Disparities in Suicidality by Gender Identity Among Medicare Beneficiaries

Abstract

Introduction:

Suicidality is higher for gender minorities than the general population, yet little is known about suicidality in disabled or older adult gender minorities.

Methods:

This study used 2009–2014 Medicare claims to identify people with gender identity–related diagnosis codes (disabled, n=6,678; older adult, n=2,018) and compared their prevalence of suicidality with a 5% random “non-gender minority” beneficiary sample (disabled, n=535,801; older adult, n=1,700,008). Correlates of suicidality were assessed (via chi-square) for each of the four participant groups separately, and then disparities within eligibility status (disabled or older adult) were assessed using logistic regression models, adjusting first for age and mental health chronic conditions, and then additionally for Medicaid eligibility, race/ethnicity, or U.S. region (each separately). The primary hypotheses were that gender minority beneficiaries would have higher suicidality, but that suicidality disparities would persist after adjusting for covariates. Data were analyzed between 2017 and 2019.

Results:

Gender minority beneficiaries had higher unadjusted suicidality compared with non-gender minority beneficiaries in the disabled cohort (18.5% vs 7.1%, p<0.001). Significant suicidality predictors in all four groups included: age (except in older adult gender minorities), Medicaid eligibility, depression or behavioral health conditions, avoidable hospitalizations, and violence victimization. In age- and mental health–adjusted logistic regression models, gender minorities had higher odds of suicidality than non-gender minority beneficiaries (disabled, OR=1.95, p<0.0001; older adult, OR=2.10, p<0.0001). Disparities were not attenuated after adjusting for Medicaid eligibility, race/ethnicity, or region.

Conclusions:

Heightened suicidality among identified gender minority Medicare beneficiaries highlights a pressing need to identify and reduce barriers to wellness in this population.

INTRODUCTION

People whose gender identity differs from societal expectations based on their sex assigned at birth (i.e., transgender or gender non-binary people; henceforth “gender minorities”) have elevated rates of depression, anxiety,^1,2 and suicide attempt or ideation.^3–11 Fifty five percent have lifetime suicidal ideation and 29% attempt suicide,¹² versus 9.2% and 2.7%, respectively, in the general population.¹³ Historical and ongoing societal marginalization of gender minorities¹⁴ results in more suicide risk factors including significant life crises⁸ (e.g., housing instability,^15,16 financial strain,^6,16 or forced sex),^3,17 substance abuse,³ days with poor mental health,¹⁸ and depression.³ Marginalization also results in gender minority–specific suicide risks like internalized stigma,¹⁵ transphobia,⁸ gender identity–based healthcare discrimination,¹⁹ and targeted discrimination or violence victimization.^3,6,10,16,20

Observed suicide prevalence among gender minorities varies by study methodology and subpopulation,^6,17,21 and most existing research focuses on the higher risk among gender minority youth^3,7,8,11 or Veterans.^5,6,16,22,23 Meanwhile, suicide risk is also elevated after age 60 years,²⁴ and there are concerns that as the U.S. “baby boomer” cohort ages, their historically high suicide attempt rates will translate to an increase in suicide attempts for people aged >65 years.²⁵ Stressors for suicide attempt in older adults include social disconnectedness, chronic physical illness, psychological stress, disability, and can also include impaired decision making.²⁶ Disability-related functional limitations are also independently associated with suicide attempt.^27,28 Older adult gender minorities have poorer physical health, more disability and depression, and higher perceived stress than older sexual minority adults,²⁹ and often lack affordable housing and family support.^30–33 Gender identity–specific worries include “that physical aging [will make] them less able to withstand physical attack” if targeted based on gender identity.³⁴ Disabled gender minorities may share some of the same risk factors as the older adult gender minority population, including elevated social isolation³⁵ and other psychosocial burdens.³⁶ Researchers have called for more studies of mental health of diverse gender minority populations across the lifespan, yet there are no large-scale studies of suicide risk in older adult gender minorities and almost no research on this topic at all for disabled gender minorities.

