Phase B deepening (2026-05-14): GBS variants (AIDP/AMAN/AMSAN/Miller-Fisher/Bickerstaff/PCB) encoded as severity_triggers — pivot from variant-specific engines because future variant engines are Phase C/D expansion targets. 5 setting playbooks span the full journey: home (EMS recognition + single-breath count + no sedation) → ed (Brighton + EGRIS + LP + NCS + IgA pre-IVIG) → icu (20-30-40 rule + autonomic monitoring) → inpatient (treatment-related fluctuation watch + CIDP transition surveillance) → outpatient (rehab + EGOS prognostic + CIDP surveillance + influenza vaccination discussion). IVIG vs PLEX equivalent per Cochrane Hughes 2014 PMID 25238327; do NOT combine. Steroids alone NOT effective in GBS (Hughes Brain 2007 PMID 17337484; EAN/PNS 2023 PMID 37814552 recommends against). Avoid succinylcholine (denervation hyperkalemia) — use rocuronium + sugammadex. CIDP/A-CIDP transition surveillance: ~5% of patients initially diagnosed as GBS reclassify as acute-onset CIDP if progression continues >8 wk or ≥3 treatment-related fluctuations (EAN/PNS 2023 PMID 37814552; Fokke Brain 2014 PMID 24163275 — n=494, 5% deteriorated after 8 wk). Influenza vaccination not contraindicated post-acute GBS per CDC ACIP; risk-benefit discussion with neurology recommended. Schema-blocked calculators surfaced as tickets in docs/framework-audit/shard-3-neuro-sym-state.md: calc.egos (EGOS van Koningsveld 2007 PMID 17537676), calc.megos (mEGOS Walgaard 2011 PMID 21403108), calc.egris (EGRIS Walgaard 2010 PMID 20517939), calc.brighton (Brighton criteria validated Fokke 2014 PMID 24163275). Currently encoded as plain-English required_assessments and severity_triggers — NO new calc.* ids introduced. §5.5.1 effect sizes: (1) IVIG vs PLEX — change in 7-grade disability at 4 wk mean difference 0.02 grade (95% CI -0.20 to 0.25), i.e. EQUIVALENT (Hughes Cochrane 2014 PMID 25238327; Hughes Brain 2007 WMD -0.02 [95% CI -0.25 to 0.20] PMID 17337484). (2) PLEX vs supportive — improved ≥1 disability grade at 4 wk RR 1.64 (95% CI 1.37–1.96); mechanical-ventilation requirement RR 0.53 (95% CI 0.39–0.74); full muscle-strength recovery at 1 yr RR 1.24 (95% CI 1.07–1.45); severe motor sequelae RR 0.65 (95% CI 0.44–0.96) (Raphaël Cochrane 2012 PMID 22786475; orig PMID 11406009). (3) EGRIS — derivation cohort n=397, 22% needed MV in week 1; score 0–7 maps to 1%→91% MV risk; AUC 0.84 (Walgaard 2010 PMID 20517939). (4) mEGOS — age + preceding diarrhoea + MRC sumscore predict inability to walk at 6 mo; AUC 0.84–0.87 at day 7 vs 0.73–0.77 at admission (Walgaard 2011 PMID 21403108). (5) EGOS — 3-variable score 1–7, predicted inability to walk at 6 mo 1%→83%; AUC 0.85 (van Koningsveld 2007 PMID 17537676). (6) Respiratory failure — ~22–25% of GBS require mechanical ventilation (Walgaard 2010 PMID 20517939). (7) IVIG dose — 2 g/kg total (0.4 g/kg/d × 5 d), GPP within 2–4 wk of onset if unable to walk unaided (EAN/PNS 2023 PMID 37814552). (8) Outcome/mortality — 3–10% die, ~20% unable to walk at 6 mo (van Doorn Lancet Neurol 2008 PMID 18848313); IGOS n=1000: walk-unaided at 1 yr 83% Europe/Americas vs 69% Bangladesh, mortality 5% vs 17%, axonal subtype 6% vs 36% (Doets Brain 