Comparison of HAS-BLED and HAS-BED Versus CHADS2

and CHA2DS2VASC Stroke and Bleeding Scores in Patients

With Atrial Fibrillation
Daniela Poli, MDa,*, Emilia Antonucci, ScDb, Vittorio Pengo, MDc, Sophie Testa, MDd, and
Gualtiero Palareti, MDe

Anticoagulation is recommended in patients with atrial fibrillation (AF) for stroke preven-
tion, and the bleeding risk associated suggests the need for a bleeding risk stratification.
HAS-BLED (hypertension, abnormal renal/liver function, stroke, bleeding history or pre-
disposition, labile international normalized ratio (INR), elderly >65 years, drugs/alcohol
concomitantly) score includes “labile INR” referred to quality of anticoagulation. However,
in naïve patients, this item is not available. In addition, stroke and bleeding risk prediction
scores shared several risk factors. The aims of our study were as follows: (1) to evaluate if the
HAS-BLED score in its refined form excluding “labile INR” (HAS-BED [hypertension,
abnormal renal/liver function, stroke, bleeding history or predisposition, elderly, drugs/
alcohol]) is still associated with bleeding risk and (2) to evaluate the predictive ability for
bleeding of both stroke and bleeding prediction models. We followed an inception cohort of
4,579 patients with AF enrolled in the Survey on anticoagulaTed pAtients RegisTer
(NCT02219984). Major bleeds were recorded. During follow-up (7,014 patient-years), 115
patients experienced a major bleeding (MB; rate 1.6 3 100 patient-years). Patients at high
risk were better identified by HAS-BLED and HAS-BED scores with respect to CHADS2
(congestive heart failure, hypertension, age >75 years, diabetes, previous stroke or transient
ischemic attack) and CHA2DS2VASc (congestive heart, failure, hypertension, age
[>75 years], diabetes, stroke/transient ischemic attack, vascular disease, age [65 to 74 years],
female gender). HAS-BLED has a slightly higher c value in comparison to CHADS2 and
CHA2DS2VASc. However, among naïve patients, the predictive ability for hemorrhage of
HAS-BED score is overlapping with CHADS2 and CHA2DS2VASc. In low stroke risk pa-
tients (CHA2DS2VASc [ 0 to 1), only 6 patients are at high bleeding risk, and none of them
experienced MB. In conclusion, in our prospective cohort of patients with AF, we found that
HAS-BLED and HAS-BED scores identify patients at high bleeding risk. However, the
predictive value for MB of HAS-BED used in naïve patients is similar to CHADS2 or
CHA2DS2VASc, suggesting that stroke stratification scores could be sufficient for tailoring
treatment. Ó 2017 Elsevier Inc. All rights reserved. (Am J Cardiol 2017;119:1012e1016)

In the management of stroke risk prevention in patients
with nonvalvular atrial fibrillation (AF), CHADS2 score
(congestive heart failure, hypertension, age >75 years,
diabetes, previous stroke or transient ischemic attack)1 has
been suggested to first stratify stroke risk, and CHA2DS2-
VASc (congestive heart, failure, hypertension, age
[>75 years], diabetes, stroke/transient ischemic attack,
a
Thrombosis Centre, Oncology Department, AziendaOspedaliero-
Universitaria Careggi, Florence, Italy; bStart Register Section, Arianna
Anticoagulazione Foundation, Bologna, Italy; cDepartment of Cardiac
Thoracic and Vascular Sciences, Thrombosis Centre, University of Padua,
Padua, Italy; dHaemostasis and Thrombosis Centre, Laboratory Depart-
ment, Hospital of Cremona, Cremona, Italy; and eCoordinator of the Survey
on anticoagulaTed pAtients-RegisTr, Bologna, Italy. Manuscript received
September 27, 2016; revised manuscript received and accepted December
22, 2016.
See page 1015 for disclosure information.
*Corresponding author: Tel: (0039) 055-7945453; fax: (0039) 055-
7946218.
