diff --git a/inbox/archive/health/2025-03-xx-frontiers-nutrition-glp1-nutrient-intake-crosssectional.md b/inbox/archive/health/2025-03-xx-frontiers-nutrition-glp1-nutrient-intake-crosssectional.md
new file mode 100644
index 000000000..7ebf745e1
--- /dev/null
+++ b/inbox/archive/health/2025-03-xx-frontiers-nutrition-glp1-nutrient-intake-crosssectional.md
@@ -0,0 +1,57 @@
+---
+type: source
+title: "Frontiers in Nutrition 2025: Cross-Sectional Study of GLP-1 Users — Near-Universal Vitamin D Shortfall, 64% Iron-Deficient, 72% Calcium-Deficient"
+author: "Frontiers in Nutrition (10.3389/fnut.2025.1566498)"
+url: https://www.frontiersin.org/journals/nutrition/articles/10.3389/fnut.2025.1566498/full
+date: 2025-03-01
+domain: health
+secondary_domains: []
+format: research-paper
+status: unprocessed
+priority: medium
+tags: [GLP-1, nutrition, micronutrients, vitamin-D, iron, calcium, protein, cross-sectional, DRI, dietary-reference-intake]
+---
+
+## Content
+
+Cross-sectional study examining nutrient intake during GLP-1 receptor agonist use.
+
+**Study design:**
+- n = 69 participants (adults using GLP-1RA for at least 1 month)
+- Participants completed 3-day food records + online survey questionnaires
+- Compared intake against Dietary Reference Intakes (DRI)
+
+**Key findings:**
+- **Vitamin D**: Only 1.4% of participants met 100% of the DRI. Mean intake 4 μg/day vs. national average of 19 μg/day — 79% below national baseline.
+- **Iron**: 64% consumed below the Estimated Average Requirement (EAR); highest prevalence among women and individuals undergoing aggressive caloric restriction.
+- **Calcium**: 72% consumed below the RDA.
+- **Protein**: 58% did not meet recommended targets (1.2–1.6 g/kg/day during weight loss per multi-society advisory).
+
+**Bottom line stated by authors:** "Participants on a GLP-1RA are not meeting the Dietary Reference Intakes for several vital nutrients through their diet."
+
+**Limitation:** Small sample (n=69), self-selected, cross-sectional design. Not representative of Medicaid or food-insecure populations — likely skews toward commercially insured, internet-accessible patients. No control group.
+
+## Agent Notes
+
+**Why this matters:** Primary data study (vs. cohort database claims study) with dietary record methodology. The 1.4% vitamin D DRI compliance figure is from this study and is the most striking specific datum in the GLP-1 nutritional literature. Despite the small n, the convergence with Urbina 2026 (n=480,825) gives confidence this isn't a sample artifact.
+
+**What surprised me:** The 1.4% vitamin D DRI compliance. This is not a marginal shortfall — it means 98.6% of GLP-1 users in this sample were not meeting even the recommended dietary intake for vitamin D, a nutrient already deficient in ~40% of the general US population.
+
+**What I expected but didn't find:** Any stratification by food security status. The study participants likely have commercial insurance and internet access (required to complete online survey). This means the deficiency rates found here may be UNDERESTIMATES for food-insecure populations, who start from a worse nutritional baseline.
+
+**KB connections:**
+- Consistent with and supportive of Urbina 2026 narrative review (`2026-01-xx-urbina-clinical-obesity-glp1-micronutrient-narrative-review.md`)
+- The 1.4% vitamin D DRI figure is specifically useful for claim writing — it's a concrete data point
+
+**Extraction hints:**
+- Use as supporting evidence for the broader nutritional deficiency claim, not as a standalone claim
+- The 1.4% vitamin D DRI compliance is the single most quotable datum from this source
+- Note sample limitation: n=69, likely commercially insured, online-accessible patients
+
+**Context:** Frontiers in Nutrition is a peer-reviewed open-access journal. Study methodology (3-day food record) is considered more reliable than dietary recall alone but has known limitations (underreporting, short capture window).
+
+## Curator Notes (structured handoff for extractor)
+
+PRIMARY CONNECTION: `2026-01-xx-urbina-clinical-obesity-glp1-micronutrient-narrative-review.md` (supporting data point)
+WHY ARCHIVED: The 1.4% vitamin D DRI compliance figure from dietary records is the most concrete datum for the nutritional deficiency claim. Small study but converges with larger systematic evidence.
+EXTRACTION HINT: Use as supporting evidence, not primary source. Archive for the 1.4% vitamin D figure specifically.
diff --git a/inbox/archive/health/2025-08-xx-lancet-preserving-clinical-skills-ai-deskilling.md b/inbox/archive/health/2025-08-xx-lancet-preserving-clinical-skills-ai-deskilling.md
new file mode 100644
index 000000000..242613786
--- /dev/null
+++ b/inbox/archive/health/2025-08-xx-lancet-preserving-clinical-skills-ai-deskilling.md
@@ -0,0 +1,63 @@
+---
+type: source
+title: "Lancet: Preserving Clinical Skills in the Age of AI Assistance — Mainstream Editorial on Colonoscopy Deskilling and Never-Skilling"
+author: "The Lancet (PIIS0140-6736(25)02075-6)"
+url: https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(25)02075-6/abstract
+date: 2025-08-01
+domain: health
+secondary_domains: [ai-alignment]
+format: editorial
+status: unprocessed
+priority: high
+tags: [clinical-AI, deskilling, never-skilling, medical-education, colonoscopy, physician-training, AI-safety, Lancet]
+flagged_for_theseus: ["Lancet editorial brings never-skilling into mainstream medicine discourse — same failure mode as Theseus's capability degradation concerns in human-AI systems"]
+---
+
+## Content
+
+The Lancet editorial "Preserving clinical skills in the age of AI assistance" (2025) documents and synthesizes the deskilling evidence emerging from clinical AI deployment, with specific focus on the colonoscopy observational study finding.
