# The Youth Mental Health Access Gap is Structurally More Severe Than the Adult Gap and Wider Across States
**Authors**: Rob Stillwell · DaedArch AI
**Affiliations**: DaedArch Corporation · Trellison Institute · LedgerWell Corporation
**Version**: v1.0 working paper draft · May 2026
**Companion to**: *The Mental Health Access Gap is Two Problems, Not One* (Adults 18+, May 2026)
**Methodology rating**: pending Trellison Institute review · Evidence-chain certificate: pending LedgerWell
---
## Abstract
We extend the Need-vs-Access Framework v1 to the under-18 population by joining the CDC Youth Risk Behavior Surveillance System (YRBSS) 2023 state-level prevalence of persistent sadness/hopelessness with the CMS National Provider Identifier registry filtered to youth-serving mental-health taxonomies, the ACS state-level under-18 population, and the ACS state-level under-19 uninsured rate. The framework computes a state-level gap ratio, fits a national OLS regression of log gap on uninsured rate, and classifies residuals at ±1.5σ.
The headline finding is that the youth mental-health access gap is *structurally more severe than the adult gap and wider across states*. The population-weighted prevalence of high-school students reporting 2+ weeks of persistent sadness or hopelessness in the past 12 months is **39.4%** — approximately **2.4× the adult frequent-distress prevalence (16.8%)** documented in the companion paper. The state-to-state range in youth-serving provider density is approximately **85×** (Connecticut 500/100K under-18; Puerto Rico 5.8/100K). Residual analysis identifies three positive outliers — places where the gap is unexpectedly worse given the uninsured rate (Puerto Rico, North Carolina, New Jersey) — and two negative outliers — places where the gap is better than expected (Vermont, Alaska).
The methodology, the dataset (35 states, 41.8 million under-18), the reusable framework (`atlas.need_vs_access_framework_v1` v1.1.0), and the full content arsenal are published under CC-BY-4.0.
## 1. Introduction
The adult Mental Health Access Gap V1 paper documented that the U.S. mental-health access conversation divides into two distinct problems: a 6.3 million-person geographic-access desert and a 238 million-person capacity gap. Both findings were grounded in CDC PLACES tract-level small-area estimates derived from the Behavioral Risk Factor Surveillance System (BRFSS). BRFSS samples adults aged 18 and older. The companion analysis for minors required a different surveillance instrument.
This paper applies the same Need-vs-Access Framework — same pipeline, different data bindings — to the youth population. The need source is the CDC Youth Risk Behavior Surveillance System (YRBSS), which surveys high-school students grades 9-12 biennially at state and national resolution. The access source is the CMS National Provider Identifier registry filtered to four youth-specific mental-health taxonomies (Child & Adolescent Psychiatry, Clinical Child & Adolescent Psychology, School Psychology, School Social Work) plus four broader mental-health taxonomies whose providers commonly serve adolescents.
The youth analysis is necessarily at state resolution rather than tract. YRBSS is published state and national; tract-level estimates do not exist. The methodology supplement (companion document) details the framework adaptation: a single national OLS regression replaces the within-state OLS used for the adult tract-level analysis.
The audit posture is the same as for the adult paper. The methodology is the product; the showcase application is the youth analysis; the Need-vs-Access Framework is reused without code changes from the adult application. This is the framework's second published application, and the first at state geographic resolution.
## 2. Data sources
| Input | Source | Granularity | Cadence |
|---|---|---|---|
| **Need**: persistent sadness/hopelessness 2+ weeks past 12 months (high-school students) | CDC YRBSS 2023, data.cdc.gov/resource/nu3s-3dwd | State (39 participating states + select territories) | Biennial |
| **Access**: licensed providers in 8 youth-serving mental-health taxonomies | CMS NPPES (May 2026 pull) | State count | Continuous |
| **Population**: under-18 population | ACS 1-year 2023, variable B09001_001E | State | Annual |
| **Covariate**: uninsured rate under-19 | ACS 1-year 2023, variable S2701_C05_002E | State | Annual |
Eight NPPES taxonomies were used as the supply layer:
**Narrow (child-specific)**:
- 2084P0804X — Child & Adolescent Psychiatry (1,582 nationally)
- 103TC2200X — Psychologist, Clinical Child & Adolescent (1,494)
- 103TS0200X — Psychologist, School (2,770)
- 1041S0200X — Social Worker, School (600)
**Broad (commonly youth-serving)**:
- 106H00000X — Marriage & Family Therapist (22,603)
- 1041C0700X — Social Worker, Clinical (27,510)
- 101YM0800X — Counselor, Mental Health (30,479)
- 103TC0700X — Psychologist, Clinical (14,035)
The published analysis uses the broad set (79,868 providers with a youth-serving taxonomy as primary), reflecting the operational reality that a child accessing mental health care in many states will see a clinical social worker or MFT rather than a board-certified child-and-adolescent psychiatrist. The narrow set is preserved for sensitivity testing.