This analysis seeks to close this critical evidentiary gap using a diagnosis code–based algorithm to identify people with gender minority–related diagnosis codes in 2009–2014 Medicare insurance claims data (who have elevated rates of mental health diagnoses^1,37,38) and compare their prevalence of suicidality to that of other Medicare beneficiaries. The hypotheses are that people with gender-minority related diagnosis codes have higher prevalence of suicidality than non-gender minority beneficiaries, that disparities persist after adjusting for age and behavioral health diagnoses, and that significant sociodemographic risk factors for suicidality mirror known risk factors in the general population.

METHODS

Study Sample

Medicare provides health insurance for U.S. citizens or permanent legal residents that are aged >65 years or who have a long-term disability. A group of gender minority Medicare beneficiaries were identified using ICD-9 diagnosis codes that appear in billing claims when people receive medical care (e.g., hormone therapy) related to gender identity and described in prior work^1,37,38: “gender identity disorder” (302.6 and 302.85) and “transsexualism” (302.50–302.53). These gender minority beneficiaries were identified in 100% of 2009–2014 claims (Inpatient, Outpatient, Skilled Nursing Facility, Home Health, Carrier, and Part D Event files), and compared with beneficiaries who do not have gender identity-related diagnoses (“non-gender minorities”) with at least one non-pharmacy claim in each year studied to ensure engagement in care.³⁸ Herein, “gender minority beneficiaries” refers to beneficiaries with gender minority–related diagnosis codes, and “non-gender minority beneficiaries” refers to beneficiaries without these codes.

For each year studied, beneficiaries were included if they were continuously enrolled in Medicare Parts A (hospital coverage) and B (outpatient services coverage), and never Part C (Medicare Advantage program; claims not available) for 12 months in that year. Analyses were stratified by original eligibility: age ≥65 versus disability (those qualifying because of end-stage renal disease were excluded). Beneficiary demographics, enrollment, and chronic disease indicators were extracted from the Master Beneficiary Summary File. The IRB of the Cambridge Health Alliance approved this research.

Measures

Suicidal behavior may be under-detected by ICD-9 codes for suicide attempt (E95*) alone³⁹ including because of “coding practices, reimbursement patterns, and uncertainty of intent,”⁴⁰ as well as the possibility that people with suicidal behavior do not receive medical treatment. To maximize the ability to detect suicidal behavior in claims data, this study measured three aspects of suicidality: suicide attempt (E95*; injuries of intentional intent), suicide ideation (V-code; V62.84), and potential suicide attempt as described in Barak–Cohen et al.⁴⁰ Potential suicide attempt codes with positive predictive value >0.70 for suicide attempt based on research combining data from electronic health records with death certificate data⁴⁰ were included: 965.* (poisoning by analgesics, antipyretics, and anti-rheumatics), 967.* (poisoning by sedatives and hypnotics), 969.* (poisoning by psychotropic agents) and 881.* (open wound of elbow forearm and wrist).⁴⁰ A composite measure for “any suicidality” refers to presence of any of the three component measures in the specified observation period.

Covariates examined for their relationship with suicide attempt or ideation in bivariate analyses included: age first observed (18–24, 25–34, 35–44, 45–54, 55–64, 65–74, 75–84, 85–94, and ≥95 years), RTI race code (white, black/African American, Hispanic, Asian/Pacific Islander, American Indian/Alaskan Native, other, or unknown race), Medicaid eligibility ever, U.S. region (Northeast, South, Midwest, or West), depression or other behavioral health Chronic Condition Warehouse flags, any avoidable hospitalizations in non-institutional settings (signaling inadequate access to preventive care),⁴¹ and any violence victimization (a risk factor for suicidal ideation⁴²)—broad case definition including ICD-9 codes for abuse, assault, and intentional injury.⁴³ Suicidality and violence victimization codes did not overlap.

Statistical Analysis

In unadjusted analyses, the prevalence of any suicidality and each component measure (attempt, ideation, or potential attempt) were compared between gender minority and non-gender minority beneficiaries at any point during 2009–2014 using chi-square tests, stratified by eligibility cohort (disabled versus older adult). Covariate correlates of any suicidality were then compared within each of the four populations separately.

