Richard H. White, MD
In patients receiving oral anticoagulant therapy, how well do scores predict major bleeding? How do they compare with each other and with subjective physician assessment?
Academic hospital, Switzerland.
515 internal medicine patients ≥ 18 years of age (median age 71 y, 64% men) receiving oral anticoagulant therapy (acenocoumarol or phenprocoumon) for any indication at hospital discharge or at presentation in the ambulatory clinic. 96% of patients were hospitalized at enrollment, and 64% started anticoagulant therapy ≥ 3 months before.
Prediction scores were Outpatient Bleeding Risk Index (OBRI) (age ≥ 65 y; stroke; gastrointestinal bleeding; recent myocardial infarction, anemia, diabetes, or creatinine > 1.5 mg/dL), Kuijer (age ≥ 60 y, female, cancer), Shireman (age ≥ 70 y; female; gastrointestinal bleeding > 10 d or < 10 d; anemia; diabetes; alcohol or drug abuse; antiplatelet therapy), modified HEMORR2HAGES (age ≥ 75 y; gastrointestinal bleeding; anemia; glomerular filtration rate [GFR] < 30 mL/min or hepatic disease; cancer; hypertension; alcohol abuse; low platelet count; fall risk), Registry of Patients with Venous Thromboembolism (RIETE) (age ≥ 75 y, major bleeding < 15 d, anemia, creatinine > 1.2 mg/dL, cancer, pulmonary embolism), modified HAS-BLED (age ≥ 65 y, stroke, gastrointestinal bleeding, creatinine ≥ 2.3 mg/dL, hepatic disease, hypertension, alcohol abuse, drug abuse), Anticoagulation and Risk Factors in Atrial Fibrillation (ATRIA) (anemia, GFR < 30 mL/min, age ≥ 75 y, prior bleeding, hypertension), and physician's subjective assessment of risk at discharge (0% to 100%).
First-time major bleeding event (fatal bleeding, with death within 7 d of bleeding episode and no alternative cause of death; symptomatic bleeding in a critical area or organ; or ≥ 20 g/L decrease in hemoglobin level or transfusion of ≥ 2 units of whole or red blood cells).
35 first-time major bleeding events occurred during the study; 12-month cumulative incidence was 6.8%. C-statistics ranged from 0.54 to 0.61 and did not differ among the 7 scores (P = 0.84) or between the scores and physician assessment (P = 0.94) (Table). Only 1 score (ATRIA) predicted major bleeding better than chance (Table).
In patients receiving anticoagulant therapy, performance of clinical prediction scores for bleeding was poor (c-statistic ≤ 0.61) and did not differ for 7 scores and physician assessment.
Clinical prediction scores for bleeding in patients receiving oral anticoagulant therapy*
*Scale descriptions and expanded abbreviations can be found under “Description of prediction guides”; other abbreviations defined in Glossary.
†Area under the receiver-operating characteristic curve.
‡Significantly better than chance.
White RH. Scores poorly predict major bleeding (c-statistics ≤ 0.61) during oral anticoagulant therapy. Ann Intern Med. ;158:JC13. doi: 10.7326/0003-4819-158-6-201303190-02013
Download citation file:
Published: Ann Intern Med. 2013;158(6):JC13.
Hematology/Oncology, Hospital Medicine, Venous Thromboembolism.
Results provided by:
Copyright © 2018 American College of Physicians. All Rights Reserved.
Print ISSN: 0003-4819 | Online ISSN: 1539-3704
Conditions of Use