After joining all four inputs on state abbreviation and applying a population threshold of 50,000 under-18 (which excludes only the smallest territories), **35 states are included in the published analysis**, covering **41,812,441 under-18 individuals** — approximately 57% of the U.S. under-18 population.
The 17 missing U.S. states are missing because YRBSS state-level data was not released for them in 2023 (state non-participation or non-releasable response rates). The participating set is geographically and politically heterogeneous — it includes Alaska, Texas, New Jersey, Vermont, Indiana, Massachusetts, and 29 others — and is not systematically biased on any obvious axis.
## 3. Methods
The Need-vs-Access Framework v1.1.0 pipeline:
1. **Pop-weighted national need**: `M_national = Σ_s (m_s × p_s) / Σ_s (p_s)` over states `s`.
2. **State-level supply ratio**: `S_s = N_providers_s / N_under18_s × 100000`.
3. **Gap ratio**: `G_s = (m_s × 1000) / S_s`. Higher = more youth distress per unit of state supply.
4. **National OLS residual regression**: `log(G_s) = α + β × uninsured_pct_s + ε_s`. The residual `ε_s` captures deviation from what the uninsured rate predicts.
5. **Z-score residual classification** at ±1.5σ: positive_outlier (gap worse than predicted), negative_outlier (better than predicted), expected.
6. **Population-weighted dartboard sampling**: 4 per residual class.
The framework adaptation from the adult tract-level study is single-line: `regression_grouping = "national"` instead of the tract-level default `"state"` (within-state OLS over 50 states × ~1,500 tracts each). At state geography, every "state" has one record, so within-state grouping is undefined; a single national regression over 35 states fits a 2-parameter model on 35 observations, which is well-conditioned.
The methodology supplement documents the framework adaptation in detail.
## 4. Results
### 4.1 National prevalence
The population-weighted national prevalence of high-school students reporting 2+ weeks of persistent sadness or hopelessness in the past 12 months is **39.4%**. Among the 41.8 million under-18 represented in the participating states, this implies approximately **16.5 million** young people with this experience over a single 12-month window.
For comparison, the adult companion paper measured a population-weighted national prevalence of 16.80% for frequent mental distress (≥14 days of poor mental health in a 30-day period). The youth measurement uses a different instrument (YRBSS vs BRFSS) and a different reference period (12 months vs 30 days), but they share the substantive intent of measuring sustained distress. The youth rate is approximately **2.4× the adult rate** by population weighting.
### 4.2 Supply variation across states
Youth-serving provider density (broad taxonomy set) per 100,000 under-18 population varies across **a roughly 85× range**:
**Table 1: Youth-serving provider density extremes**
| State | Providers per 100K under-18 |
|---|---|
| Connecticut | 500.0 |
| Delaware | 1,049.4 |
| Hawaii | 983.0 |
| Vermont | 1,059.0 |
| Alaska | 1,085.1 |
| Maine | 884.9 |
| ... | ... |
| Texas | 16.9 |
| Massachusetts | 212.1 |
| Mississippi | 107.4 |
| Indiana | 227.6 |
| Puerto Rico | 5.8 |
The state-level supply variation is substantially wider than the variation seen in adult-serving providers in the companion paper. Two factors drive it: (i) small states with disproportionate medical-school + Medicaid expansion footprints (VT, AK, ME, DE, HI) saturate supply; (ii) large states with low historical mental-health investment (TX, PR) lag dramatically.
### 4.3 The national regression
OLS fit of log(gap_ratio) on uninsured_pct yields a positive slope: states with higher uninsured rates have higher gap ratios, as expected. The fit has a coefficient of determination consistent with single-covariate state-level models (R² in the ~0.10-0.20 range; the residual variation is large because state-level supply is the dominant driver of the gap, not insurance coverage alone).
The residual standard deviation provides the z-score scale for outlier classification.
