Valley Foundation, the Ellison Medical Foundation, and the Kaiser Permanente Community Benefits Program. The Genetic Epidemiology Research on Adult Health and Aging (GERA) genotypic data was funded by the National Institutes of Health (RC2 AG036607 [Schaefer and Risch]). Data came from a grant, the Resource for Genetic Epidemiology Research in Adult Health and Aging (RC2 AG033067; Schaefer and Risch, PrincipaI Investigators), awarded to the RPGEH and the University of California San Francisco Institute for Human Genetics.

Supplementary Material

References

, Xu

, Tornello

. Stability of self-reported same-sex and both-sex attraction from adolescence to young adulthood. Arch Sex Behav, 2016; 45(3):651–659; doi: 10.1007/s10508-015-0541-1

Mock

, Eibach

. Stability and change in sexual orientation identity over a 10-year period in adulthood. Arch Sex Behav, 2012; 41(3):641–648; doi: 10.1007/s10508-011-9761-1

Ott

, Wypij

, Corliss

, et al. Repeated changes in reported sexual orientation identity linked to substance use behaviors in youth. J Adolesc Health, 2013; 52(4):465–472; doi: 10.1016/j.jadohealth.2012.08.004

Diamond

. Female bisexuality from adolescence to adulthood: Results from a 10-year longitudinal study. Dev Psychol, 2008; 44(1):5–14; doi: 10.1037/0012-1649.44.1.5

Katz-Wise

, Todd

. The current state of sexual fluidity research. Curr Opin Psychol, 2022; 48:101497; doi: 10.1016/j.copsyc.2022.101497

Dickson

, van Roode

, Cameron

, et al. Stability and change in same-sex attraction, experience, and identity by sex and age in a New Zealand birth cohort. Arch Sex Behav, 2013; 42(5):753–763; doi: 10.1007/s10508-012-0063-z

Mittleman

. Sexual fluidity: Implications for population research. Demography, 2023; 60(4):1257–1282; doi: 10.1215/00703370-10898916

Lilly

, Satherley

, Sibley

, et al. Fixed or fluid? Sexual identity fluidity in a large national panel study of New Zealand adults. J Sex Res, 2024; 61(9):1351–1366; doi: 10.1080/00224499.2023.2289517

Scheitle

, Wolf

. Religion and sexual identity fluidity in a national three-wave panel of U.S. adults. Arch Sex Behav, 2018; 47(4):1085–1094; doi: 10.1007/s10508-017-0979-4

10.

Campbell

. Sexual fluidity among millennial women: Journeys across a shifting sexual landscape. Palgrave Macmillan: Cham, Switzerland; 2022.

11.

Redcay

, Counselman-Carpenter

, Proctor

. The impact of fluid attraction and fluid identity on mental health. J Gay Lesbian Ment Health, 2019; 23(1):83–98; doi: 10.1080/19359705.2018.1528915

12.

Coulaud

, Salway

, Adams

, et al. Knowledge gaps in existing research exploring sexual fluidity and mental health among young adults. J Epidemiol Community Health, 2024; 78(9):556–560; doi: 10.1136/jech-2023-221844

13.

Frost

, Meyer

. Minority stress theory: Application, critique, and continued relevance. Curr Opin Psychol, 2023; 51:101579; doi: 10.1016/j.copsyc.2023.101579

14.

Everett

. Sexual orientation identity change and depressive symptoms: A longitudinal analysis. J Health Soc Behav, 2015; 56(1):37–58; doi: 10.1177/0022146514568349

15.

Srivastava

, Hall

, Krueger

, et al. Sexual identity fluidity, identity management stress, and depression among sexual minority adolescents. Front Psychol, 2022; 13:1075815; doi: 10.3389/fpsyg.2022.1075815

16.

Sumerau

, Mathers

LAB

, Moon

. Foreclosing fluidity at the intersection of gender and sexual normativities. Symbol Interact, 2020; 43(2):205–234; doi: 10.1002/symb.431

17.

Lick

, Durso

, Johnson

. Minority stress and physical health among sexual minorities. Perspect Psychol Sci, 2013; 8(5):521–548; doi: 10.1177/1745691613497965

18.

Fricke

, Sironi

. Sexual fluidity and BMI, obesity, and physical activity. SSM Popul Health, 2020; 11:100620; doi: 10.1016/j.ssmph.2020.100620

19.

Database of Genotypes and Phenotypes. Research Program on Genes, Environment and Health (RPGEH). National Center for Biotechnology Information: Bethesda, MD; 2018. Available from: https://www.ncbi.nlm.nih.gov/projects/gap/cgi-bin/study.cgi?study_id=phs000788.v1.p2 [Last accessed: October 21, 2024].

20.

Kaiser Permanente. Kaiser Permanente Research Bank. Kaiser Permanente: Hyattsville, MD; 2019. Available from: https://researchbank.kaiserpermanente.org/ [Last accessed: October 21, 2024].

21.

HealthMeasures. List of adult measures—available PROMIS® measures for adults. Northwestern University: Chicago, IL; 2024. Available from: https://www.healthmeasures.net/explore-measurement-systems/promis/intro-to-promis/list-of-adult-measures [Last accessed: October 21, 2024].

22.

Revicki

, Kawata

, Harnam

, et al. Predicting EuroQol (EQ-5D) scores from the patient-reported outcomes measurement information system (PROMIS) global items and domain item banks in a United States sample. Qual Life Res, 2009; 18(6):783–791; doi: 10.1007/s11136-009-9489-8

23.