+
+**Core clinical finding referenced:**
+An observational study published contemporaneously found that experienced colonoscopists lost proficiency in colon polyp detection when routine AI support was switched off. After endoscopists had been using AI for three months, their unassisted adenoma detection rate (ADR) fell from 28% to 22% — a 22% relative reduction in unassisted detection capability.
+
+**Three-pathway taxonomy adopted by Lancet editorial:**
+- **Deskilling**: existing expertise lost through disuse (the colonoscopy finding)
+- **Mis-skilling**: AI errors adopted as correct clinical patterns
+- **Never-skilling**: foundational competence never acquired because AI precedes skill development in training
+
+**Editorial's framing:** As AI assumes a growing role in clinical practice, concern is mounting that off-loading clinical tasks and reasoning will lead to loss of skills (deskilling), adopting errors or bias from AI (mis-skilling), or failure to achieve competence (never-skilling).
+
+**Key problem identified:** Medical schools and postgraduate clinical training programs have been slow to integrate AI education into curricula. Most medical students lack understanding of the basic technical principles underlying AI. Medical education accreditation standards typically exclude AI competencies.
+
+**What the editorial does NOT provide:** Specific intervention protocols at scale. The editorial raises the alarm as a "design question" without empirically validated mitigation programs. Proposed measures (AI-off drills, pre-AI competency baselines, structured assessment before AI output review) exist as prescriptions, not validated implementations.
+
+**STAT News coverage (August 12, 2025):** "As AI spreads through health care, is the technology degrading providers' skills?" — mainstream media confirmation that the finding crossed from academic to public health discourse.
+
+**Mainstream acknowledgment significance:** The Lancet is the world's most read general medical journal. Publication of this editorial signals that the deskilling concern has moved from speculative/academic to mainstream clinical concern.
+
+## Agent Notes
+
+**Why this matters:** The Springer AI Review already documented the three-pathway model (archived `2025-08-xx-springer-clinical-ai-deskilling-misskilling-neverskilling-mixed-method-review.md`). What's different here is the institutional weight: The Lancet editorial converts the academic taxonomy into a mainstream clinical and educational policy concern. This is the never-skilling claim's "crossing the Rubicon" moment — from research literature to institutional acknowledgment.
+
+**What surprised me:** The editorial raises the alarm WITHOUT providing specific validated interventions. The world's most prestigious medical journal is publishing "we have a serious problem" without "here is the evidence-based solution." This is unusual for Lancet editorials, which typically accompany research papers with clinical guidance. The absence of prescriptive mitigation suggests the field genuinely doesn't know yet how to solve this at scale.
+
+**What I expected but didn't find:** Any health system or medical school reporting a systematic "AI-off drill" program with outcomes data. The mitigation proposals remain prescriptive, not empirical. The never-skilling detection problem (no baseline to compare against) remains unsolved — no medical school is running prospective competency assessments before AI exposure.
+
+**KB connections:**
+- Extends existing archive `2025-08-xx-springer-clinical-ai-deskilling-misskilling-neverskilling-mixed-method-review.md` — the Lancet adds institutional weight and the specific colonoscopy ADR finding
+- Supports existing KB claim: [[human-in-the-loop clinical AI degrades to worse-than-AI-alone because physicians both de-skill from reliance and introduce errors when overriding correct outputs]]
+- The never-skilling concept is NOT yet in KB claims — claim candidate still pending extraction
+- FLAG @Theseus: The Lancet editorial's structure (we know the problem, we don't know the solution at scale) parallels alignment concerns about human capability degradation in AI-dominated domains. Never-skilling is the clinical training manifestation of a broader capability degradation problem.
+
+**Extraction hints:**
+- Primary: extend/update the existing deskilling claim to include three-pathway taxonomy
+- Secondary: write a specific "never-skilling" claim: "Never-skilling in clinical AI is structurally invisible because it lacks a pre-AI baseline for comparison, requiring prospective competency assessment before AI exposure to detect — and no current training institution runs this assessment at scale"
+- Tertiary: the "Lancet acknowledgment without solution" is itself notable — the mainstream is aware of the problem but has no validated intervention. This is a different quality of concern than "academic debate."
+
+**Context:** The Lancet editorial is not a research paper — it's an opinion/perspective piece. The observational study it references (colonoscopy ADR finding) is the empirical evidence. STAT News August 12, 2025 confirms the finding achieved mainstream press coverage. The combination (Lancet editorial + STAT News) = the deskilling concern achieving public health discourse status, not just clinical research status.
+
+## Curator Notes (structured handoff for extractor)
+
+PRIMARY CONNECTION: [[human-in-the-loop clinical AI degrades to worse-than-AI-alone because physicians both de-skill from reliance and introduce errors when overriding correct outputs]]
+WHY ARCHIVED: Lancet publication is the institutional moment when deskilling/never-skilling moved from academic concern to mainstream clinical and educational policy concern. The absence of proven mitigation programs is as important as the evidence of the problem.
