Current issues of ACP Journal Club are published in Annals of Internal Medicine


Risk for bleeding in patients starting antiocoagulants

ACP J Club. 1991 Mar-April;114:55. doi:10.7326/ACPJC-1991-114-2-055

Source Citation

Landefeld CS, McGuire E 3d, Rosenblatt MW. A bleeding risk index for estimating the probability of major bleeding in hospitalized patients starting anticoagulant therapy. Am J Med. 1990;89:569-78. [PubMed ID: 2239976]



To develop and validate a risk index for bleeding in hospitalized patients who were starting anticoagulant therapy.


Cohort study of hospitalized patients.


1 university hospital in Boston (derivation) and 1 in Cleveland (validation).


Index derivation: Data from 2 retrospective samples at 1 hospital were combined and re-analyzed to produce a multivariable estimate of bleeding risk based on baseline features of an inception cohort of 617 hospitalized patients starting long-term (> 10 d) anticoagulants. Validation: The index was applied prospectively to an inception cohort at a second hospital. Physicians' estimates of bleeding risk were obtained at the start of treatment and compared with estimates from the index. For the validation study, the index was applied prospectively to a separate inception cohort of 394 consecutive patients identified from pharmacy records of use of heparin or warfarin.

Assessment of prognostic factors

Data about variables hypothesized to be associated with bleeding were gathered daily.

Main outcome measures

Bleeding was classified according to a bleeding severity index, based on amount, rate, and consequences of bleeding, by assessors who were unaware of individuals' risk factor profiles.

Main results

61 of 1011 patients (6%) developed major bleeding before discharge. 4 independent risk factors were identified for the risk index: the number of specific comorbid conditions; intravenous heparin use in patients aged 60 years or older; a maximum prothrombin time or partial thromboplastin time ratio ≥ 2.0 times control; and worsening liver function during therapy. In the validation study, major bleeding occurred in 2.5% (6 of 235) of those with a low risk for bleeding according to the index, in 15.6% (15 of 96) of those with medium risk, and in 19.0% (12 of 63) of those with high risk. The index did less well in the validation study than in the derivation study (P < 0.005). The index and physicians' estimates had similar accuracy in predicting major bleeding. The combination of the index and the physicians' predictions improved overall accuracy somewhat, with increased sensitivity (P = 0.03) but decreased specificity (P < 0.01) .


Despite differences in baseline characteristics between the population in which the index was developed and that in which it was tested, the index maintained part of its ability to distinguish among subgroups of patients with different bleeding risks on anticoagulants. In predicting major bleeding, the index complemented clinical judgment but did not perform better by itself.

Sources of funding: American Heart Association; Northeast Ohio Affiliate; Research Fund of December 1942.

Address for article reprint: Dr. C.S. Landefeld, UCSF, 3333 California St., Suite 380, San Francisco, CA 94118 USA. Fax: 415-514-0702.


In this study, a previously derived index for predicting a major or minor bleeding event during in-hospital anticoagulation was prospectively validated in another hospital. It is not clear why patients receiving thrombolytic agents in addition to anticoagulants were included, but they constituted only 1% of the population.

The data required to derive the index score are easy to obtain. The scoring system is clearly described, and its explicitness offers a considerable advantage over the use of clinical judgment, which may vary considerably among physicians and institutions. The fact that the index was predictive in two quite different populations suggests that it is probably generalizable to most tertiary care institutions.

The high sensitivity seen with the combination of index score and clinical judgment means that the development of major bleeding can virtually be ruled out in a patient with a low index score and a very low clinical probability of bleeding. The authors sensibly suggest that before instituting anticoagulation in someone with a high index score, the indications for anticoagulation should be reassessed, any other factors that might increase the risk for bleeding should be addressed (for example, concomitant aspirin use), and the level of anticoagulation should be closely monitored.

Finally, it should be pointed out that the high rate of bleeding observed in this study has not been seen with chronic outpatient use of warfarin in patients with myocardial infarction or nonvalvular atrial fibrillation, in whom the yearly risk for major bleeding has been less than 2% (1, 2). Of course, the majority of these patients would have had low index scores.

Andreas Laupacis, MD, MSc
University HospitalLondon, Ontario, Canada


1. Stroke Prevention in Atrial Fibrillation Study Group Investigators. Preliminary report of the Stroke Prevention in Atrial Fibrillation Study. N Engl J Med. 1990;322:863-8.

2. Smith P, Arnesen H, Holme I. The effect of warfarin on mortality and reinfarction after myocardial infarction. N Engl J Med. 1990;323:147-52.