LGBTQ+ mental health, in the data

first, the rule this page is built on

Being lesbian, gay, bisexual or transgender is not a mental illness, and documenting disparities is not diagnosing identity. This page reports measured differences in distress, treatment and risk between populations. It says nothing about what any identity is. That distinction is not a courtesy — it is the settled position of the classification systems themselves, and the history of how they got there is worth having on the table, because this community has lived both sides of what a diagnostic category can do.

The receipts, from the primary documents:

• December 1973 — the American Psychiatric Association's Board of Trustees voted to remove homosexuality from the DSM-II. The debate is preserved in the American Journal of Psychiatry symposium “Should Homosexuality Be in the APA Nomenclature?” [23]. The 1974 printing of DSM-II replaced the diagnosis with “sexual orientation disturbance”; DSM-III (1980) carried “ego-dystonic homosexuality”; DSM-III-R (1987) removed that too [24].
• 2013 — DSM-5 replaced “gender identity disorder” with “gender dysphoria”, relocating the clinical focus from identity to distress; the DSM-5-TR update was, in the APA's own words, “endorsed by the DSM Steering Committee, APA Assembly, and Board of Trustees to use culturally-sensitive and less stigmatizing language” [25].
• 2019 / 2022 — ICD-11, adopted at the 72nd World Health Assembly on 25 May 2019 and in effect since 1 January 2022, moved gender incongruence out of the mental-disorders chapter into “Conditions related to sexual health”. The WHO's own explanation: classification as a mental disorder “can cause enormous stigma”, and the change “reflects current knowledge that trans-related and gender diverse identities are not conditions of mental ill-health” [26]. Its predecessor, ICD-10, had filed F64.0 “Transsexualism” under Mental and Behavioural Disorders.

Note what this history also demonstrates: diagnostic categories are committee decisions. A vote created the diagnosis and a vote removed it. That cuts in every direction on this page, and we will not pretend otherwise when it is convenient.

the ruler: four different quantities that get called “depression rates”

Every exhibit below carries one of four measurement types, and they are not interchangeable:

• self-reported symptom frequency — “how often do you feel depressed?” Not a diagnosis, not a validated screener (NHIS summary statistics).
• screened — a short validated instrument crossed a threshold (K6, GAD-2/PHQ-2). A positive screen is not a diagnosis.
• assessed — survey responses run through diagnostic-criteria logic (NSDUH's DSM-5-based major-depressive-episode recode, its model-based “any mental illness”). Closer to clinical criteria; still not a clinician's diagnosis.
• diagnosed — “has a doctor or other health professional ever told you…” — a report of diagnostic behaviour, not of disease.

And here is the first finding. The federal statistical system publishes a diagnosed anxiety/depression series for all adults (d89q-62iu), broken out by exactly four dimensions: total, age, sex, race/ethnicity [2]. It publishes an orientation breakout of symptom frequency and treatment receipt (25m4-6qqq) [1]. No published federal table crosses sexual orientation with the diagnosed construct — the cross exists only in restricted microdata analysis. Every “LGB diagnosis rate” you have read from federal data was actually one of the other three quantities. This page will not make that substitution.

the published federal backbone: orientation × symptoms and treatment, 2019–2024

The NHIS Adult Summary Health Statistics are the longest clean published federal series crossing sexual orientation with mental-health measures — six annual years, all inside the single post-2019-redesign NHIS, so the years are the survey's own presentation and comparable within the series [1]. The survey publishes three orientation groups. Every panel below is adults 18 and over; orientation is not crossed with age in this product.

Treatment receipt shows the same ordering — which is worth pausing on, because it contradicts the assumption that a disparity in distress must mean a disparity in being left untreated. Bisexual adults report the most distress and the most counseling and the most medication:

The cost-barrier series is the one that does not fit the “more treatment, problem solved” reading. Alongside the highest treatment rates, bisexual adults also report — every single year — the highest rate of not getting needed mental health care because of cost:

the NSDUH snapshot — and why it is a snapshot, not a trend

SAMHSA's National Survey on Drug Use and Health measures different constructs — assessed ones, run through DSM-5-based logic — and publishes sexual-identity breakouts in a dedicated slide deck rather than in its detailed tables (which contain no sexual-identity table in any release we checked) [3][4]. The 2023 numbers, from the deck itself:

Among LGB+ adolescents specifically, the deck reports 44.6% with a past-year major depressive episode, 76.3% of those with severe impairment (slide 28) [3].

Why there is no trend line here. The NSDUH sexual-identity series spans five non-comparable regimes: 2015–2019 (item asked of adults only, three categories) · 2020 (short collection year) · 2021–2022 (multimode redesign — SAMHSA's own release: estimates “should not be compared with previous years”) · 2023 (item changed: all respondents 12+, six categories — the LGB+ deck itself says 2023 should not be compared with 2022 and earlier) · 2024 onward (the item was asked in the field and stripped from the public-use file — zero occurrences of the variable in the 684-page codebook, a removal the data user's guide does not document) [4][5]. Five regimes, never one line. Anyone who shows you an NSDUH “LGB trend” across these years is splicing what the agency says not to splice.