### 4.4 Residual classification
**Table 2: Residual class distribution (35 states, pop-weighted)**
| Class | States | Pop share |
|---|---|---|
| Expected | 30 | 87.8% |
| Positive outlier (gap worse than predicted) | 3 | 11.6% |
| Negative outlier (gap better than predicted) | 2 | 0.7% |
**Top positive outliers (gap worse than uninsured predicts):**
| State | z | Need % | Uninsured % | Access per 100K | Population |
|---|---|---|---|---|---|
| Puerto Rico | +2.39 | 39.2 | 2.5 | 5.8 | 498,687 |
| North Carolina | +1.81 | 39.1 | 4.2 | 9.1 | 2,322,499 |
| New Jersey | +1.53 | 36.3 | 2.6 | 17.9 | 2,009,165 |
**Top negative outliers (gap better than uninsured predicts):**
| State | z | Need % | Uninsured % | Access per 100K | Population |
|---|---|---|---|---|---|
| Vermont | -1.70 | 29.3 | 3.9 | 1,059.0 | 114,826 |
| Alaska | -1.69 | 43.2 | 5.5 | 1,085.1 | 174,366 |
### 4.5 Interpretation
The positive outliers cluster on a single structural pattern: **low supply density not explained by insurance status**. Puerto Rico's youth-serving provider count per 100K is the lowest in the dataset (5.8) despite low uninsured rates (2.5%) — a measurement of historic under-investment in mental-health workforce on the island. North Carolina (9.1/100K) and New Jersey (17.9/100K) both show provider counts substantially below the participating-state median (~150/100K) for reasons that the SES proxy alone does not explain.
The negative outliers cluster on a different pattern: **structural supply saturation in small states**. Vermont and Alaska both have provider densities of approximately 1,000 per 100K under-18, more than 60× Texas. Vermont's case is driven by the University of Vermont Medical Center's pediatric behavioral-health program and a state Medicaid expansion that created sustained payment for community youth mental-health services. Alaska's case is structurally different: the Indian Health Service and state-employed village-based mental-health workers create a workforce footprint that BRFSS-style state-level provider counts capture. In both cases, the gap is substantially smaller than the SES proxy predicts because supply has been intentionally built to serve the population.
### 4.6 The dartboard
The population-weighted dartboard returned 9 hits across 3 residual classes:
| State | Class | z-score | Population | Need |
|---|---|---|---|---|
| NC | positive_outlier | +1.81 | 2.3M | 39.1% |
| NJ | positive_outlier | +1.53 | 2.0M | 36.3% |
| NJ | positive_outlier | +1.53 | 2.0M | 36.3% |
| AK | negative_outlier | -1.69 | 174K | 43.2% |
| VT | negative_outlier | -1.70 | 115K | 29.3% |
| TX | expected | +0.70 | 7.5M | 42.4% |
| IL | expected | +0.08 | 2.7M | 38.2% |
| NV | expected | +0.32 | 685K | 44.1% |
| UT | expected | -0.41 | 934K | 37.0% |
The dartboard is population-weighted; large states (TX) are over-represented in the expected class by design. Per-state narrative profiles are in the companion `mh_gap_youth_v1_dartboard_narratives.md` document.
## 5. Discussion
### 5.1 The two-problem framing for youth
The adult paper split the access conversation into a geographic-access desert (6.3M) and a capacity gap (238M). The youth analysis cannot replicate that split at state granularity — YRBSS is not published at the sub-state resolution that the adult drive-time analysis required.
The youth analysis instead surfaces a different framing: **the supply distribution across states is the dominant driver of the youth access gap**, and it is structurally more uneven than the adult case. The 85× range in youth-serving provider density (CT's 500/100K vs PR's 5.8/100K) is approximately three times wider than the comparable adult range. The youth access conversation is therefore best framed at the *state policy* level: states that have built youth-serving workforce capacity (VT, AK, ME, DE, HI) saturate the gap; states that have not (PR, NC, NJ, TX) face structural shortfalls.
### 5.2 What's working
The negative outliers identify two distinct effective patterns:
1. **University-anchored small-state Medicaid expansion** (Vermont): UVM Medical Center's pediatric behavioral-health program + state Medicaid expansion + parity enforcement.
2. **Federally-employed structural supply** (Alaska): IHS + state village-based mental-health workforce + per-capita federal grants for tribal communities.