+EXTRACTION HINT: Two claims worth extracting separately: (1) update existing deskilling claim with three-pathway taxonomy and colonoscopy ADR evidence; (2) write never-skilling as a distinct new claim emphasizing the baseline-absence problem that makes it structurally invisible.
diff --git a/inbox/archive/health/2025-11-28-stateline-kff-state-medicaid-glp1-coverage-retreat.md b/inbox/archive/health/2025-11-28-stateline-kff-state-medicaid-glp1-coverage-retreat.md
new file mode 100644
index 000000000..708da2380
--- /dev/null
+++ b/inbox/archive/health/2025-11-28-stateline-kff-state-medicaid-glp1-coverage-retreat.md
@@ -0,0 +1,63 @@
+---
+type: source
+title: "States Retreat from GLP-1 Obesity Coverage: 4 States Cut, 13 Remain (Down from 16)"
+author: "Stateline / KFF Health News"
+url: https://stateline.org/2025/11/28/states-retreat-from-covering-drugs-for-weight-loss/
+date: 2025-11-28
+domain: health
+secondary_domains: []
+format: article
+status: unprocessed
+priority: high
+tags: [GLP-1, Medicaid, state-policy, access, obesity, coverage, equity, semaglutide]
+---
+
+## Content
+
+States are retreating from covering GLP-1 medications for weight loss in Medicaid, driven by cost pressures and state budget challenges. As of January 2026, only 13 state Medicaid programs cover GLP-1s for obesity treatment under fee-for-service, down from 16 states in 2025. Four states eliminated coverage effective January 1, 2026:
+
+**California**: Eliminated coverage for GLP-1s when used for weight loss effective January 1, 2026. Maintains coverage for other medically accepted indications (diabetes, cardiovascular disease prevention). Largest state Medicaid program by enrollment.
+
+**Pennsylvania**: Medicaid stopped covering GLP-1s for weight loss for adults 21 and older starting January 1, 2026. Children and young adults under 21 retain coverage (federal law requires Medicaid to cover all medically necessary treatments for people under 21).
+
+**South Carolina**: Ended coverage January 1, 2026.
+
+**New Hampshire**: Ended coverage effective January 1, 2026.
+
+**Michigan**: Did not eliminate coverage but restricted to beneficiaries with BMI ≥40 with strict prior authorization criteria, effective January 1, 2026.
+
+**Additional states considering restrictions**: Rhode Island, Wisconsin, and others are evaluating new limitations.
+
+Primary stated reason across all states: cost. GLP-1 medications (Wegovy, Zepbound) cost $800-$1,000+/month at list price. States cite significant costs associated with coverage and recent state budget challenges including federal funding cuts.
+
+**Federal context**: The BALANCE model (CMS CMMI) was announced in January 2026 as a voluntary mechanism to expand coverage through negotiated drug pricing, launching in Medicaid in May 2026 and Medicare Part D in January 2027. However, participation is voluntary for states, manufacturers, and Part D plans — states that cut coverage would need to voluntarily opt back in through BALANCE.
+
+**Medicare Bridge**: CMS launched a Medicare GLP-1 Bridge program (July 1 - December 31, 2026) at $50/month copay. Critical limitation: Low-Income Subsidy (LIS) beneficiaries cannot use their cost-sharing subsidies for the Bridge — the $50/month copay applies even to the poorest Medicare beneficiaries.
+
+## Agent Notes
+
+**Why this matters:** This is the structural documentation of the access infrastructure collapse happening simultaneously with the evidence that GLP-1 continuous delivery is required for effect. Session 21 established that GLP-1 benefits revert within 1-2 years of cessation; this source documents that the population with highest metabolic disease burden (Medicaid) is losing access to the continuous delivery infrastructure. The compounding failure thesis isn't theoretical — it's being actively created by policy.
+
+**What surprised me:** California cut coverage. California is generally the most progressive state on healthcare access. If California is cutting GLP-1 obesity coverage despite being a leading health access state, this represents a more fundamental cost-sustainability problem than I initially modeled. It's not just red-state cuts — blue-state cost pressures are creating the same outcome.
+
+**What I expected but didn't find:** Any state EXPANDING coverage in 2026. The net direction is entirely negative — retreats, restrictions, and the only federal offset (BALANCE) is voluntary and months away from launching. No state is moving toward broader coverage.
+
+**KB connections:**
+- Directly confirms the access infrastructure dismantling flagged in Session 21
+- The 13-state coverage rate (26% of states) means 74% of Medicaid beneficiaries in states without obesity GLP-1 coverage
+- The Michigan BMI ≥40 restriction (vs FDA-approved ≥30 threshold) creates a coverage gap for the 30-39 BMI range where preventive intervention is most cost-effective
+- Connects to: [[value-based care transitions stall at the payment boundary]] — even "value-based" framing can't overcome $1,000/month drug prices
+- Connects to: [[GLP-1 receptor agonists are the largest therapeutic category launch in pharmaceutical history but their chronic use model makes the net cost impact inflationary through 2035]]
+
+**Extraction hints:**
+- Claim candidate: "State Medicaid GLP-1 obesity coverage is contracting, not expanding — 4 states eliminated coverage in 2026 while BALANCE's voluntary launch mechanism offers no guaranteed offset — creating an access infrastructure gap for the population with highest metabolic disease burden"
+- Frame as: knowledge (GLP-1 effectiveness) advancing while access infrastructure deteriorates — the institutional distribution failure pattern from Session 19 (SELECT trial finding)
+- The California cut is worth flagging specifically — California cutting = cost problem that ideological commitment can't overcome
+
+**Context:** KFF is the authoritative tracker of state Medicaid policy changes. The Stateline article synthesizes state-by-state cuts from multiple journalists. The pattern across states with very different political compositions (CA, PA, SC, NH) suggests this is a fiscal response, not an ideological one.