E-mail address: polida@aou-careggi.toscana.it (D. Poli).
vascular disease, age [65 to 74 years], female gender) to
further stratify patients at low risk.2 These stroke risk scores
have been widely validated, and there is a strong correlation
between the different classes of risk and the occurrence of
ischemic events.3 However, the bleeding risk associated
with anticoagulation suggests the need for a bleeding risk
stratification. The HAS-BLED score (hypertension,
abnormal renal/liver function, stroke, bleeding history or
predisposition, labile international normalized ratio (INR),
elderly >65 years, drugs/alcohol concomitantly) has been
proposed to stratify bleeding risk, and this model is widely
used and popular.3 A recent study compared different
bleeding risk prediction scores, including HAS-BLED, and
found that, although the HAS-BLED score performed better
than other bleeding scores,4 all had a modest performance in
predicting any bleeding. In addition, bleeding and stroke
risk models include hypertension, age, and history of stroke
as “shared” risk factors that could help to explain why
bleeding risk rises with the increase of the stroke risk.5,6
Moreover, the applicability of HAS-BLED in naïve pa-
tients for whom the item “labile INR” is not available for

0002-9149/17/$ - see front matter Ó 2017 Elsevier Inc. All rights reserved. www.ajconline.org
http://dx.doi.org/10.1016/j.amjcard.2016.12.007
 Arrhythmias and Conduction Disturbances/Bleeding Risk Stratification in Atrial Fibrillation Patients 1013

definition could be limited. Therefore, we applied to a Table 1

cohort of patients with AF included in the Survey on anti- Clinical characteristics of patients (n ¼ 4,579)
coagulaTed pAtients RegisTer (START-Register) the HAS- Men 2504 (54.7%)
BLED score in its refined form HAS-BED (hypertension, Median (IQR) age 76 (70-82)
abnormal renal/liver function, stroke, bleeding history or Follow-up period (years) 7014
predisposition, elderly, drugs/alcohol) with the aim to Median (IQR) follow-up period (years) 1.4 (0.6-2.3)
evaluate the predictive ability for bleeding risk of stroke and Heart failure 674 (14.7%)
bleeding risk stratification models. Hypertension 3647 (79.6%)
Diabetes Mellitus 909 (19.9%)
Coronary artery disease 826 (18.0%)
Methods Peripheral artery disease 279 (6.1%)
Moderate renal impairment (eGFR 30-60 ml/min)* 1981 (43.2%)
The START register is an observational, multicenter, Previuos Stroke/TIA 707 (15.4%)
prospective cohort study that includes adults (>18 years) History of major bleeding 157 (3.4%)
who start anticoagulation therapy, whatever the clinical Patients on VKA treatment 3531 (77.2%)
indication for the therapy, the drug and dosage used.7 The Patients on DOACs 1048 (22.8%)
aim of the START register is to collect data on effectiveness Apixaban 360 (34.3%)
and safety of anticoagulant treatments, on determinants of Dabigatran 398 (38.0%)
adverse events in anticoagulated patients, as well as on their Rivaroxaban 290 (27.7%)
Quality of anticoagulant treatment*
quality of life and compliance to treatment. The register has
Time in therapeutic range 67 (54%-77%)
been approved on October 2011 (n ¼ 142/2010/o/oss) by Time above therapeutic range 10 (4%-17%)
the Ethical Committee of the Institution of the Coordinating Time below therapeutic range 20 (12%-31%)
Member (G. P.) (Azienda Ospedaliero-Universitaria, Poli- Concomitant antiplatelet drugs 755 (16.5%)
clinico S. Orsola-Malpighi, Bologna, Italy) and is registered Dual antiplatelet therapy 59 (1.3%)
in Clinical Trials.gov Identifier: NCT02219984. The
participating centers are asked to consecutively include DOAC ¼ direct oral anticoagulant; TIA ¼ transient ischemic attack.