Adjusted analyses compared gender minority with non-gender minority odds of any suicidality and each component measure in 2009–2014 (each year modeled as a repeated observation per person) using logistic regression models, stratified by eligibility cohort. The National Academy of Medicine (formerly known as the Institute of Medicine) defines health disparities⁴⁴ as those differences between groups that remain after adjusting for need (health status) and preferences: The first set of models therefore adjusted for need using age (continuous) and behavioral health status (17 binary behavioral health indicators from the Master Beneficiary Summary File) in line with prior work.^38,45 Adjusted models therefore characterized the remaining disparity in suicidality even after accounting for differences in the distribution of age and behavioral health status in the two samples. The second set of models further adjusted any suicidality for Medicaid dual eligibility, race/ethnicity, or U.S. region to examine whether these factors mitigate any observed disparities.

A Bonferonni-adjusted p-value of 0.00083 was used to determine significance for primary analyses (p=0.05/n=60 comparisons). Clustered error structure at the beneficiary level accounted for within- and between-subject variance (SAS, version 9.4 PROC GENMOD).

RESULTS

Descriptive information about these cohorts was described previously.³⁸ Identified gender minority beneficiaries were younger (among the disabled), more often dually eligible for Medicaid, and had more mental health diagnoses than non-gender minority beneficiaries.³⁸ Disabled gender minority beneficiaries were more likely than disabled non-gender minorities in 2009–2014 to have a suicide attempt (5.1% vs 1.3%), suicidal ideation (15.4% vs 3.5%), and potential suicide attempt (6.5% vs 4.3%) (Table 1, all p<0.0001), as well as any suicidality (18.5% vs 7.1%). Older adult gender minorities were more likely to have a suicide attempt (0.5% vs 0.1%) and suicidal ideation (1.1% vs 0.4%, both p<0.0001), but less likely to have a potential suicide attempt (1.7% vs 3.1%, p<0.0002) than older adult non-gender minorities; any suicidality prevalence did not differ significantly.

Table 1.

Percent of Continuously Enrolled Medicare Beneficiaries With Suicidality 2009‒2014, by Cohort and Gender Minority Status