+
+## Curator Notes (structured handoff for extractor)
+
+PRIMARY CONNECTION: [[GLP-1 receptor agonists are the largest therapeutic category launch in pharmaceutical history but their chronic use model makes the net cost impact inflationary through 2035]]
+WHY ARCHIVED: Confirms access infrastructure collapse — not theoretical, documented in real policy choices across ideologically diverse states including California. Creates specific divergence candidate: "access infrastructure is being dismantled precisely as continuous-treatment evidence makes it most necessary."
+EXTRACTION HINT: Focus on two angles: (1) cost-sustainability of the GLP-1 continuous-treatment model for public payers; (2) the California datum as evidence that this is a structural cost problem, not a political one.
diff --git a/inbox/archive/health/2025-xx-penn-ldi-ajmc-glp1-adherence-lower-income-barriers.md b/inbox/archive/health/2025-xx-penn-ldi-ajmc-glp1-adherence-lower-income-barriers.md
new file mode 100644
index 000000000..541b21959
--- /dev/null
+++ b/inbox/archive/health/2025-xx-penn-ldi-ajmc-glp1-adherence-lower-income-barriers.md
@@ -0,0 +1,68 @@
+---
+type: source
+title: "GLP-1 Adherence Collapse at Year 1-2 — Lower-Income Groups Show Higher Discontinuation; Medicaid PA More Restrictive Than FDA"
+author: "Penn LDI / AJMC / Multiple sources"
+url: https://ldi.upenn.edu/our-work/research-updates/patients-face-new-barriers-for-glp-1-drugs-like-wegovy-and-ozempic/
+date: 2025-01-01
+domain: health
+secondary_domains: []
+format: article
+status: unprocessed
+priority: high
+tags: [GLP-1, adherence, discontinuation, Medicaid, low-income, access-barriers, prior-authorization, commercial-insurance, equity]
+---
+
+## Content
+
+Synthesis of adherence and access barrier evidence for GLP-1 obesity therapy:
+
+**AJMC adherence study (commercially insured, n=16 million+ patients without diabetes, 2021):**
+- 1-year adherence for Wegovy: 36%
+- 1-year adherence for Ozempic: 47%
+- 2-year adherence (follow-on study, presented April 2025): only 14.3% of patients still on therapy
+- This is COMMERCIAL insurance — the best-coverage, highest-income population
+
+**Discontinuation determinants:**
+- Higher discontinuation: lower-income groups, multiple health conditions, age over 65
+- High costs, lack of insurance coverage, and adverse effects drive discontinuation
+- For lower-income populations: out-of-pocket cost is cited as the primary barrier even when drugs are technically covered
+
+**Medicaid prior authorization specifics:**
+- 70% of Medicaid PA policies specify conditions more restrictive than FDA-approved criteria
+- Typical PA requirements: documented diet/exercise failure, specific BMI thresholds above FDA minimum, specific comorbidity combinations
+- Prior authorization is functionally a clinical gatekeeping mechanism that the healthcare system uses to limit access beyond what the FDA deems clinically appropriate
+
+**Penn LDI framing:**
+- "Patients face new barriers" — not old barriers, new ones emerging in 2025-2026 as states cut coverage, Medicaid implements stricter PA, and insurance denials persist
+
+**The arithmetic of the access gap:**
+- If 36-47% of commercially insured patients (with the best coverage) adhere at year 1, and GLP-1 benefits require continuous delivery...
+- Then Medicaid patients — with PA more restrictive than FDA, higher cost barriers, higher burden of social determinants affecting adherence — likely have substantially lower adherence rates
+- The compounding: (lower adherence) × (higher baseline metabolic disease burden) × (continuous delivery required for effect) = the population most needing the intervention has the least sustained access to it
+
+## Agent Notes
+
+**Why this matters:** The 14.3% two-year adherence figure in commercially insured patients is the most alarming datum in the GLP-1 adherence literature. Combined with the Session 20 finding (GLP-1 benefits revert within 1-2 years of cessation), 85.7% of commercially insured patients on GLP-1s are not achieving durable metabolic benefit — because they've discontinued before the rebound occurs. For Medicaid patients with additional barriers, the number is likely worse.
+
+**What surprised me:** That the 14.3% two-year adherence figure is from COMMERCIAL insurance (April 2025 presentation). I expected adherence to be better in commercial populations. The fact that even well-insured patients can't sustain GLP-1 therapy past 2 years at scale means the adherence problem isn't primarily financial — there's a broader behavioral/pharmacological challenge that financial coverage alone doesn't solve. This COMPLICATES the access-as-solution narrative.
+
+**What I expected but didn't find:** A direct study comparing Medicaid vs. commercial insurance adherence rates for GLP-1 obesity treatment. That comparison doesn't appear to exist yet as a published study — likely because Medicaid coverage has been so limited that there's no large population to study. The direct comparison is a genuine research gap.
