Anticoagulation: a clinical prediction rule can identify patients at increased risk of major bleeding

Clinical bottom line (level 1a)

  1. One in eight out-patients on warfarin have a major bleed within 4 years of starting therapy.
  2. A clinical prediction rule can help rank patients for their risk of having a major bleed.
Beyth et al: American Journal of Medicine 1998; 105: 91-99
Expires October 2003

The study

Inception cohort study with objective outcomes, adjusted for confounding factors, validated in an independent set of patients.

Setting: outpatient anticoagulation clinic, university hospital, USA

264 patients (aged mean 60, 53% female) starting out-patient anticoagulation (mainly for venous thromboembolism- acting as a validation cohort for a clinical prediction rule


A proportional hazards model was used to adjust for confounding factors in the derivation group.

?100% followed for 48 months
Outcomes studied:
  • major bleeding: loss of 2 units in 7 days ore less or otherwise life-threatening (e.g. intracranial bleeding)
  • major bleeding in high-risk patients
  • major bleeding in intermediate-risk patients
  • major bleeding in low-risk patients

    • Outpatient bleeding risk index
    • Sum the following risk factors - score one for each positive item
      • aged 65 or more
      • history of stroke
      • history of gastrointestinal bleed
      • recent myocardial infarction, Hct < 30%, Cr > 133 micromol/l or diabetes mellitus
    • Score
      • 0: low-risk
      • 1-2: intermediate-risk
      • 3-4: high-risk

    The evidence

    outcome time to outcome number of patients/total number %
    (95% CI)    		 NNF
    (95% CI)
    major bleeding: 48 months 32/264 12.1%
    (8.18% to 16.1%)    		 8
    (6 to 12)
    major bleeding in high-risk patients 48 months 10/18 53%
    (11% to 97%)    		 2
    (1 to 9)
    major bleeding in intermediate-risk patients 48 months 20/166 12%
    (5% to 19%)    		 8
    (5 to 20)
    major bleeding in low-risk patients 48 months 2/80 3%
    (0% to 8%)    		 40
    (13 to infinity)

    Comments

    1. Intraobserver reliability (K = 0.95) and interobserver reliability (K = 0.87) was high.
    2. Physicians' estimate of the risk of major bleeding did not correlate with the observed frequency.

    Citation

    1. Beyth RJ, Quinn LM, Landefeld CS: prospective evaluation of an index for predicting the risk of major bleeding in outpatients treated with warfarin. American Journal of Medicine 1998; 105: 91-99
    Search Terms: reviewed in EBM journal
    Contributor: Chris Ball and Clare Wotton, October 1999
    Reviewer: Graham Cooke

    Clinical Question.
    Patient patients on warfarin
    Intervention or Exposure clinical prediction rule
    Outcome risk of major bleed