* Available for 3,531 patients on vitamin K antagonist.
patients who start an anticoagulant treatment. Baseline
patient’s clinical features are recorded by participants on
web-based case report forms and include demographic and We used descriptive analysis expressed as median and
clinical characteristics of patients, clinical indication for interquartile range for continuous variables. Following a test
treatment, associated risk factors for thromboembolic com- of statistical normality, analyses were performed using the
plications or bleeding occurring during treatment, laboratory Fisher’s exact test. Incidence rates for bleeding events were
routine data, type of anticoagulant drug used and dose (or calculated as the number of events per 100 patient-years
expected therapeutic range), and use of concomitant drugs. (pt-yrs) of observation. For this calculation, observation
For patients treated with vitamin K antagonists (VKAs), all started at the beginning of follow-up and ended when pa-
INR controls, the subsequent dosing prescriptions and in- tients died, stopped treatment for any reason, or experienced
formation at each visit about possible clinical events and a MB event. All variables included in the analysis were
changes in the medical history are automatically captured collected at baseline. The c statistic, a measure of the area
through informatics. Follow-up of enrolled patients is under the receiver-operating characteristic curve, quantified
mandatory for at least 1 year but is recommended to be the predictive validity of the classification schemes and
indefinite. Participants are required to regularly follow-up tested the hypothesis that these classification schemes per-
all enrolled patients at least quarterly, by phone call or formed significantly better than chance. The SPSS statistical
ambulatory visit. An ambulatory follow-up visit is manda- software package, version 19, for Windows (SPSS, Chi-
tory at least annually. cago, Illinois) was used for data processing.
This registry is an on-going cohort with enrollment
on-going at the time of the present analysis. In this study, we
Results
present results of patients with AF enrolled in the START
register at June 30, 2015. Patients with AF are stratified for We followed an inception cohort of 4,579 patients with
stroke risk evaluation according to CHADS21,8 and AF starting anticoagulation and enrolled in the START
CHA2DS2VASc2 scores, whereas baseline bleeding risk is register, patients characteristics are listed in Table 1. The
evaluated using HAS-BLED score.9 All models have been distribution of patients according to stroke and bleeding risk
categorized in low and high risk. High stroke risk was scores is reported in Table 2. During follow-up, 115 patients
defined for CHADS2 a score 1 and for CHA2DS2VASc a experienced a MB events (rate 1.6  100 pt-yrs), and 13
score 2; high bleeding risk was defined for HAS-BLED hemorrhages (11.3%, rate 0.2  100 pt-yrs) were fatal. The
3. HAS-BLED score considers among the 7 items also distribution of MBs in relation to the considered scores is
the quality of anticoagulation under VKAs treatment. reported in Table 3 (as categorized form).
However, this information is not available for naïve patients The predictive ability for hemorrhage of HAS-BLED
enrolled in our study. For this reason, we calculate this and HAS-BED bleeding scores was compared with the
bleeding score also without the item: “labile INR,” and we predictive ability for hemorrhage of the 2 stroke risk scores
renamed it as HAS-BED score. Major bleeding (MB) events CHADS2 and CHA2DS2VASc, and results are showed in
were defined as recommended by the International Society Table 4 (as continuous form) and Table 5 (as categorized
on Thrombosis and Haemostasis.10 form). HAS-BLED score has a slightly higher c value for
1014 The American Journal of Cardiology (www.ajconline.org)

Table 2 Table 5
Distribution of patients according to stroke and bleeding scores (n ¼ 4,579) Predictive ability for hemorrhage of the scores (categorized)
Score CHADS2 CHA2DS2Vasc HAS-BLED* HASBED† c statistic p value 95% CI

0 361 (7.9%) 100 (2.2%) 93 (2.7%) 160 (3.5%) HAS-BLED 0.59 0.001 0.539-0.643
1 1215 (26.5%) 300 (6.6%) 476 (13.6%) 780 (17.1%) HAS-BED 0.52 0.4 0.468-0.579
2 1643 (35.9%) 682 (14.9%) 1300 (37.2%) 2061 (45.0%) CHADS2 0.54 0.1 0.494-0.596
3 775 (16.9%) 1129 (24.7%) 1068 (30.6%) 1206 (26.5%) CHA2DS2VASc 0.51 0.8 0.455-0.561
4 450 (9.8%) 1165 (25.4%) 454 (13.0%) 305 (6.7%)
5 123 (2.7%) 647 (14.1%) 91 (2.6%) 36 (0.8%)
6 12 (0.3%) 395 (8.6%) 6 (0.3%) 2 (0.0%)
7 136 (3.0%) Table 6
8 20 (0.4%) Distribution of major bleeds in patients (n ¼ 400) at low stroke risk
9 5 (0.1%) (CHA2DS2VASc score 0 to 1) stratified by HAS-BLED and HAS-BED
scores
* Available for 3,494/3,531 patients (99.0%) on VKAs.