+
+**KB connections:**
+- Supports Session 20's finding: `2026-04-08-bcbs-glp1-persistence-doubled.md` — BCBS persistence data (also commercial)
+- The continuous-treatment model (Sessions 20-21): 85.7% non-adherers won't achieve durable benefit
+- The access infrastructure collapse (this session, multiple sources): Medicaid coverage cuts
+- Together: the population with highest metabolic burden has both lowest access AND likely lowest adherence
+
+**Extraction hints:**
+- Claim: "GLP-1 two-year adherence is only 14.3% in commercially insured patients, meaning the continuous-delivery infrastructure required for durable metabolic benefit is not being maintained even in the best-coverage population — and is almost certainly lower in Medicaid and uninsured populations"
+- This is a complicating finding: the problem isn't only access (coverage), it's also adherence (sustained delivery). The solution requires BOTH coverage AND support infrastructure.
+- Note the Medicaid PA finding (70% more restrictive than FDA) as an administrative gatekeeping mechanism above clinical evidence.
+
+**Context:** Penn LDI (Leonard Davis Institute of Health Economics at University of Pennsylvania) is a leading health policy research institution. The AJMC study (16 million patients) is one of the largest real-world adherence analyses for GLP-1 in obesity treatment.
+
+## Curator Notes (structured handoff for extractor)
+
+PRIMARY CONNECTION: Continuous-treatment model (Session 21 musing) and the GLP-1 adherence literature thread
+WHY ARCHIVED: The 14.3% two-year adherence figure in the BEST-coverage population reveals that the access problem is not just financial — it's behavioral/pharmacological adherence combined with financial barriers. This complicates the "expand coverage → solve the problem" narrative in a KB-valuable way.
+EXTRACTION HINT: Two claims: (1) GLP-1 2-year adherence at 14.3% even in commercial insurance; (2) the combination of low adherence + continuous-delivery requirement = most patients aren't achieving durable benefit even when covered. The Medicaid PA (70% more restrictive than FDA) is a separate, extractable claim.
diff --git a/inbox/archive/health/2026-01-05-kff-balance-model-glp1-coverage-gap-analysis.md b/inbox/archive/health/2026-01-05-kff-balance-model-glp1-coverage-gap-analysis.md
new file mode 100644
index 000000000..314a6e88b
--- /dev/null
+++ b/inbox/archive/health/2026-01-05-kff-balance-model-glp1-coverage-gap-analysis.md
@@ -0,0 +1,68 @@
+---
+type: source
+title: "KFF: BALANCE Model for GLP-1s — What It Does and Doesn't Offset"
+author: "KFF Health News"
+url: https://www.kff.org/medicare/what-to-know-about-the-balance-model-for-glp-1s-in-medicare-and-medicaid/
+date: 2026-01-05
+domain: health
+secondary_domains: []
+format: analysis
+status: unprocessed
+priority: high
+tags: [GLP-1, BALANCE-model, CMS, Medicare, Medicaid, coverage, access, obesity, policy]
+---
+
+## Content
+
+The BALANCE (Better Approaches to Lifestyle and Nutrition for Comprehensive hEalth) Model is a CMS CMMI voluntary test to expand GLP-1 coverage in Medicare Part D and Medicaid for weight management.
+
+**What it does:**
+- Negotiates drug pricing with manufacturers (Eli Lilly, Novo Nordisk agreements completed)
+- Enables states and Part D plans to cover GLP-1s for obesity under a statutory waiver
+- Requires participating enrollees to receive lifestyle support alongside medication
+- Medicaid launch: rolling May-December 2026 (deadline for state notification: July 31, 2026)
+- Medicare Part D launch: January 2027
+
+**What it doesn't do (critical limitations):**
+1. **Voluntary for everyone** — states, manufacturers, and Part D plans all choose to participate. No entity is required to join. No participating state list has been published as of April 2026.
+2. **Doesn't fix January 2026 cuts** — California, Pennsylvania, South Carolina, and New Hampshire eliminated coverage effective January 1, 2026. These states would need to voluntarily opt into BALANCE to restore coverage. BALANCE launching in May 2026 creates a 4+ month coverage gap even for states that participate.
+3. **Medicare Bridge LIS exclusion** — The Medicare GLP-1 Bridge (July-December 2026, $50/month copay) explicitly excludes Low-Income Subsidy beneficiaries from their cost-sharing subsidies. The poorest Medicare beneficiaries face full $50/month copay.
+4. **Lifestyle support requirement** — BALANCE requires participants to engage with evidence-based lifestyle supports. This is clinically appropriate but may create additional access barriers for populations with limited time, digital access, or health literacy.
+5. **No guarantee of price adequacy** — CMS negotiated with manufacturers but hasn't disclosed the negotiated prices. The level of discount achieved may not make drugs affordable for states facing budget constraints.
+
+**Coverage gap math:**
+- 16 states covered GLP-1 obesity treatment in Medicaid as of 2025
+- 13 states cover in January 2026 (net -3 states in 12 months)
+- BALANCE offers potential recovery, but only for states that opt in voluntarily
+- Net effect in Q1-Q2 2026: coverage is worse than 2025, with no confirmed offset
+
+**The access inversion problem:**
+- States with highest metabolic disease burden (Southern states, rural states) tend to have lowest GLP-1 coverage rates
+- States that can afford coverage (larger tax base, better fiscal health) are cutting due to cost
+- The populations most in need (Medicaid enrollees with comorbid obesity + metabolic disease) face the highest access barriers
+
+## Agent Notes
+
+**Why this matters:** BALANCE is the official "answer" to access concerns — but it's a voluntary mechanism that doesn't guarantee coverage for any specific population. The gap between BALANCE as a policy mechanism and BALANCE as an access guarantee is large. This is the disconfirmation test for whether the "compounding failure" thesis is being offset by policy: ANSWER IS NO. The offset mechanism exists on paper but isn't operational and requires voluntary adoption from the same state budgets that just cut coverage.