Variable	Originally eligible because of disability			Originally eligible because of age ≥65 years
	Non-GM disabled (N=535,801) % (SE)		GM disabled (N=6,678) % (SE)	Non-GM older adults (N=1,700,008) % (SE)		GM older adults (N=2,018) % (SE)
Differences between groups
Any suicidality (suicide attempt, ideation, or potential attempt)	7.1 (0.04)		18.5 (0.48)	3.5 (0.01)		2.9 (0.38)
Between-group χ2 p-value		p<0.0001			p<0.0001
Suicide attempt	1.3 (0.02)		5.1 (0.27)	0.1 (<0.01)		0.5 (0.16)
Between-group χ2 p-value		p<0.0001			p<0.0001
Suicidal ideation	3.5 (0.03)		15.4 (0.44)	0.4 (<0.01)		1.1 (0.23)
Between-group χ2 p-value		p<0.0001			p<0.0001
Potential suicide attempt	4.3 (0.03)		6.5 (0.3)	3.1 (0.01)		1.7 (0.29)
Between-group χ2 p-value		p<0.0001			p<0.0002
Significance within each group
Any suicidality, by age first observed, years
18‒24	11.5 (0.32)		25.1 (2.67)	n/a		n/a
25‒34	12.7 (0.18)		23.2 (1.21)	n/a		n/a
35‒44	11.3 (0.13)		21.8 (1.07)	n/a		n/a
45‒54	8.5 (0.08)		18.0 (0.89)	n/a		n/a
55‒64	5.1 (0.05)		14.0 (0.92)	n/a		n/a
65‒74	4.5 (0.07)		8.2 (1.43)	1.9 (0.01)		3.25 (0.52)
75‒84	5.8 (0.13)		Too small to reporta	4.6 (0.03)		Too small to reporta
85‒94	7.2 (0.28)		0.0	7.2 (0.05)		Too small to reporta
Within-group χ2 p-value	p<0.0001		p<0.0001	p<0.0001		p=0.0013
Any suicidality by race/ethnicity
White, not Hispanic	7.9 (0.04)		20.2(0.58)	3.8 (0.02)		3.1 (0.41)
Black/African American	4.9 (0.07)		13.1 (1.04)	1.0 (0.03)		Too small to reporta
Hispanic	5.9 (0.11)		12.6 (1.46)	2.1 (0.05)		Too small to reporta
Other	5.5 (0.37)		29.2 (6.56)	1.5 (0.12)		Too small to reporta
Asian/Pacific Islander	4.6 (0.25)		14.9 (3.54)	1.3 (0.06)		0
American Indian/Alaskan	8.9 (0.41)		27.7 (4.91)	3.2 (0.22)		Too small to reporta
Native
Unknown race	6.0 (0.49)		Too small to reporta	0.9 (0.09)		0
Within-group χ2 p-value	p<0.0001		p<0.0001	p<0.0001		Not estimable
Any suicidality by Medicaid dual eligibility status
Never dually eligible	4.7 (0.04)		11.7 (0.77)	3.4 (0.01)		2.3 (0.37)
Dually eligible ever	9.5 (0.06)		20.9 (0.58)	4.1 (0.04)		5.7 (1.23)
Within-group χ2 p-value	p<0.0001		p<0.0001	p<0.0001		p=0.0008
Any suicidality by region
South	6.8 (0.05)		16.8 (0.88)	3.7 (0.02)		3.0 (0.68)
West	7.7 (0.09)		18.7 (0.90)	3.8 (0.03)		3.3 (0.78)
Northeast	7.0 (0.08)		18.7 (1.01)	3.0 (0.03)		Too small to reporta
Midwest	7.5 (0.07)		20.2 (1.03)	3.1 (0.03)		3.2 (0.88)
Within-group χ2 p-value	p<0.0001		p=0.0894	p<0.0001		p=0.7534
Any suicidality, by depression flag in any year
No depression flag	2.2 (0.03)		2.7 (0.35)	2.3 (0.01)		1.3 (0.3)
Any depression flag	13.0 (0.07)		26.3 (0.66)	7.1 (0.04)		7.2 (1.1)
Within-group χ2 p-value	p<0.0001		p<0.0001	p<0.0001		p<0.0001
Any suicidality, by any behavioral health condition flags
No chronic condition flags	1.2 (0.03)		1.6 (0.40)	1.5 (0.01)		Too small to reporta
Chronic condition flag in any year	9.6 (0.05)		21.4 (0.54)	5.6 (0.03)		5.3 (0.72)
Within-group χ2 p-value	p<0.0001		p<0.0001	p<0.0001		p<0.0001
Any suicidality by avoidable hospitalizations
No avoidable hospitalization	4.0 (0.03)		10.5 (0.45)	1.6 (0.01)		1.2 (0.29)
Any avoidable hospitalization	11.7 (0.07)		37.3 (1.08)	6.1 (0.03)		7.2 (1.09)
Within-group χ2 p-value	p<0.0001		p<0.0001	p<0.0001		p<0.0001
Any suicidality 2009–2014 by violence victimization status
No documented violence victimization	6.3 (0.03)		15.5 (0.46)	3.4 (0.01)		2.7 (0.36)
Documented violence victimization	34.7 (0.38)		56.2 (2.21)	16.5 (0.56)		Too small to reporta
Within-group χ2 p-value	p<0.0001		p<0.0001	p<0.0001		Not estimable

Note: Boldface indicates Bonferonni-adjusted statistical significance of p<0.0008 for significance. Non–gender minority patients were from a 5% random sample of Medicare beneficiaries in each year who had at least one claim and who were not identified as gender minority beneficiaries. All beneficiaries in both cohorts were continuously enrolled in Medicare Parts A and B (and not Part C) for 12 months in each year studied. Inclusion criteria for non–gender minority patients also included at least one non-pharmacy claim in the year studied. Beneficiaries with end-stage renal disease were excluded.

