Major bleeding risk on anticoagulation.
Applies to / doesn’t apply to3
Applies to: Atrial fibrillation. Valvular AF and mitral stenosis were excluded from the derivation cohort.
- Valvular atrial fibrillation or mitral stenosis — both excluded from the derivation cohort.
- End-stage kidney disease.
- Active cancer.
Systolic BP above 160 mmHg. A different question from CHA₂DS₂-VASc’s history of hypertension, so it is asked separately and never carried.
Chronic dialysis, renal transplant, or serum creatinine ≥200 µmol/L (≥2.26 mg/dL).
Cirrhosis, or bilirubin above 2× the upper limit of normal with AST, ALT or ALP above 3×.
Previous stroke, particularly lacunar. TIA and embolism are not counted here.
Prior major bleeding, or a predisposition to bleed. The derivation study operationalised this as anaemia, which was the data available to it.
Unstable or high INR, or time in therapeutic range under 60%. Structurally does not apply on a DOAC — the derivation cohort predates them entirely.
Concomitant antiplatelet agents or non-steroidal anti-inflammatory drugs.
≥8 UK units/week (~64 g ethanol; ≈4–5 US standard drinks)
What this score estimates
Estimates the risk of major bleeding within one year in patients with atrial fibrillation, most of whom were taking a vitamin K antagonist. Nine items, one point each. A major bleed was intracranial bleeding, bleeding requiring hospitalisation, a haemoglobin fall greater than 2 g/dL, or bleeding requiring transfusion — excluding haemorrhagic stroke. The paper prints this threshold as "2 g/L", which is a typo: a 2 g/L fall is clinically trivial, and the ISTH 2005 definition it cites specifies 2 g/dL.
Pearls & pitfalls6
- Discrimination is modest. The derivation cohort reported a C statistic of 0.72; pooled across 39 later validations it is 0.63 (95% CI 0.61–0.65, Gao 2021), and guideline supporting text cites figures as low as 0.58. Calibration is imperfect too: Zhu 2015 found the score under-predicts bleeding in the middle of its range.
- Derived in 2003–2004, entirely before direct oral anticoagulants, so labile INR structurally does not apply on a DOAC. Scores intended for DOAC-treated patients exist, but the evidence does not currently establish that any of them predicts better than this one.
- When labile INR is scored absent because the patient is on a DOAC — where the item structurally does not apply — the resulting total is not comparable to the published per-score bleeding rates, which were derived in vitamin K antagonist–era patients for whom the item could be present.
- Of 5,272 patients discharged alive, 3,456 had known one-year bleeding follow-up — 25% missing, acknowledged by the authors.
- Rates above a score of 5 are not published: score 6 had two patients and no bleeds, and scores 7–9 had no patients at all.
- Performs poorly in end-stage kidney disease and in active cancer, both of which carry bleeding risk the nine items do not capture.
What the guidelines say2
States that bleeding risk scores should not be used in isolation to determine eligibility for oral anticoagulation, but to identify and modify bleeding risk factors.
View guideline source ↗Recommends assessment and management of modifiable bleeding risk factors, and no longer names a specific bleeding score.
Guideline positions are stated as published. Curie.MD Calc does not publish treatment direction.
Where this comes from5
- Pisters R, Lane DA, Nieuwlaat R, et al. Chest. 2010;138(5):1093-1100. View source ↗
- Lip GYH, Frison L, Halperin JL, Lane DA. J Am Coll Cardiol. 2011;57(2):173-80. View source ↗
- Schulman S, Kearon C. Definition of major bleeding in clinical investigations of antihaemostatic medicinal products in non-surgical patients. J Thromb Haemost. 2005;3(4):692-4. View source ↗
- Gao X, Cai X, Yang Y, et al. Front Cardiovasc Med. 2021;8:757087. View source ↗
- Zhu W, He W, Guo L, Wang X, Hong K. Clin Cardiol. 2015;38(9):555-61. View source ↗