# Per-State Outcome Profile Narratives — Phase B

**Companion to**: `mh_gap_youth_outcomes_v1_article.md` §S3
**Source**: `analysis_outputs.mh_gap_youth_outcomes_v1`
**Sampling**: 5 framework-identified outliers (PR, NC, NJ positive; VT, AK negative) + 7 pre-specified comparator states to span demographic + outcome diversity (TX, CA, MA, WV, NM, MT, NY)

This is not the framework's randomized dartboard (the Youth V1 dartboard sampled 9 states by population-weighted residual stratification). This is a structured comparison set chosen to show the gap-vs-outcome relationship across the full range of state characteristics. Per-state narratives anchor the §5 Discussion's three-explanation argument for the non-protective supply signal.

---

## POSITIVE OUTLIERS — gap worse than uninsured rate predicts

### 1. Puerto Rico · z = +2.39 (strongest signal in dataset)

**Profile**: 39.2% sad/hopeless. 5.8 youth-serving providers per 100K under-18 — by far the lowest in the dataset. 2.5% uninsured under-19 (ASES program is comprehensive). YRBSS considered suicide 14.4%, made plan 11.9%, attempted 10.9%. All-age suicide AADR: not available (NCHS bi63-dtpu excludes territories). Drug overdose 2023: 26.1 per 100K.

**Reading**: The framework's strongest positive-outlier signal in the youth study. The post-Hurricane Maria health-system disruption + decades of federal Medicaid-payment cap policy left the territory with workforce density 50-100× below comparable states. YRBSS suicide indicators are *below national mean* — Puerto Rico's youth-distress measurement is high, but the suicide-attempt prevalence (10.9%) is actually slightly above national average (9.8%) and below states like Alaska (19.0%). The supply gap is the dominant signal; the mortality consequences are less extreme than the supply gap alone would suggest.

**Policy lever**: Federal — full Medicaid parity for the territory, Title V MCH grant expansion, IHS-style federal direct-employment for mental-health workforce in underserved territories.

### 2. North Carolina · z = +1.81

**Profile**: 39.1% sad/hopeless. 9.1 providers per 100K. 4.2% uninsured under-19. YRBSS considered 18.2%, made plan 15.9%, attempted 9.5%. All-age AADR: 14.3 per 100K (close to national mean of 16.7). Drug OD 2023: 37.3 per 100K.

**Reading**: NC has a substantial youth-distress prevalence, low uninsured rate, but the youth-serving workforce is thin (third-lowest density in our 35-state set). The all-age suicide AADR sits near the national mean and the YRBSS attempted-suicide rate is similarly near-mean. The framework flags NC as a workforce-build-out priority; the mortality data does not flag NC as a suicide-prevention crisis state.

**Policy lever**: CCBHC implementation has not been scaled in NC; the state did expand Medicaid in 2023. The framework's residual signal here predicts that workforce expansion is the most likely policy lever to close the gap.

### 3. New Jersey · z = +1.53 (most striking outcome divergence)

**Profile**: 36.3% sad/hopeless. 17.9 providers per 100K. 2.6% uninsured under-19 (near national floor). YRBSS considered 14.0%, made plan 11.1%, attempted **5.2%** (lowest in dartboard, well below national 9.8%). All-age AADR **8.3** (3rd lowest in country). Drug OD 30.2.

**Reading**: NJ is the cleanest demonstration of the framework's gap-vs-mortality divergence. The framework flags NJ as a positive outlier (supply worse than uninsured predicts). The mortality data shows NJ is one of the *lowest-suicide states in the country* (AADR 8.3 vs MT's 28.9). The framework's identification is correct (NJ has structural workforce supply lower than its insurance landscape implies); the mortality outcome is independently determined by demographic, urban, socioeconomic, and policy factors that do not align with the simple supply-vs-coverage residual.

The framework here surfaces the right *workforce-priority* state without claiming NJ is a *suicide-prevention-priority* state. The two are different.

