Bipolar disorder dossier — APA 2024 + CANMAT/ISBD 2018 Yatham + NICE 2024 + BALANCE Geddes Lancet 2010 + STEP-BD Sachs NEJM 2007 Axis 1 Acute mania: lithium 300 TID target 0.8–1.2 OR valproate 20–30 mg/kg target 50–125; combo with SGA for YMRS ≥30 or psychotic features Axis 2 Bipolar depression: quetiapine 300 QHS, lurasidone 40–120 with food, cariprazine 1.5–3, lamotrigine slow-titrate to 200; NEVER antidepressant monotherapy (STEP-BD Sachs NEJM 2007) Axis 3 Maintenance: lithium preferred per BALANCE (Geddes Lancet 2010) target 0.6–0.8; lifelong for BP I; lamotrigine for depression-predominant BP II Severity triggers (9): severe_mania_ymrs_30, active_si_bipolar, psychotic_features_in_mania, catatonia_in_bipolar (Bush-Francis), lithium_toxicity, nms_on_antipsychotic, pregnancy_on_valproate, mixed_features_with_si (NEW — highest SI risk phenotype, cross-routes to psych.suicidality.ed.core.v1), rapid_cycling_emergence (NEW — discontinue antidepressant, optimise mood stabiliser, refer specialist) Three setting playbooks: ED (agitation containment + disposition), inpatient (stabilisation to therapeutic levels), outpatient (lifelong maintenance + IPSRT/CBT + LAI aripiprazole for non-adherence per Calabrese AJP 2017) Sibling differentiation vs psych.depression.core.v1 — antidepressant monotherapy contraindicated in bipolar; cross-routes to psych.suicidality.ed.core.v1 (mixed-features SI), psych.first-episode-psychosis.core.v1 (mood-incongruent psychosis), psych.alcohol_withdrawal.core.v1 + psych.opioid_use_disorder.core.v1 (substance-induced mania) Deepened 2026-05-14 (shard-5-obped-id depth-pass-1): authored co-located _briefs/psych.bipolar-disorder.core.v1.md (sepsis-template-style — Frame, phenotype matrix 9-axis, settings 5-tier, Bayesian linkage YMRS+lithium-level+CK+MDQ LRs, severity triggers gap analysis, 12-phase author order, Stage-A API checklist, schema proposal cache) + _research-bundles/psych.bipolar-disorder.core.v1.md (companion bundle anchoring APA 2024, CANMAT/ISBD 2018, NICE 2024, STEP-BD, BALANCE, Cipriani 2005, BOLDER, Loebel lurasidone, Durgam cariprazine, Pompili 2013, Calabrese LAI 2017, Pagnin ECT 2004). Bumped last_reconciled 2026-05-13 → 2026-05-14. Deepened 2026-05-14: fixed broken pointers — manifest blanked (out-of-shard-scope per peds.febrile-infant / psych.suicidality / psych.alcohol_withdrawal precedent); design_brief now resolves to authored co-located brief. Deepened 2026-05-14: repointed 4 unresolved workup IDs to registered umbrellas — workup.catatonia_screen → workup.delirium (DELIRIUMS workup covers catatonia + reversibles); workup.lithium_toxicity → workup.encephalopathy (lithium tox presents as AMS); workup.nms_screen → workup.hyperthermic_toxidromes (covers SS/NMS/MH exact match); workup.substance_induced_mania → workup.severe_agitation (5-fold differential including drug intox). Deepened 2026-05-14: removed 3 unregistered calculator IDs (calc.ymrs, calc.cssrs, calc.bmi) — YMRS workflow remains embedded in workup actions + severity_triggers (severe_mania_ymrs_30) + Axis 1 step_up_when language; C-SSRS referenced via workup.suicide_risk (registered umbrella); BMI monitoring continues via panel.glucose_a1c + setting_playbook narratives. All three flagged for future clinical-tools-registry batch. Phenotype matrix (subtype × current-episode × psychotic × rapid-cycling × catatonia × substance-induced × seasonal × postpartum × treatment-resistant) is encoded indirectly via severity_triggers + setting_playbooks + sibling_differentiation. First-class TS field for phenotype matrix is schema-blocked — see docs/framework-audit/shard-5-obped-id-state.md Schema-blocked queue. Bayesian linkage (MDQ LR+ ≈ 7 / LR− ≈ 0.3 in psychiatric outpatients per Hirschfeld 2000; family-history of BP I in 1st-degree relative LR+ ≈ 7–10 for bipolar in MDE presentation; YMRS ≥30 LR+ ≈ 8 for inpatient need; classic euphoric mania LR+ ≈ 3 for lithium response per Bowden 2001; mixed/rapid-cycling LR− ≈ 0.5 for lithium response → valproate preferred; lithium serum >1.5 LR+ ≈ 8 for toxicity, haemodialysis at >4.0 or >2.5 + symptoms per EXTRIP 2015; CK ≥1000 + temp ≥38 + rigidity on AP LR+ ≈ 15 for NMS per Caroff 2015; T_treat = moderate-severe YMRS = mood stabiliser + SGA; T_test = mild BP II hypomania without functional impairment = watchful waiting + MDQ-tracked; cross-dossier routing to psych.suicidality.ed.core.v1 + psych.depression.core.v1 + psych.first-episode-psychosis.core.v1 + psych.alcohol_withdrawal.core.v1 documented in co-located research bundle). ROS/DDx LR seed data audited by npm run audit:ros-ddx-coverage (cross-cutting; not touched by this shard). CRITICAL anti-pattern warnings baked in: (1) NEVER antidepressant monotherapy in bipolar (STEP-BD Sachs NEJM 2007); (2) valproate absolute contraindication in pregnancy (FDA 2013 black-box); (3) lamotrigine slow titration mandatory (SJS); halve dose with valproate; (4) lurasidone with food ≥350 kcal; (5) olanzapine IM no concurrent IM benzodiazepine; (6) ziprasidone QTc ECG; (7) carbamazepine HLA-B*15:02 Asian-ancestry screen; (8) lithium narrow therapeutic index — Q5-7d levels titration; dehydration/NSAID/ACEI/diuretic precipitate toxicity; (9) do not discontinue mood stabiliser abruptly — rebound mania within weeks (Suppes 1991); (10) antidepressant 8-week post-remission discontinuation to minimise switch risk; (11) mixed features = highest SI risk → C-SSRS at every visit; (12) sleep deprivation = mania trigger (Harvey 2008). PRODUCTION blockers: (1) calc.ymrs + calc.cssrs + calc.bmi not in clinical-tools-registry — flagged for future calc-registry batch; (2) RxCUIs need RxNav validation via scripts/research/rxnav-validate.ts; (3) targeted test file pending (relies on dossier-contract.test.ts); (4) ECT atoms (number of sessions, bilateral vs unilateral, frequency) not yet registered; (5) prisma/seed/manifests/ pointer out-of-shard-scope per precedent — manifest field blanked.