# The Framework Predicts Capacity Gaps, Not Mortality — Executive Brief
**A Trellison Institute Working Paper — Executive Brief**
*May 2026 · 2-page summary of `mh_gap_youth_outcomes_v1_article.md`*
*Phase B follow-up to Youth Mental Health Access Gap V1 (May 2026)*
---
## The finding in one line
The Need-vs-Access Framework's gap measures *strongly* predict within-instrument youth suicide ideation and attempts (r > +0.75), *modestly* predict state all-age suicide AADR (r ≈ +0.34), and *do not* predict mortality at state level via the supply-side measure or the residual classification.
## The numbers
**Strong: within-YRBSS coupling** (n = 30-34 states):
| Need (sad/hopeless 2+ wks) × | r |
|---|---:|
| YRBSS considered suicide | **+0.817** |
| YRBSS made plan | **+0.800** |
| YRBSS attempted suicide | **+0.762** |
**Moderate: across-instrument** (n = 34-35):
| Need × | r |
|---|---:|
| All-age suicide AADR (NCHS 2017) | +0.338 |
| Drug overdose deaths per 100K (2023) | +0.192 |
**Null: supply-side metric** (n = 32-35):
| `access_value` (providers per 100K under-18) × | r |
|---|---:|
| YRBSS considered | −0.030 |
| YRBSS made plan | +0.000 |
| YRBSS attempted | +0.109 |
| All-age AADR | +0.143 |
| Drug OD | **+0.211** |
**Null with negative direction: framework residual** (n = 34):
| `residual_z` × | r |
|---|---:|
| All-age AADR | −0.296 |
| Drug OD | −0.045 |
## What this means in plain English
1. **The framework's need metric is validated.** State-level sad/hopeless prevalence is a strong cross-sectional proxy for state-level youth suicide ideation and attempts. The Youth V1 study chose the right need signal.
2. **State-level provider supply does not predict reduced mortality.** Counting NPI-registered youth-serving providers per 100K under-18 — the closest a public researcher can come to state-level supply measurement — is uncorrelated with state suicide AADR and modestly *positively* correlated with state drug-OD rate. **The "more providers prevents more deaths" expectation does not hold at state cross-section.**
3. **The framework's residual signal is orthogonal to outcomes.** States flagged as positive outliers (PR, NC, NJ — supply worse than uninsured predicts) actually have *lower* mortality than negative outliers (VT, AK). The residual measure captures a policy-relevant capacity gap given the insurance landscape — not a mortality-incidence signal.
4. **Drug overdose is a different question.** Higher-coverage states have *higher* drug OD rates. The "deaths of despair" geography (WV, NM, OH, DC) doesn't align with mental-health-care access in the way suicide does. Drug overdose mortality requires a separate analytical pipeline.
## What this means for policy
| Question | Use which framework output |
|---|---|
| Which states should build more youth-serving workforce capacity? | `residual_class` = positive_outlier (PR, NC, NJ) — these states have undersupplied workforce given their insurance coverage |
| Which states have the highest youth suicide ideation? | High `need_value` (IN 47%, AR 45.1%, OK 44.9% — see Youth V1 paper) |
| Which states should be prioritized for suicide-prevention intervention? | High all-age AADR (MT 28.9, AK 27.0, WY 26.9 — *not* the same list as the workforce-priority list) |
| Which states show the policy stack that works for capacity? | `residual_class` = negative_outlier (VT, AK) — UVM Medical Center + Medicaid expansion + Designated Agencies (VT) or IHS + tribal health organizations (AK) |
The framework surfaces the *capacity-build-out priority*. It is not a suicide-prevention triage tool. Treating it as one would mismatch the intervention to the problem.
## Outlier state outcome profiles
| State | gap class | sad/hopeless | access/100K | YRBSS attempted | All-age AADR | Drug OD/100K |
|---|---|---:|---:|---:|---:|---:|
| AK | negative_outlier | 43.2% | 1,085 | **19.0%** | **27.0** | 43.4 |
| VT | negative_outlier | 29.3 | 1,059 | 7.4 | 18.3 | 39.7 |
| NJ | positive_outlier | 36.3 | 17.9 | **5.2** | **8.3** | 30.2 |
| NC | positive_outlier | 39.1 | 9.1 | 9.5 | 14.3 | 37.3 |
| PR | positive_outlier | 39.2 | 5.8 | 10.9 | n/a | 26.1 |
| TX | expected | 42.4 | 16.9 | 12.3 | 13.4 | 19.0 |
| MA | expected | 34.0 | 212 | 7.2 | 9.5 | 36.1 |
| WV | expected | 43.8 | 229 | n/a | 21.1 | **79.1** |
| MT | expected | 43.3 | 263 | 11.3 | 28.9 | 15.4 |
## Method, in one paragraph
Pearson correlation between the five Need-vs-Access Framework gap measures (need, covariate, access, log_gap, residual_z) and five outcome measures (3 YRBSS state-level youth suicide indicators + NCHS all-age suicide AADR 2017 + NCHS VSRR drug OD 2023). Sources joined on state 2-letter abbreviation. Sample size 24-35 states depending on outcome (YRBSS state participation varies by question). Drug OD rate computed as 12-month-ending deaths / ACS 2023 state population × 100K. No multivariate adjustment in v1.0; multivariate residual analysis planned for v1.1.
## Limitations (8 in detail in the working paper)
1. Cross-sectional design; no causal identification.
2. AADR is 2017 (6-year lag vs 2023 gap measures).
3. All-age mortality; youth-specific (10-19) requires CDC WONDER XML POST + license acceptance.
4. YRBSS state coverage limited (35 of 50 states + territories in v1.0).
5. NPPES counts structure not capacity.
6. Cross-state confounding (rural geography, opioid-exposure history, demographic composition) not adjusted.
7. No FBI juvenile arrest data (API key pending).
8. No HCUP ER visit data (licensed access required).
## What's next
The framework's value proposition — surface where states have under-built youth-serving workforce capacity relative to their insurance landscape — is preserved by this finding. What is qualified is the claim that the framework's outputs are directly outcome-predictive. The two claims are different, and the published Phase B paper draws the distinction explicitly.
v1.1 work queued: FBI juvenile arrest correlation, CDC WONDER youth-specific mortality, multivariate residual analysis. Each is a turn of work that extends the analysis but does not invalidate the v1.0 finding.
## Reference
Full working paper: `docs/mh_gap_youth_outcomes_v1_article.md` (~3,500 words, 8 sections + supplementary).
Dataset: `analysis_outputs.mh_gap_youth_outcomes_v1` (35 states); CSV at `s3://daedarch-public-media/youth_mental_health_outcomes/dataset_v1/mh_gap_youth_outcomes_v1_state.csv`.
Hub: https://trellison.com/research/youth-mental-health-outcomes-correlation (under construction).
**Methodology rating**: Trellison Institute, pending review.
**Evidence chain certificate**: LedgerWell, pending.
**Working paper version**: v1.0 draft.