One more provenance note: SAMHSA's earlier annual LGB reports — the 2018, 2019, 2020 and 2021–2022 editions — all return 404 on live samhsa.gov as of 2026-08-26 and survive only in the Internet Archive [6]. The 2023 products are live today. We cite the archived captures, because those are the citations that will still work next year.

the transgender data layer: one probability sample, one ended federal series, and labeled convenience data

For transgender adults, the honest starting point is that no federal probability series publishes diagnosed-condition prevalence — at all, in any year. What exists instead is three kinds of evidence, each with a different label.

The probability anchor. TransPop (2016–2018) is, in ICPSR's words, “the first national probability sample of transgender individuals in the United States” — Gallup-screened, with a transgender sample of 274 adults and 1,162 cisgender comparisons [9]. Its distress measure is screened (Kessler-6). An N of 274 means wide confidence intervals; the study's value is its design, not its precision, and every TransPop number should be read with that in mind.

The one federal series that ever existed — now ended. The Household Pulse Survey is the only federal dataset that ever published a gender-identity breakout of a mental-health measure. Its construct is screened — two-item GAD-2/PHQ-2 instruments — and it is an experimental product with a low response rate; both labels are mandatory. With those caveats displayed, here is the entire series, start to end:

The convenience data, carried under its own label. The U.S. Transgender Survey — the largest survey of transgender people in the country (27,715 respondents in 2015; 92,329 in 2022) — is an anonymous online opt-in sample, and its own methodology chapter says so plainly [11]:

“respondents in this study were not randomly sampled… it is not appropriate to generalize the findings in this study to all transgender people”— 2015 USTS report, chapter 2 — the survey's own words

With that disclaimer attached and never detached: 39% of 2015 respondents and 44% of 2022 respondents met the K6 threshold for serious psychological distress in the past month (screened). These numbers describe the people who chose to take the survey. They are not population prevalence, and this page does not average them with anything. (The 2022 “Early Insights” report publishes no suicidality content — numbers you may have seen attributed to it are not in it [11].)

the suicidality data

Youth (YRBS 2023). The Youth Risk Behavior Survey is the national surveillance of high-school students, grades 9–12; every measure is a self-report questionnaire item. 2023 is also the first year the national YRBS asked about transgender identity — meaning the right-hand table below is the first of its kind, with no earlier national series behind it [7][8].

Adults (NSDUH 2023, assessed/self-report). Serious thoughts of suicide in the past year, age-adjusted, by sexual identity (slide 41): straight 4.1% · gay male 8.1% · lesbian female 11.7% · bisexual 16.7% · some other term 16.7%; overall 5.3% [3]. Among LGB+ adolescents: 34.5% had serious thoughts of suicide, 15.9% made a plan, 10.1% attempted (slide 39) — and SAMHSA's own caveat travels with those numbers: “Many adolescents did not want to answer questions about suicide, which suggests that some adolescents could have had these thoughts but did not feel comfortable disclosing that information” [3].

Transgender adults (TransPop, probability sample, self-report). In the TransPop transgender sample: lifetime suicidal ideation 81.3%, recent ideation 44.4%, recent attempt 6.9% [9][10]. N = 274 — the intervals are wide; the design is what makes the numbers citable at all.

What none of this is: death data. All of the above measures thoughts, plans and attempts, from surveys of living people. Suicide death rates by orientation or gender identity do not exist and cannot exist from current federal data — the next section explains why, and treats that absence as the finding it is.

the panels that are deliberately empty

what was being counted, and stopped

The empty panels above are not all old news. Several of them were in the process of being filled — and the filling stopped, recently and documentably. This is the third time this site has documented public health information disappearing from live federal sources (the FDA AI tracker and the NIMH record are the other two), and the method is the same: every row below cites the agency's own artifacts, with Internet Archive captures bracketing each removal, so none of it depends on trusting us.

The one motive statement this page will make is the one an agency made itself: the 2025 NHIS survey description discontinues the gender-identity questions with an explicit Executive Order 14168 compliance note [27]. Everything else in the table is dates and documents. Readers can draw their own conclusions; the point this page insists on is narrower and harder to argue with: five years from now, “there is no federal data on transgender mental health” will be true partly because the instruments that were beginning to produce it were switched off — and the switch-off is itself now part of the public record, preserved above.

prescribing and diagnosis in care systems: the treated-prevalence layer

What exists on prescribing and clinical diagnosis is claims- and EHR-based, and it all shares one construct that must be stated every time: treated prevalence. Every number below requires a coded diagnosis or a prescription fill inside a care system, and the transgender cohorts are themselves identified by ICD codes — so these studies describe clinically visible patients, not populations. Comparing them to the survey numbers above is a category error, and popular writeups make it constantly.