2018 PMID 30247567). (9) CSF — albuminocytologic dissociation rises 49% (day 1) → 88% (≥2 wk), time-dependent (Fokke Brain 2014 PMID 24163275). Resolving cross-dossier routes (verified to exist as dossier .ts files 2026-05-18): respiratory-failure / EGRIS-high / FVC<20 → ICU airway pathway; demyelinating NCS subtype → neuro.gbs-aidp.v1; axonal NCS + anti-GM1/GD1a → neuro.gbs-aman.v1; anti-GQ1b + ophthalmoplegia/ataxia/areflexia triad → neuro.gbs-miller-fisher.v1; demyelinating-relapse >8 wk (A-CIDP) → neuro.peripheral-neuropathy.v1; sensory-level cord mimic → neuro.transverse-myelitis.v1; fatigable/fluctuating preserved-reflex weakness → neuro.mg-crisis.core.v1. (neuro.cidp.core.v1 / neuro.botulism.v1 / neuro.tick-paralysis.v1 do NOT yet exist on disk → kept as descriptive sibling_engine_id placeholders.) Special populations (as data): PREGNANCY — IVIG is the preferred immunomodulator (PLEX feasible but volume shifts; corticosteroids not used); GBS in pregnancy does not mandate early delivery; multidisciplinary OB/neuro/ICU (EAN/PNS 2023 PMID 37814552). PAEDIATRIC — IVIG hastens recovery vs supportive care (low-quality evidence, Hughes Cochrane 2014 PMID 25238327); same 2 g/kg dosing; generally better prognosis than adults. AUTONOMIC DYSFUNCTION — present in a majority of severe GBS, a leading mortality cause; continuous ECG + cautious treatment of labile BP/arrhythmia (van Doorn 2008 PMID 18848313). AXONAL vs DEMYELINATING — same acute IVIG/PLEX dose; axonal (AMAN/AMSAN) has slower/poorer recovery and warrants early rehab planning (Doets IGOS 2018 PMID 30247567). ANTI-PATTERN — corticosteroid monotherapy is INEFFECTIVE in GBS and weakly harmful (oral steroids WMD -0.82 favouring no steroids); EAN/PNS recommends AGAINST oral steroids and weakly against IV steroids (Hughes Brain 2007 PMID 17337484; EAN/PNS 2023 PMID 37814552). §5.5.2 Bayesian layer shipped as ros-ddx seed triple (prisma/seed/ros-and-ddx/neuro.gbs.core.v1.{ros,differentials,finding-lrs}.ts — the repo EXEMPLAR; existing rows preserved, +14 LR rows / +4 differentials / +2 ROS / +3 conditional-dependency notes appended this pass). All seed evidence_source PMIDs reconciled to PubMed-MCP-verified anchors; full was→actual→corrected table in neuro.gbs.core.v1._research-bundle.md. NO RxCUI/drug code added or changed this pass. PMID provenance: all 12 evidence.pmids + all reconciled seed PMIDs PubMed-MCP-VERIFIED 2026-05-18 (no NEEDS_SOURCE_REVIEW remaining in evidence.pmids). The only outstanding NEEDS_SOURCE_REVIEW is the Wendell/Levine Neurohospitalist 2011 MG-crisis review (prior seed PMID 23983832 was a wrong-article mis-attribution; canonical PMID not PubMed-MCP-confirmable this pass — MG-crisis discrimination rows re-anchored to verified EAN/PNS 2023 + Yuki NEJM 2012, which carry the actual reflex/sensory/serology discriminators). 6 differential siblings: MG-crisis + MS-flare + neuro.gbs-aidp/aman/miller-fisher + transverse-myelitis are EXISTING real engines (engine_id resolves to dossier files 2026-05-18); CIDP/tick-paralysis/botulism remain descriptive placeholder sibling_engine_id (no dossier file yet). sibling_engine_id is not registry-resolved (no audit check).