† Number of patients Number of major bleeds
Available for 4,550/4,579 patients (99.4%) excluding the item “labile
INR.” HAS-BLED low risk 394/400 (98.5%) 8
HAS-BED low risk 400/400 (100%) 8
HAS-BLED high risk 6/400 (1.5%) 0
Table 3 HAS-BED high risk 0/400 0
Distribution and rate of major bleedings in relation to the scores
(categorized)
Low risk High risk Relative risk P
N N (95% CI) stroke predictive models CHADS2 and CHA2DS2-VASc
(rate x100 pt-yrs) (rate x100 pt-yrs) scores are able to identify patients at high risk for bleeding,
HAS-BLED 45 (1.1) 70 (2.3) 2.0 (1.4-3.0) 0.002
even if HAS-BLED and HAS-BED scores perform better
HAS-BED 57 (1.2) 58 (2.4) 1.9 (1.3-2.8) 0.0006 for that purpose. However, as expected, the predictive
CHADS2 29 (1.2) 86 (1.9) 1.5 (1.0-2.4) 0.05 ability for bleeding risk is modest for all the scores and
CHA2DS2VASc 8 (1.4) 107 (1.7) 1.1 (0.6-2.8) 0.6 among naïve patients is similar for both bleeding and stroke
risk stratification models.
Previous studies suggest that bleeding risk increases
according with the increase of CHADS212 and CHA2DS2-
Table 4
Predictive ability for hemorrhage of the scores (continuous)
VASc scores.5 This observation is not surprisingly because
some key risk factors for ischemic stroke are also risk fac-
c statistic p value 95% CI tors for MB. In fact, the term “shared” risk factors has been
HAS-BLED 0.61 0.000 0.560-0.667 adopted to explain this contemporary influence of some
HAS-BED 0.58 0.002 0.530-0.639 common clinical conditions to both bleeding and thrombotic
CHADS2 0.58 0.002 0.531-0.638 risks.5,6 Several parameters included in risk prediction rules
CHA2DS2VASc 0.56 0.021 0.509-0.618 are “shared” risk factors: age, hypertension, ischemic stroke,
diabetes mellitus, and renal impairment. When the presence
of “shared” risk factors predominates, both risk scores
obviously increase concordantly.
bleeding risk in comparison to CHADS2 and CHA2DS2-
Among the “shared” risk factors, age shows a paradig-
VASc. When we analyze the HAS-BED score, its predic-
matic behavior with a clear increase of both risks in aging.
tive ability for hemorrhage is overlapping that of CHADS2
Fear of bleeding among elderly patients is well known and
and CHA2DS2VASc (Tables 4 and 5). In addition, to
represents an important item that leads to the underuse of
evaluate whether the use of the bleeding risk stratification
anticoagulation, as it has been frequently claimed.13,14
model could help the decision of treating patients at low
However, several data suggest that advanced age is a
stroke risk, we stratify low stroke risk patients identified
more potent risk factor for ischemic stroke than intracerebral
(using CHA2DS2VASc) for both HAS-BLED and
hemorrhage.5,15,16 In contrast, the history of previous stroke
HAS-BED scores (Table 6). Four hundred patients at low
itself is an independent risk factor for both recurrent cerebral
stroke risk were identified; in this group, 98.5% and 100%
ischemia and cerebral bleeding, as it has been outlined by
of patients were classified at low risk for bleeding
several investigators.17e20 As a matter of fact, aging is one
with HAS-BLED and HAS-BED, respectively. Among
of the items indicated in all the available scores.4,21e24
them, 8 MB events were recorded; instead, no bleeding
Analogous considerations could be done for hypertension
events occurred among patients classed at high risk
and diabetes mellitus.
for bleeding.