^aValue is suppressed in compliance with Centers for Medicare and Medicaid Services requirements that reporting not include cell sizes with n<12. We exclude people ages 95 and older because cell sizes in this age group are too small to report frequencies of suicide.

GM, gender minority.

Significant predictors of any suicidality in all four groups included: younger age and being aged >75 years (not significant for older adult gender minorities), white and American Indian/Alaskan Native race/ethnicity (not estimable in older adult gender minorities), Medicaid eligibility, living in the West (non-gender minority beneficiaries only), having behavioral health chronic condition flags, avoidable hospitalizations (less prevalent for gender minorities in both groups; results available upon request), and violence victimization (more prevalent in gender minorities in the disabled population; results available upon request).

In logistic regression models of disabled beneficiaries adjusting for age and behavioral health conditions, gender minority disabled beneficiaries had significantly higher odds compared with non-gender minority beneficiaries of any suicidality (OR=1.95, 95% CI=1.82, 2.09), as well as of suicide attempt (OR=1.70, 95% CI=1.50, 1.94), suicidal ideation (OR=2.23, 95% CI=2.05, 2.43), and potential suicide attempt (OR=1.36, 95% CI=1.22, 1.50) (Table 2, all p<0.0001; full regression output Appendix Table 1).

Table 2.

ORs for Suicidality in Each Year 2009‒2014 by Eligibility Cohort and Gender Minority Status

Cohort	OR GM vs Non-GM (95% CI)	p-value
Suicidality (any) including suicide attempt, suicidal ideation, or potential attempt
Older adults	2.10 (1.60, 2.75)	<0.0001
Disabled adults	1.95 (1.82, 2.09)	<0.0001
Suicide attempt
Older adults	4.37 (2.26, 8.46)	<0.0001
Disabled adults	1.70 (1.50, 1.94)	<0.0001
Suicidal ideation
Older adults	2.61 (1.56, 4.37)	0.0003
Disabled adults	2.23 (2.05, 2.43)	<0.0001
Potential suicide attempt
Older adults	1.43 (1.01, 2.00)	0.0412
Disabled adults	1.36 (1.22, 1.50)	<0.0001

Note: Results are from authors’ analysis of 2009‒2014 Medicare Research Identifiable Files. Boldface indicates Bonferonni-adjusted statistical significance of p<0.0008 for significance. All models adjust for age and behavioral health conditions (as binary indicators). Outcomes are modeled as repeated observations per individual; clustered error structure at the beneficiary level accounted for within- and between-subject variance.

GM, gender minority.

Older adult gender minorities showed similar patterns of higher odds compared with non-gender minority older adults of any suicidality (OR=2.10, 95% CI=1.60, 2.75, p<0.0001), suicide attempt (OR=4.37, 95% CI=2.26, 8.46, p<0.0001), and suicidal ideation (OR=2.61, 95% CI=1.56, 4.37, p<0.0003), but no significant difference in potential suicide attempt (p=0.0412) (Table 2; full regression output Appendix Table 2).

All age- and behavioral health–adjusted disparities remained significant after adjustment for dual eligibility, race/ethnicity, and region. No interaction terms were significant (Appendix Table 3).

DISCUSSION

Medicare beneficiaries with gender identity–related diagnosis codes in 2009–2014 had higher unadjusted prevalence of suicide attempt and suicidal ideation (both disabled and older adult cohorts) and potential suicide attempt (disabled cohort only), as well any suicidality (disabled only). Shared predictors of any suicidality included: age (younger adult or aged >75 years; except in older adult gender minorities), Medicaid dual eligibility, behavioral health diagnoses, avoidable hospitalizations, and violence victimization (not estimable in older adult gender minorities). After adjusting for age and behavioral health diagnoses, gender minority beneficiaries still had higher odds of suicide attempt and ideation (disabled and older adults), potential suicide attempt (disabled only), and any suicidality measure (disabled and older adults). Further adjustment for Medicaid eligibility, race/ethnicity, or U.S. region did not mitigate this disparity.