+
+**What surprised me:** The Medicare Bridge LIS exclusion. Low-Income Subsidy beneficiaries are, by definition, the lowest-income Medicare participants. Creating a program to expand access to GLP-1s and then explicitly excluding cost-sharing protections for the poorest beneficiaries is a structural contradiction. The $50/month copay is a meaningful barrier for someone on $800-900/month SSI.
+
+**What I expected but didn't find:** Any committed list of states that have signed up for BALANCE as of April 2026. The model was announced January 2026, state notification deadline is July 31, 2026. We're 4 months post-announcement and no public participation list. This is consistent with states needing time to evaluate, but it means there's no confirmed coverage expansion yet.
+
+**KB connections:**
+- The "structural separation" of BALANCE enrollment from state coverage cuts means the compounding failure pattern (Session 21) is NOT being offset
+- Connects to: [[value-based care transitions stall at the payment boundary because 60 percent of payments touch value metrics but only 14 percent bear full risk]] — voluntary models face similar participation limitations
+- The LIS exclusion is a specific instance of access being structurally inverted: program designed for access, structured to exclude the lowest-income
+
+**Extraction hints:**
+- Claim candidate: "The BALANCE model offers voluntary GLP-1 coverage expansion but does not offset the January 2026 state coverage retreats — creating a net coverage gap for Medicaid beneficiaries in 2026 that voluntary participation mechanisms cannot close in the near term"
+- The LIS exclusion is extractable as a specific claim about how access programs can replicate access inversions through their own design
+- Consider connecting to Session 19's "SELECT trial finding" pattern: knowledge advancing while infrastructure retreats
+
+**Context:** KFF's analysis is the authoritative source for Medicare/Medicaid policy interpretation. The NCPA (National Community Pharmacists Association) formally announced the model January 5, 2026. Multiple law firm analyses (Mintz, ReedSmith) confirm the voluntary structure and limitations.
+
+## Curator Notes (structured handoff for extractor)
+
+PRIMARY CONNECTION: [[value-based care transitions stall at the payment boundary because 60 percent of payments touch value metrics but only 14 percent bear full risk]]
+WHY ARCHIVED: The BALANCE model is the policy response to GLP-1 access concerns, and its voluntary structure means it provides no guaranteed offset to the January 2026 coverage cuts. This is direct evidence that compounding access failures are not being systematically addressed.
+EXTRACTION HINT: Focus on the gap between BALANCE as mechanism vs. BALANCE as guarantee. The LIS exclusion is the sharpest evidence of structural access inversion.
diff --git a/inbox/archive/health/2026-01-xx-urbina-clinical-obesity-glp1-micronutrient-narrative-review.md b/inbox/archive/health/2026-01-xx-urbina-clinical-obesity-glp1-micronutrient-narrative-review.md
new file mode 100644
index 000000000..995f42e2b
--- /dev/null
+++ b/inbox/archive/health/2026-01-xx-urbina-clinical-obesity-glp1-micronutrient-narrative-review.md
@@ -0,0 +1,62 @@
+---
+type: source
+title: "GLP-1 Micronutrient Deficiencies: Narrative Review of 6 Studies (n=480,825) — Iron, Calcium, Vitamin D, Protein Deficits Systematic"
+author: "Urbina et al., Clinical Obesity (Wiley)"
+url: https://onlinelibrary.wiley.com/doi/10.1111/cob.70070
+date: 2026-01-01
+domain: health
+secondary_domains: []
+format: research-paper
+status: unprocessed
+priority: high
+tags: [GLP-1, micronutrients, nutritional-deficiency, iron, calcium, vitamin-D, protein, semaglutide, safety, monitoring]
+---
+
+## Content
+
+Systematic narrative review of micronutrient and nutritional deficiencies associated with GLP-1 receptor agonist therapy. Structured PubMed and Cochrane search (January 2019 – May 2025), 6 studies meeting inclusion criteria, encompassing 480,825 adults.
+
+**Key quantitative findings:**
+
+- **Vitamin D**: 7.5% deficiency at 6 months, 13.6% at 12 months. Mean vitamin D intake of 4 μg/day — significantly lower than estimated national average of 19 μg/day. Only 1.4% of GLP-1 users met 100% of the Dietary Reference Intake (DRI) for vitamin D.
+
+- **Iron**: GLP-1 users demonstrate 26–30% lower ferritin levels than SGLT2 inhibitor comparators. 64% of GLP-1RA users consumed below the estimated average requirement (EAR) for iron. Iron absorption drops markedly after 10 weeks of semaglutide (prospective pilot, n=51).
+
+- **Calcium**: 72% of users consumed below the Recommended Dietary Allowance (RDA) for calcium.
+
+- **Protein**: 58% did not meet recommended protein intake targets (1.2–1.6 g/kg/day during active weight loss per OMA/ASN guidance).
+
+- **Thiamine and cobalamin**: Deficits increase over time (consistent pattern).
+
+**Mechanism**: GLP-1-induced appetite suppression is non-selective — it reduces total caloric intake including micronutrient-rich foods. Delayed gastric emptying alters absorption kinetics. The drugs do not distinguish between "calories to reduce" and "nutrients to maintain."