Among the several bleeding scores available, the HAS-
BLED model has been proposed and validated for applica-
Discussion
tion in daily clinical practice for its simplicity. HAS-BLED
Our study confirms that HAS-BLED score is useful to has become very popular in Europe, as recommended by the
identify patients at high risk for bleeding,11 even when European Cardiology Guidelines.3 In our data, the HAS-
HAS-BED is applied in naïve patients. In addition, the BLED score shows a slightly better performance in
 Arrhythmias and Conduction Disturbances/Bleeding Risk Stratification in Atrial Fibrillation Patients
bleeding prediction with respect to the 2 stroke risk strati-
fication models considered, accordingly with previous
studies.11 This score includes, among the different items, the
“labile INR” to indicate a worst quality of anticoagulation as
a risk factor for bleeding. However, in patients naïve to
anticoagulation, both with VKAs and DOACs (direct oral
anticoagulants), the item “labile INR” is not available for
definition, and to our knowledge, a specific validation of the
score in naïve patients is lacking. In fact, the HAS-BLED
score is usually applied in a different form with respect to
its validation, and this could change its predictive ability for
bleeding. In our data, HAS-BLED score defined as HAS-
BED score is still able to detect patients at high risk for
bleeding, but the predictive ability for bleeding risk
expressed as c statistic is not higher than that of CHADS2
and CHA2DS2VASc scores.
It is widely accepted that patients at high risk for stroke
(CHADS2 >1 and CHA2DS2VASc 2) should be anti-
coagulated for stroke prevention. Instead, uncertainty still
remains on the benefit of treating patients at low stroke
risk, and in this group of patients, a bleeding risk estima-
tion could help the decision. However, when we stratify
our low stroke risk patients in relation to bleeding models,
very few patients at high bleeding risk were identified,
according with the overlapping between the rising of stroke
and bleeding risk. Therefore, again this further stratifica-
tion seems not to facilitate physicians in everyday clinical
settings. This could contribute to explain the still low
adherence in clinical practice to guidelines for anti-
coagulation prescription in patients with AF.25 The rec-
ommendations of ESC guidelines clearly indicate that
“HAS-BLED per se should not be used to exclude patients
from anticoagulation and a HAS-BLED score 3 should
require caution and regular review are recommended, as
well as efforts to correct the potentially reversible risk
factors for bleeding.”3 However, in clinical practice, this
recommendation is frequently misinterpreted, and a sort of
algebraic difference is usually done between the bleeding
and stroke scores for clinical decision. It should be noted
that in the recently published ESC guidelines26 the role of
bleeding risk scores has been reconsidered taking in to
account these aspects.
Our study has potential limitations that should be
acknowledged. First of all, the major part of participating
centers is anticoagulation clinics adherent to the Italian
Federation of Anticoagulation Clinics. Thus, quality of
anticoagulation control is expected to be higher than that
usually seen in patients managed in other settings. As a
matter of fact, in our cohort, the rate of MB and of ce-
rebral bleeding are lower than that observed in the study
of Roldan et al11; nevertheless, in this study, only patients
on stable anticoagulation were considered differently
from our inception cohort study. In addition, patients
included in our cohort have been considered suitable for
starting anticoagulation; therefore, we cannot exclude a
possible selection by the participating anticoagulation
clinic.
Disclosures
The authors have no conflicts of interest to disclose.
1015
Supplementary Data
Supplementary data related with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.amjcard.
2016.12.007.
1. Gage BF, Waterman AD, Shannon W, Boechler M, Rich MW,
Radford MJ. Validation of clinical classification schemes for predicting
stroke: results from the National Registry of Atrial Fibrillation. JAMA
2001;285:2864e2870.