---

## NEGATIVE OUTLIERS — gap better than uninsured rate predicts

### 4. Vermont · z = −1.70

**Profile**: 29.3% sad/hopeless (lowest in dartboard). 1,059 providers per 100K. 3.9% uninsured under-19. YRBSS considered: missing (n/a). Made plan 13.7%, attempted 7.4%. All-age AADR 18.3 (slightly above national mean). Drug OD 39.7.

**Reading**: VT shows the supply-policy success pattern at scale (UVM Medical Center pediatric behavioral-health + state Medicaid expansion + the Designated Agency system). The framework correctly identifies VT as a negative outlier — its youth-serving workforce is much larger than its insurance landscape alone would predict. YRBSS attempted is below national mean (7.4 vs 9.8); all-age AADR is moderate (18.3) — not the lowest in the country (NJ at 8.3 is lower) but lower than the high-AADR cluster (MT 28.9, AK 27.0).

VT illustrates that a successful policy stack produces measurable downstream effects (low YRBSS attempted), but not always a dominant signal at the all-age mortality scale — because all-age suicide is shaped by demographic factors (age structure, gun ownership, isolation, occupational stress) that VT's policy stack doesn't fully control.

### 5. Alaska · z = −1.69 (most striking gap-vs-mortality contradiction)

**Profile**: 43.2% sad/hopeless (above national average). 1,085 providers per 100K. 5.5% uninsured under-19 (slightly worse than national). YRBSS considered 22.6%, made plan 20.5%, attempted **19.0%** (highest in dartboard). All-age AADR **27.0** (3rd highest, after MT). Drug OD 43.4.

**Reading**: AK is the central counterexample for the "supply prevents harm" narrative. The framework correctly identifies AK as a negative outlier — its youth-serving workforce density is among the highest in the country, driven by IHS plus Alaska Native tribal health organizations' federally-employed mental-health workers. Yet AK has the highest YRBSS attempted-suicide rate (19.0% vs 9.8% national mean) and the third-highest all-age suicide AADR (27.0 vs 16.7 mean).

This is the policy-relevant non-result. Alaska has built workforce supply at scale through a federal channel (IHS) — and the suicide outcome has not improved. The drivers of Alaskan youth suicide (rural isolation, firearm access, intergenerational trauma in Indigenous communities, climate-change-driven cultural disruption, methamphetamine + alcohol exposure) operate on a different axis from workforce supply. The framework correctly identifies AK as well-supplied; the framework does not predict the mortality outcome.

This is the clearest demonstration of why the residual_z signal is not a suicide-risk metric.

---

## EXPECTED CASES (comparator states)

### 6. Texas · z = +0.70 (largest state, expected class)

**Profile**: 42.4% sad/hopeless. 16.9 providers per 100K (low). 8.9% uninsured under-19 (highest in dataset). YRBSS considered 21.1%, made plan 18.3%, attempted 12.3%. All-age AADR 13.4. Drug OD 19.0 (low).

**Reading**: TX is the largest state by under-18 population (7.5M) and sits within the expected residual band. The high uninsured rate predicts the supply gap; the suicide and drug OD rates are at-to-below the national mean despite the high distress prevalence. The reading is that Texas has a *high need* and *low supply* situation, but the regression model predicts most of the gap from the uninsured rate alone. Policy lever: both Medicaid expansion (to close the uninsured gap) and workforce build-out (to close the supply gap).

### 7. Massachusetts · z = −0.08 (expected, high-supply state)

**Profile**: 34.0% sad/hopeless (below mean). 212 providers per 100K. 1.2% uninsured under-19 (lowest in country — MA's near-universal coverage). YRBSS considered 15.8%, made plan 12.3%, attempted 7.2%. All-age AADR 9.5. Drug OD 36.1.