For sexual orientation and prescribing, the population-level evidence is not American: Swedish register data links survey-measured orientation to the national prescribed-drug register [17]. In the U.S., the nearest thing on the horizon is a sleeper: the Medicare Current Beneficiary Survey quietly added a sexual-orientation item in fall 2023 — it survives in the 2025 instrument — but it has not appeared in any public-use file. If it ever reaches the restricted files, it would be one data request away from linkable Part D antidepressant fills. Nothing of the sort is published today, and this page will not pretend otherwise.

the field's explanation — quoted, not asserted

This page draws no causal conclusion about why the disparities above exist. The research field has a dominant explanatory model, and the honest move is to show it to you as what it is — the field's model, with its citation — and let it sit beside the data. From the abstract of the paper that formalized it, Meyer's 2003 Psychological Bulletin review [18]:

“The author offers a conceptual framework for understanding this excess in prevalence of disorder in terms of minority stress—explaining that stigma, prejudice, and discrimination create a hostile and stressful social environment that causes mental health problems. The model describes stress processes, including the experience of prejudice events, expectations of rejection, hiding and concealing, internalized homophobia, and ameliorative coping processes.”— Meyer IH, Psychological Bulletin 2003;129(5), abstract [18]

The disparity estimates the model was built to explain are meta-analytic and consistent across four decades of studies:

• King M et al., BMC Psychiatry 2008;8:70. Lifetime suicide attempt in lesbian, gay and bisexual people: pooled risk ratio 2.47 (95% CI 1.87–3.28) vs heterosexual people; depression and anxiety disorders at least 1.5 times higher (RR range 1.54–2.58); lifetime suicide attempt in gay/bisexual men RR 4.28 (2.32–7.88).
• Marshal MP et al., Journal of Adolescent Health 2011;49(2):115–123. Sexual-minority youth: suicidality OR 2.92 vs heterosexual youth, with a severity gradient — ideation 1.96, plans 2.20, attempts 3.18, attempts requiring medical attention 4.17; depression effect size d = 0.33.
• Wittgens C et al., Acta Psychiatrica Scandinavica 2022;145(4):357–372. Population-based studies only (519,414 heterosexual; 10,178 lesbian/gay; 14,410 bisexual participants): odds ratios 1.97–2.89 for lesbian/gay adults and 2.70–4.81 for bisexual adults across depression, anxiety and suicidality — and no evidence the disparities shrank over time, except for alcohol-use disorder.
• Pellicane MJ, Ciesla JA, Clinical Psychology Review 2022;91:102113. Transgender and gender-diverse populations, 85 studies: distal stressors, expected rejection, internalized transphobia and identity concealment were each associated with depression, suicidal ideation and suicide attempts.

Note what the strongest recent meta-analysis (population-based studies only) found beyond the disparities themselves: no evidence they shrank over time, except for alcohol-use disorder [21]. And note the finding that repeats in every dataset on this page, from NHIS symptom frequency to NSDUH assessment to the meta-analyses: bisexual adults show the widest gaps, a pattern the field's bisexual-specific review confirms and popular coverage almost always flattens into a single “LGBT” number.

what this page does not show

• That any identity is a disorder. The classification systems settled this, in writing, decades ago for orientation and in 2019 for gender incongruence — see the receipts at the top.
• Why the disparities exist. The minority-stress model is the field's explanation, quoted above as such. This page asserts no cause in any direction.
• Diagnosed-prevalence disparities from federal tables. No such table exists; every federal figure here is symptom frequency, screened, assessed or treatment receipt, and is labeled.
• Suicide death rates by orientation or identity. Structurally impossible from current federal data. Attempts and ideation, from surveys, are what exist.
• That high treatment rates mean the system is working, or that disparities mean it is failing. The data shows bisexual adults with the most counseling, the most medication and the most cost-blocked care simultaneously. That is a description, not a verdict.
• Anything about you. Population statistics do not diagnose anyone, in either direction.

what to ask — turning this into agency

If something on this page is relevant to your care, the useful move is a better conversation, not a self-diagnosis or a discontinued prescription:

• “Is what I'm describing a diagnosis, a screen, or your clinical impression?” — the construct question, pointed at your own chart.
• “What would change your assessment?” — the question that distinguishes an assessment from a label.
• If you're LGBTQ+ and looking for care: ask a prospective clinician what experience they have with LGBTQ+ patients, and what their approach is. A clinician who treats that as a fair question is telling you something; so is one who doesn't.
• If cost is the barrier — the data above says that for bisexual adults especially it often is — ask directly about sliding-scale fees, community mental-health centers, and whether your state's marketplace plans cover telehealth therapy.
• If you take psychiatric medication: nothing here is a reason to stop. Changes belong in a conversation with your prescriber, including tapering plans if a change is ever warranted.