+
+**Clinical implication stated by authors**: "Micronutrient deficiencies during GLP-1RA therapy are a common consequence rather than a rare adverse effect."
+
+**Monitoring gap**: 92% of patients had no dietitian visit in the 6 months prior to GLP-1 prescription (from complementary study). Multi-society advisory (OMA/ASN/ACLM/Obesity Society) recommends proactive nutritional monitoring and supplementation but protocol adoption lags at scale.
+
+## Agent Notes
+
+**Why this matters:** This is the systematic literature synthesis confirming that what was seen in single large cohorts is robust across studies. The n=480,825 across 6 studies means this isn't one health system's data — it's a meta-level confirmation of the nutritional deficiency pattern. The framing — "common consequence, not rare adverse effect" — should change how GLP-1 prescribing infrastructure is designed.
+
+**What surprised me:** The 1.4% vitamin D DRI compliance figure. This means 98.6% of GLP-1 users are NOT meeting vitamin D intake needs through diet. Combined with already-high population-level vitamin D deficiency rates (approximately 40% in the US generally), GLP-1 users are starting from a disadvantaged baseline and making it significantly worse. This is not a marginal nutritional concern — it's near-universal.
+
+**What I expected but didn't find:** Any stratification of deficiency rates by socioeconomic status, food security, or Medicaid vs. commercial insurance status. The review analyzed GLP-1 users generally — no breakdown for the food-insecure population where baseline micronutrient deficiency is already elevated. The food-insecure + GLP-1 double-jeopardy remains an inference, not a direct measurement (see research gap note in Session 21).
+
+**KB connections:**
+- Supplements and extends: existing archive `2026-04-08-glp1-nutritional-deficiency-signal.md` (different source, overlapping findings but broader systematic methodology)
+- Reinforces the monitoring infrastructure argument: if 64% iron-deficient, 72% calcium-deficient, 58% protein-deficient — the software layer providing dietary tracking becomes medically essential
+- Directly relevant to the OMA/ASN/ACLM advisory already archived: the advisory was right to flag nutritional monitoring as essential infrastructure
+- Connects to atoms-to-bits argument: continuous dietary monitoring alongside GLP-1 delivery is the natural moat position
+
+**Extraction hints:**
+- Primary claim: "GLP-1 receptor agonist therapy produces systematic micronutrient deficiencies in the majority of users — 64% iron-deficient, 72% calcium-deficient, 58% protein-deficient — with only 1.4% of users meeting vitamin D dietary requirements, making nutritional monitoring infrastructure a clinical necessity not an optional enhancement"
+- Note scope carefully: "common consequence, not rare adverse effect" is the claim's core precision
+- The 1.4% vitamin D compliance figure is the most concrete single datum for the headline claim
+
+**Context:** Urbina et al. published in Clinical Obesity (Wiley), a peer-reviewed journal of the World Obesity Federation. The narrative review methodology is appropriate for synthesizing heterogeneous study designs. The 6-study cutoff is a limitation — this is a rapidly evolving field — but the convergence across studies strengthens the directional conclusion.
+
+## Curator Notes (structured handoff for extractor)
+
+PRIMARY CONNECTION: Existing `2026-04-08-glp1-nutritional-deficiency-signal.md` archive + OMA/ASN advisory archive
+WHY ARCHIVED: Systematic multi-study synthesis (not a single cohort) confirming nutritional deficiency as a common consequence. The framing upgrade — "common consequence, not rare adverse effect" — elevates this from a signal to a clinical fact requiring infrastructure response.
+EXTRACTION HINT: Claim should emphasize the near-universality of specific deficits (iron: 64%, calcium: 72%, vitamin D: 98.6% not meeting DRI) rather than just prevalence statistics. The monitoring gap (92% no dietitian visit) is the infrastructure claim that follows.
diff --git a/inbox/archive/health/2026-06-xx-pubmed-glp1-hfpef-systematic-review-meta-analysis-mortality-hospitalization.md b/inbox/archive/health/2026-06-xx-pubmed-glp1-hfpef-systematic-review-meta-analysis-mortality-hospitalization.md
new file mode 100644
index 000000000..35a3e4c94
--- /dev/null
+++ b/inbox/archive/health/2026-06-xx-pubmed-glp1-hfpef-systematic-review-meta-analysis-mortality-hospitalization.md
@@ -0,0 +1,75 @@
+---
+type: source
+title: "GLP-1 Agonists in HFpEF: Meta-Analysis of 6 RCTs (n=4,043) Shows 27% Mortality/Hospitalization Reduction — Divergence with ACC 'Insufficient Evidence' Stance"
+author: "PubMed (BMC Cardiovascular Disorders / Springer Nature)"
+url: https://pubmed.ncbi.nlm.nih.gov/40637782/
+date: 2026-06-01
+domain: health
+secondary_domains: []
+format: research-paper
+status: unprocessed
+priority: high
+tags: [GLP-1, HFpEF, heart-failure, meta-analysis, semaglutide, tirzepatide, mortality, cardiovascular, divergence-candidate]
+---
+
+## Content
+
+Systematic review and meta-analysis examining GLP-1 receptor agonist impact on cardiovascular outcomes in heart failure with preserved ejection fraction (HFpEF).
