Bleed stratification with transfusion thresholds, Forrest-stigma therapy, and failure escalation

(A) Score Rockall first: age, shock, comorbidity, diagnosis, major stigmata and rebleeding predict death — identifying 15% at presentation and 26% after endoscopy as low-risk early discharges.^[1] Transfuse at seven: 921 randomised bleeders with half the restrictive arm never transfused — survival 95% against 91% with less rebleed and fewer adverse events.^[2] Infuse PPI without delaying the scope: high-dose IV PPI as 80 mg bolus plus 8 mg hourly with scope within 24 hours of resuscitation.^[3] Pre-scope omeprazole downgrades lesions: 638 randomised patients with therapy needed in 19.1% against 28.4% — fewer active ulcers, more clean bases, shorter stays.^[7]

(B) Treat spurting, oozing and visible vessels as high risks; watch flat spots and clean bases home on oral PPI.^[3] Reprice oozing honestly: placebo rebleeds at 22.5% spurting, 17.6% clot, 11.3% vessel and 4.9% oozing — with oozing equal on PPI or placebo — so routine high-dose IV PPI after oozing haemostasis needs re-evaluation.^[4]

(C) Predict her failure upfront: hypotension, low haemoglobin, fresh blood, active bleeding and large ulcers predict adrenaline-heater-probe failure — her large actively bleeding ulcer with hypotension was always high-risk.^[8] Clip recurrence: 66-patient randomised trial with further bleeding 57.6% standard against 15.2% over-scope — surgery once per arm.^[13] Clip before embolizing: matched death 22.5% embolization against 5% clips with longer ICU stays — so second failed haemostasis goes to embolization or surgery per the ladder.^[14][3]