Health Measures. PROMIS scale v1.2—global health. Northwestern University: Chicago, IL; 2024. Available from: https://www.healthmeasures.net/index.php?option=com_instruments&view=measure&id=778 [Last accessed: April 18, 2025].

24.

Kronk

, Everhart

, Ashley

, et al. Transgender data collection in the electronic health record: Current concepts and issues. J Am Med Inform Assoc, 2022; 29(2):271–284; doi: 10.1093/jamia/ocab136

25.

Krieger

. Methods for the scientific study of discrimination and health: An ecosocial approach. Am J Public Health, 2012; 102(5):936–944; doi: 10.2105/AJPH.2011.300544

26.

Arel-Bundock

, Greifer

, Heiss

. How to interpret statistical models using marginaleffects for R and python. J Stat Soft, 2024; 111(9); doi: 10.18637/jss.v111.i09

27.

Sullivan

, Salter

, Ryan

, et al. Bias and precision of the “multiple imputation, then deletion” method for dealing with missing outcome data. Am J Epidemiol, 2015; 182(6):528–534; doi: 10.1093/aje/kwv100

28.

Greifer

. Weightit: Weighting for covariate balance in observational studies. Greifer: Cambridge, MA; 2024. Available from: https://ngreifer.github.io/WeightIt/ [Last accessed: October 21, 2024].

29.

van Buuren

, Groothuis-Oudshoorn

. mice: Multivariate imputation by chained equations in R. J Stat Softw 2011; 45(3):1–67; doi: 10.18637/jss.v045.i03

30.

White

, Royston

, Wood

. Multiple imputation using chained equations: Issues and guidance for practice. Stat Med, 2011; 30(4):377–399; doi: 10.1002/sim.4067

31.

HealthMeasures. Meaningful change for PROMIS®. Northwestern University: Chicago, IL; 2024. Available from: https://www.healthmeasures.net/score-and-interpret/interpret-scores/promis/meaningful-change [Last accessed: October 21, 2024].

32.

McClure

, Sayah

, Xie

, et al. Instrument-defined estimates of the minimally important difference for EQ-5D-5L index scores. Value Health, 2017; 20(4):644–650; doi: 10.1016/j.jval.2016.11.015

33.

Bishop

, Moran

, Turpin

, et al. Binge drinking disparities by grade, race and ethnicity, sexual orientation, and gender identity. Pediatrics, 2024; 154(3):e2023063978; doi: 10.1542/peds.2023-063978

34.

Hinds

, Loukas

, Perry

. Explaining sexual minority young adult cigarette smoking disparities. Psychol Addict Behav, 2019; 33(4):371–381; doi: 10.1037/adb0000465

35.

Mathew

, Bhatt

, Colangelo

, et al. Life-course smoking trajectories and risk for emphysema in middle age: The CARDIA lung study. Am J Respir Crit Care Med, 2019; 199(2):237–240; doi: 10.1164/rccm.201808-1568LE

36.

Britton

, Ben-Shlomo

, Benzeval

, et al. Life course trajectories of alcohol consumption in the United Kingdom using longitudinal data from nine cohort studies. BMC Med, 2015; 13:47; doi: 10.1186/s12916-015-0273-z

37.

Ross

, Salway

, Tarasoff

, et al. Prevalence of depression and anxiety among bisexual people compared to gay, lesbian, and heterosexual individuals: A systematic review and meta-analysis. J Sex Res, 2018; 55(4–5):435–456; doi: 10.1080/00224499.2017.1387755

38.

Pachankis

, Mahon

, Jackson

, et al. Sexual orientation concealment and mental health: A conceptual and meta-analytic review. Psychol Bull, 2020; 146(10):831–871; doi: 10.1037/bul0000271

39.

Hsieh

, Ruther

. Sexual minority health and health risk factors: Intersection effects of gender, race, and sexual identity. Am J Prev Med, 2016; 50(6):746–755; doi: 10.1016/j.amepre.2015.11.016

40.

Katz-Wise

, Ranker

, Gordon

, et al. Sociodemographic patterns in retrospective sexual orientation identity and attraction change in the sexual orientation fluidity in youth study. J Adolesc Health, 2023; 72(3):437–443; doi: 10.1016/j.jadohealth.2022.10.015

41.

Bowleg

. When black + lesbian + woman ≠ black lesbian woman: The methodological challenges of qualitative and quantitative intersectionality research. Sex Roles, 2008; 59(5–6):312–325; doi: 10.1007/s11199-008-9400-z

42.

Chun

KYS

, Singh

. The bisexual youth of color intersecting identities development model: A contextual approach to understanding multiple marginalization experiences. J Bisex, 2010; 10(4):429–451; doi: 10.1080/15299716.2010.521059

43.

Poteat

, Logie

. A case for strengths-based approaches to addressing intersectional stigma in HIV research. Am J Public Health, 2022; 112(S4):S347–S349; doi: 10.2105/AJPH.2021.306659

44.

Forrest

, Beccia

, Exten

, et al. Intersectional prevalence of suicide ideation, plan, and attempt based on gender, sexual orientation, race and ethnicity, and rurality. JAMA Psychiatry, 2023; 80(10):1037–1046; doi: 10.1001/jamapsychiatry.2023.2295

45.

Jylha

. What is self-rated health and why does it predict mortality? Towards a unified conceptual model. Soc Sci Med, 2009; 69(3):307–316; doi: 10.1016/j.socscimed.2009.05.013

46.

Diamond

. Sexual fluidity in male and females. Curr Sex Health Rep, 2016; 8(4):249–256; doi: 10.1007/s11930-016-0092-z

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.01 MB

0.02 MB

0.01 MB

0.02 MB

0.14 MB