+
+**Study characteristics:**
+- 6 studies (5 RCTs + 1 cohort study)
+- n = 4,043 patients total
+- Studies evaluated: 5 semaglutide, 1 tirzepatide
+
+**Primary finding:**
+- GLP-1 agonists reduced composite outcome of **all-cause mortality + heart failure hospitalization by 27%** (HR 0.73; 95% CI: 0.60–0.90)
+
+**Supporting real-world evidence (complementary study — US health care claims data 2018–2024):**
+- Semaglutide initiators: HR 0.58 (42% risk reduction) vs. sitagliptin for composite of HF hospitalization + all-cause mortality
+- Tirzepatide initiators: HR 0.42 (58% risk reduction) vs. sitagliptin
+- Study design: two cohort studies emulating STEP-HFpEF-DM and SUMMIT trials, national claims data
+
+**AJMC pooled STEP-HFpEF analysis:**
+- GLP-1s reduced adverse HF events by approximately 40% in HFpEF patients (Pharmacy Times / AJMC analysis)
+
+**ACC 2025 HFpEF scientific statement (from prior archive `2025-06-xx-jacc-acc-scientific-statement-obesity-adults-heart-failure.md`):**
+- "Symptoms improve with GLP-1 in obese HFpEF; mortality/hospitalization endpoint evidence is 'insufficient to confidently conclude' benefit"
+- 2023 ACC Expert Consensus: GLP-1 agonists "may be considered" (weak recommendation) for obese individuals with DM and HFpEF
+
+**The evidence tension:**
+- Trial evidence interpretation (ACC): STEP-HFpEF tested mortality/hospitalization as secondary composite endpoint — not powered for this outcome — therefore "insufficient"
+- Meta-analysis interpretation: pooling 6 studies yields 27% reduction with HR 0.73 (CI 0.60–0.90) — statistically significant
+- Real-world evidence: 42–58% risk reduction in national claims data
+- Resolution question: Does pooling secondary endpoints across multiple underpowered trials produce valid primary evidence, or does it compound the underpowering problem?
+
+**Clinical penetration context (from Session 21 archives):**
+- ~6.7–6.9M HFpEF patients in US; ~2.2M are obese and theoretically eligible
+- Total STEP-HFpEF + SUMMIT trial enrollment: ~1,876 patients
+- Clinical penetration: research-scale, not population-scale
+
+## Agent Notes
+
+**Why this matters:** This is a genuine divergence candidate. The same body of evidence is being interpreted differently by different evaluative frameworks — ACC's methodological strictness (secondary endpoints = insufficient) vs. meta-analysis synthesis (27% from pooled evidence). Both interpretations are defensible. The divergence has clinical implications: if GLP-1s reduce mortality in obese HFpEF, undertreatment at population scale represents preventable deaths. If the effect is a statistical artifact of pooling secondary endpoints, broad adoption creates risk.
+
+**What surprised me:** The real-world evidence (42-58% reduction) is substantially larger than the trial-based meta-analysis (27%). This is unusual — typically RCT effects exceed real-world effects due to selection bias and protocol adherence. The larger real-world effect might reflect: (1) the sitagliptin comparator being worse than placebo, (2) selection of patients who are more adherent than average trial participants, or (3) the GLP-1 mechanisms working better in real-world comorbidity complexity than in clean trial populations. This needs scrutiny.
+
+**What I expected but didn't find:** Any ACC/AHA update to the "may be considered" recommendation incorporating the new meta-analysis evidence. The ACC 2023 guidance predates most of this evidence; a 2025 update was found in the health archive (`2025-06-xx`), but the specific mortality endpoint characterization needs checking.
+
+**KB connections:**
+- Existing archive: `2025-06-xx-jacc-acc-scientific-statement-obesity-adults-heart-failure.md`
+- Existing archive: `2026-04-08-glp1-semaglutide-tirzepatide-cardiac-mechanism.md` — weight-independent cardiac mechanism
+- Existing archive: `2024-xx-journal-cardiac-failure-glp1-hfpef-malnutrition-sarcopenia-caution.md` — the opposing caution
+- Together these three archives create a genuine divergence: benefit evidence + safety concern (sarcopenic obesity paradox) + mechanism uncertainty
+
+**Extraction hints:**
+- This source is PRIMARILY a divergence-trigger — propose `domains/health/divergence-glp1-hfpef-mortality-evidence-vs-guideline-caution.md`
+- The divergence should link: (1) this meta-analysis, (2) ACC "insufficient evidence" characterization, (3) sarcopenic obesity paradox caution, (4) real-world vs. trial magnitude discrepancy
+- The "What Would Resolve This" section: a dedicated HFpEF outcomes RCT powered for mortality/hospitalization as PRIMARY endpoint
+
+**Context:** Published in BMC Cardiovascular Disorders (Springer Nature), peer-reviewed cardiology journal. Meta-analysis methodology note: 5 RCTs included had mortality/hospitalization as secondary, not primary, endpoints — this is the ACC's stated reason for caution. The study is legitimate evidence but the pooling methodology deserves scrutiny.
+
+## Curator Notes (structured handoff for extractor)
+
+PRIMARY CONNECTION: `domains/health/divergence-` candidate linking GLP-1 HFpEF benefit evidence vs. guideline caution
+WHY ARCHIVED: Creates a genuine knowledge base divergence between RCT-pooling methodology (27% benefit) and ACC's methodological strictness (secondary endpoints = insufficient for confident conclusion). Divergences are the KB's highest-value content.
+EXTRACTION HINT: Do NOT write as a single claim. Write as a divergence file: `divergence-glp1-hfpef-mortality-benefit-vs-guideline-caution.md`. The divergence is more valuable than any single claim that could be extracted.