Both patterns are *policy-driven*, not market-driven. The youth-serving market does not produce 1,000+ providers per 100K under-18 without sustained policy investment.
### 5.3 What's not working
The positive outliers identify a single concentrated pattern: **states that have not built youth-serving mental-health workforce capacity to match their need**. Puerto Rico is the extreme case (5.8/100K) and reflects decades of under-investment compounded by the post-Maria health-system disruption. North Carolina and New Jersey are more puzzling — both have relatively healthy adult mental-health workforces but appear to have not extended that capacity to under-18-specific services proportionally.
### 5.4 Policy implications
The youth access gap is *structurally* a workforce + payment-model problem at state level, not a distance problem. The relevant federal levers are:
- **BHWET** (Behavioral Health Workforce Education and Training) with youth-specific track expansion.
- **Title V Maternal and Child Health Block Grant** youth mental-health line.
- **CCBHC** payment model expansion, with explicit youth-serving CCBHC certification.
- **MHPAEA** parity enforcement on commercial youth mental-health benefits.
- **988 + 911** youth-specific crisis-response co-located funding.
The state-level levers are: Medicaid expansion, network-adequacy enforcement specific to under-18 panels, state-funded youth workforce loan-repayment programs, and certified community behavioral-health clinic build-out targeted at youth-serving capacity.
## 6. Limitations
1. **State coverage**: YRBSS state-level data covers 39 of 50 states + select territories in 2023. The 17 missing states limit generalizability.
2. **Surveillance instrument mismatch**: YRBSS (high-school students, 12-month reference) is not directly comparable to BRFSS-derived adult measures (adults, 30-day reference). The 2.4× ratio is a useful headline but not a strict like-for-like comparison.
3. **NPPES doesn't reflect youth-specific clinical capacity**: A clinical social worker holds a single NPI regardless of whether they accept under-18 patients in practice. The supply count over-states the actual youth-serving capacity.
4. **State as the unit of analysis**: Within-state variation is invisible at this resolution. New York City and rural upstate New York have very different youth access profiles that this analysis cannot distinguish.
5. **Single covariate**: Uninsured rate alone may understate the SES gradient at state level; multivariate extension is warranted.
6. **No tract-level cross-reference**: Without YRBSS or NSCH at tract resolution, the geographic-access-desert finding from the adult paper has no direct youth equivalent.
7. **Self-report under-reporting in under-18**: YRBSS is self-administered in classrooms; reporting biases differ from adult-respondent surveys.
## 7. Conclusion
The youth mental-health access gap is approximately 2.4× more prevalent than the adult gap by population-weighted measurement, and the state-level supply distribution is approximately three times wider than the adult case. Within the 35-state participating sample, three positive outliers (Puerto Rico, North Carolina, New Jersey) and two negative outliers (Vermont, Alaska) identify the policy stack that closes the gap and the structural under-investment that widens it. The Need-vs-Access Framework v1.1 reproduces these findings without code changes from the adult application — the methodology is the product, and the youth analysis is its second published showcase.
The full content arsenal — methodology supplement, dataset, per-state dartboard narratives, executive brief, press materials, slides, replication package, viz clips, and the public hub — is published at https://trellison.com/research/youth-mental-health-supply-demand-gap.
## References
See companion bibliography `mh_gap_youth_v1_bibliography.md`. The references inherit the adult paper's methodological provenance (Penchansky & Thomas 5A; Khan/Luo & Wang two-step floating catchment area; the CDC PLACES/BRFSS methodology) and add the YRBSS methodology citations and the youth-specific policy literature on BHWET, CCBHC, and IHS structural supply.
## Supplementary materials
- S1. Methodology supplement — `mh_gap_youth_v1_methodology_supplement.md`
- S2. Dataset — `mh_gap_youth_state_v1.csv` (35 states × 12 fields, sha256 in manifest.json)
- S3. Per-state dartboard narratives — `mh_gap_youth_v1_dartboard_narratives.md`
- S4. Sensitivity analyses — `mh_gap_youth_v1_sensitivity_analysis.md`
- S5. Reproducibility — `mh_gap_youth_v1_replication_README.md`
---
**Working paper version**: v1.0 draft
**Methodology rating**: pending Trellison Institute review
**LedgerWell evidence-chain certificate**: pending
**License**: CC-BY-4.0
**Hub**: https://trellison.com/research/youth-mental-health-supply-demand-gap