**Reading**: MA is what an insurance-saturated + university-medical-center state looks like at scale. The framework places MA in the expected band — its supply-coverage relationship is at the regression line. YRBSS attempted is below national mean (7.2 vs 9.8); all-age AADR is among the lowest (9.5). Drug OD is moderate-to-high (36.1) reflecting the East-Coast opioid corridor. This is a baseline expected case showing what insurance + supply + integrated care produces.

### 8. West Virginia · z = −0.21 (expected, drug-OD epicenter)

**Profile**: 43.8% sad/hopeless (highest in dartboard). 229 providers per 100K. 3.1% uninsured under-19 (low). YRBSS considered 24.8%. All-age AADR 21.1. Drug OD **79.1** (highest in country).

**Reading**: WV has the highest youth distress prevalence in the analysis and a moderately high all-age suicide rate (21.1) — but the dominant mortality story is drug overdose (79.1 per 100K). The framework places WV in the expected residual band (workforce supply matches insurance coverage roughly as predicted). The drug overdose epidemic in WV is driven by economic dislocation, OxyContin prescribing history, and supply-side drug availability — none of which the framework's gap measure captures. This is the canonical "deaths of despair" state.

### 9. New Mexico · z = +0.08 (expected, high-AADR state)

**Profile**: 36.2% sad/hopeless. 74 providers per 100K. 5.5% uninsured under-19. YRBSS considered 15.1%, attempted 8.5%. All-age AADR **23.3** (4th highest). Drug OD 48.0.

**Reading**: NM has very high all-age suicide AADR (23.3) tied to the rural Mountain West suicide pattern. The framework places NM in the expected residual band. The high mortality reflects geographic, demographic, and historical factors (Indigenous population, firearm access, rural isolation) that the supply-vs-uninsured residual does not predict.

### 10. Montana · z = −0.84 (highest all-age AADR in dataset)

**Profile**: 43.3% sad/hopeless. 263 providers per 100K. 6.5% uninsured under-19. YRBSS considered 26.1% (highest in dartboard), made plan 21.4%, attempted 11.3%. All-age AADR **28.9** (highest in dataset). Drug OD 15.4 (low).

**Reading**: MT has the highest all-age suicide AADR (28.9) and the highest YRBSS considered-suicide rate. The framework places MT in the expected residual band; its supply density (263 per 100K) is moderate for a small Western state. The drivers of MT's suicide epidemic are well-documented: rural isolation, firearm prevalence, ranching/agriculture economic stress, alcohol exposure, lack of crisis-response infrastructure outside Helena/Billings.

MT illustrates that *the suicide-prevention-priority list* (MT, AK, WY, NM, ID) and *the workforce-build-out-priority list* (PR, NC, NJ) are different lists. The framework correctly does not flag MT as a positive outlier on workforce gap — and would be wrong to use as a state-level suicide-prevention guide.

---

## What this collection of profiles says, taken together

The 5 framework-identified outliers fall into two clean structural patterns:

- **Positive outliers (PR, NC, NJ)**: workforce undersupplied for the insurance landscape. Mortality outcomes range from below-mean (NJ) to slightly-above-mean (NC). Policy lever: workforce capacity expansion.
- **Negative outliers (VT, AK)**: workforce overbuilt for the insurance landscape. Mortality outcomes range from moderate (VT) to highest-in-class (AK). VT's policy stack works; AK's high supply does not overcome the underlying demographic, firearm, and cultural drivers.

The 5 expected-class states (TX, MA, WV, NM, MT) show the full range of mortality outcomes (from MA's 9.5 AADR to MT's 28.9) while sitting near the regression line on the supply-vs-uninsured relationship.

**The clean takeaway**: the framework is well-calibrated for identifying workforce-build-out priorities given the insurance landscape. It is not a mortality-incidence predictor at state granularity. Policymakers should use the framework's outputs to triage *workforce capacity programs* and use independent mortality data (NCHS, CDC WONDER) to triage *suicide-prevention programs*. The two lists barely overlap, and conflating them mismatches the intervention to the problem.