These findings corroborate other studies showing higher suicide rates among gender minority populations generally,^{3–9,11,15,16,21,22} in a population of disabled or older adult Medicare beneficiaries and confirm in these cohorts many factors associated with suicide risk in the general population, including: younger adulthood,^13,46–48 older adulthood,⁴⁹ depression or mood disorders,^13,48 disability,⁴⁶ violence victimization,⁵⁰ lower incomes,⁴⁷ unemployment,⁵¹ and to some extent race^52,53 and region⁴⁶ (although evidence is mixed and may be driven by varied access to firearms^47,54).

Prior studies using survey data to measure past-year or lifetime suicidality would be expected to have much higher observed prevalence of suicidality than is observable in claims data, where only instances of suicidality which are treated in medical settings, coded as suicide-related, and observed in the observation window can be counted. However, prior studies of transgender Veterans in electronic health record databases spanning nearly 2 decades found a 13.8% prevalence of suicidal ideation and 7.1% prevalence of suicide attempt,¹⁶ which, despite its longer observation window, was similar to present findings in the disabled gender minority population during a 6-year period.

This is the first study to quantify that a clear disparity in suicidality exists for disabled and older adult Medicare beneficiaries with gender identity–related diagnosis codes who have similar levels of behavioral health diagnoses as other beneficiaries, and that these disparities are robust to further adjusting for the limited sociodemographic data available in Medicare claims. This suggests that other risk factors that are more prevalent for gender minorities, but not observable in Medicare claims, may be responsible for the observed suicidality disparities. Gender minority people experience specific minority stressors⁵⁵ at home, school, work, and within the community,⁵⁶ and these stressors are associated with increased psychological distress⁵⁷ and suicidal ideation.²⁰ Optimism, more reasons for living, social or family support,^4,58 and access to transition-related medical care for those gender minorities who need it^23,59 serve as protective factors for gender minorities, whereas disability- and aging-related health and emotional burdens may be compounded by discrimination or internalized gender minority stigma.^15,55

These findings should prompt increased efforts to meet the wellness needs of these populations. Recommendations for suicide prevention strategies among older adult gender minorities include addressing stigma and victimization²⁹; increasing physical activity²⁹; and fostering resilience including through social support and community belongingness,^29,60,61 self-acceptance, personal agency, and spirituality.³⁴ But there is little to no research on suicidality specifically addressing disabled gender minorities, who likely experience multiple interacting forms of marginalization.

Adopting or adapting existing recommendations requires population-specific efforts. Reducing structural stigma⁶² toward gender minorities, including through gender identity employment nondiscrimination policies,⁵ may hold promise for reducing suicide risk. Transgender older adults may feel paradoxically less social support and community belongingness despite having larger social networks than lesbian, gay, or bisexual older adults²⁹; interventions may therefore need to help foster deeper connectedness within existing social networks. One successful model for promoting wellness in gender minority youth grapples with both acceptance and rejection within one’s family, peer, and community spheres (The Family Acceptance Project⁶³). Adapting similar multi-intervention strategies for older adult or disabled gender minorities may be necessary. Promoting physical activity for older adult and disabled minorities may be further complicated by concerns in public spaces about body image, safe and equitable access to changing facilities and restrooms, and fears of harassment or victimization.^29,64

Despite the higher suicidality among identified gender minority Medicare beneficiaries, mental health treatment patterns for gender minority Medicare beneficiaries show declining or stable rates of outpatient mental health visits alongside increased psychotropic medication prescribing.⁴⁵ These patterns call into question whether gender minority Medicare beneficiaries face financial or other barriers to outpatient treatment as have been identified for other high-need gender minority populations,⁶⁵ and whether they are receiving adequate psychosocial support from mental health providers to build the aforementioned resiliency tools. Less frequent outpatient use coupled with greater reliance on medications may lead to fewer opportunities to identify suicidal ideation and prevent suicide attempt. People avoid reporting suicidal ideation owing to “stigma, overreaction [from providers], and loss of autonomy,” from involuntary hospitalization.⁶⁶ The current lack of sufficient training about medical caring for gender minority people¹⁴ often leads to “…uncertainty and ambivalence in the medical encounter [that] upsets the normal balance of power in provider–patient relationships,”⁶⁷ and concerns about disclosure of suicidal ideation are likely to be more pronounced for gender minorities who do not have a provider they can trust fully. Gender minority–affirming suicide risk assessments should also address gender minority–specific trauma, body image issues, concerns about outness/disclosure,⁶⁸ and may require consideration as to how mental health professionals’ frequent “gatekeeping” role^69,70 for transition-related care may further complicate gender minorities’ willingness to fully disclose suicidality.

Limitations

This study had several limitations. First, as discussed in more detail in prior work,^{37,38,45,71–73} the Centers for Medicare and Medicaid Services algorithm does not capture people without gender dysphoria diagnoses, and may therefore be less likely to capture beneficiaries who are non-binary or gender fluid, who do not want some types of gender-affirming care, (e.g., hormones), or who receive such care through clinics or other sources not reimbursed by Medicare. Limited analyses in a single health system combining ICD-9 codes and keyword searches available in clinical notes indicated that methods that use ICD-9 codes alone have a positive predictive value of 56% for identifying gender minorities among those identified in electronic health record data,⁷⁴ but the dearth of population-level gender minority research means that the relative size of the unobserved gender minority population is difficult to estimate at this time. The generalizability of these findings to all gender minority Medicare beneficiaries, including those without gender dysphoria diagnoses, is therefore limited. Second, claims-based behavioral health diagnoses do not adequately capture severity of mental health conditions or undiagnosed mental health need; therefore, some of the remaining observed difference in suicidality may still be due to unobserved differences in behavioral health need (which would nevertheless remain of extreme public health importance). Third, claims data cannot capture lifetime suicide attempt, which may for example partially explain lower rates of suicidality in older adults if those beneficiaries are “healthy survivors.” Fourth, claims data underestimate the prevalence of suicidality for the aforementioned reasons, and may especially underestimate prevalence for underserved populations like transgender people of color.⁷⁵ Fifth, the potential suicide attempt codes may also be capturing instances of non-suicidal self-injury, although differences in the prevalence of these codes was not a driver of results. Sixth, the small number of older adult gender minorities limited the power to assess some suicidality risk factors in this age group. Seventh, these analyses include 2009–2014 data, and changes to the ICD-10 coding system in 2015 introduced major changes to the codes for suicide-related outcomes; therefore, future diagnosis code–based estimates of suicidality will not be directly comparable. In spite of these data limitations, these analyses provide much-needed data on suicide risk for older adult and disabled gender minorities with gender dysphoria–related diagnosis codes.

CONCLUSIONS

Future research in this area will need to consider the significant changes in medical coding related to gender identity. The ICD-10 coding system, updated in 2015, included a new code for “history of sex reassignment,” and analyses from the Centers for Medicare and Medicaid Services show that these codes identify more older adult gender minorities in recent years of data.⁷² The DSM-5 was updated to use “gender dysphoria” rather than “gender identity disorder” (meaning that the DSM-5 and ICD-10 now use different terminology). The WHO recently announced that in ICD-11, a diagnosis of “gender incongruence” will be used, and will appear among the codes for sexual, rather than mental, health conditions.⁷⁶ These changes reflect a shift away from stigmatizing gender minorities, and future studies may therefore have larger gender minority sample sizes.

The significant observed disparity in suicidality between gender minority and non-gender minority Medicare beneficiaries from 2009 to 2014 highlights a pressing need to reduce barriers to wellness in this population. Future work should explore how to adapt multilevel models that may protect against suicidality, including by reducing stigma and discrimination as well as fostering resiliency, for the specific needs of gender minorities who are older or who have a disability. Continued efforts are needed to document mental healthcare utilization patterns and quality of mental health care for gender minorities with mental health needs. Finally, routine capture of self-reported gender identity data in health databases would facilitate more robust understanding of health needs and disparities for people of all gender minority identities.