2. Lip GY, Nieuwlaat R, Pisters R, Lane DA, Crijns HJ. Refining clinical
risk stratification for predicting stroke and thromboembolism in atrial
fibrillation using a novel risk factor-based approach: the Euro Heart
Survey on atrial fibrillation. Chest 2010;137:263e272.
3. Camm AJ, Lip GY, De Caterina R, Savelieva I, Atar D, Hohnloser SH,
Hindricks G, Kirchhof P; ESC Committee for Practice Guidelines-
CPG. 2012 focused update of the ESC guidelines for the management
of atrial fibrillation: an update of the 2010 ESC guidelines for the
management of atrial fibrillation. Developed with the special contri-
bution of the European Heart Rhythm Association. Eur Heart J
2012;33:2719e2747.
4. Apostolakis S, Lane DA, Guo Y, Buller H, Lip GY. Performance of the
HEMORR(2)HAGES, ATRIA, and HAS-BLED bleeding risk predic-
tion scores in non-warfarin anticoagulated atrial fibrillation patients.
J Am Coll Cardiol 2013;61:386e397.
5. McGrath ER, Kapral MK, Fang J, Eikelboom JW, Conghaile A,
Canavan M, O’Donnell MJ. Which risk factors are more associated
with ischemic stroke than intracerebral hemorrhage in patients with
atrial fibrillation? Stroke 2012;43:2048e2054.
6. Camm AJ, Savelieva I. “R” for “renal” and for “risk”: refining risk
stratification for stroke in atrial fibrillation. Circulation 2013;127:
169e171.
7. Antonucci E, Poli D, Tosetto A, Pengo V, Tripodi A, Magrini N,
Marongiu F, Palareti G; START-Register. The Italian START-Register
on anticoagulation with focus on atrial fibrillation. PLoS One 2015;10:
e0124719.
8. Gage BF, van Walraven C, Pearce L, Hart RG, Koudstaal PJ,
Boode BS, Petersen P. Selecting patients with atrial fibrillation for
anticoagulation - stroke risk stratification in patients taking aspirin.
Circulation 2004;110:2287e2292.
9. Pisters R, Lane DA, Nieuwlaat R, de Vos CB, Crijns HJ. Lip GYA
novel user-friendly score (HAS-BLED) to assess 1-year risk of major
bleeding in patients with atrial fibrillation: the Euro Heart Survey.
Chest 2010;138:1093e1100.
10. Schulman S, Kearon C; Subcommittee on Control of Anticoagulation
of the Scientific and Standardization Committee of the International
Society on Thrombosis and Haemostasis. Definition of major bleeding
in clinical investigations of antihemostatic medicinal products in non-
surgical patients. J Thromb Haemost 2005;3:692e694.
11. Roldán V, Marín F, Manzano-Fernández S, Gallego P, Vílchez JA,
Valdés M, Vicente V, Lip GY. The HAS-BLED score has better pre-
diction accuracy for major bleeding than CHADS2 or CHA2DS2-
VASc scores in anticoagulated patients with atrial fibrillation.
J Amcollcardiol 2013;62:2199e2204.
12. Poli D, Antonucci E, Marcucci R, Fatini C, Alterini B, Mannini L,
Falciani M, Abbate R, Gensini GF, Prisco D. Risk of bleeding in very
old atrial fibrillation patients on warfarin: relationship with ageing and
CHADS2 score. Thromb Res 2007;121:347e352.
13. Mazzaglia G, Filippi A, Alacqua M, Cowell W, Shakespeare A,
Mantovani LG, Bianchi C, Cricelli C. A national survey of the man-
agement of atrial fibrillation with antithrombotic drugs in Italian pri-
mary care. Thromb Haemost 2010;103:968e975.
14. Holt TA, Hunter TD, Gunnarsson C, Khan N, Cload P, Lip GY. Risk
of stroke and oral anticoagulant use in atrial fibrillation: a cross
sectional survey. Br J Gen Pract 2012;62:e710ee717.
15. Poli D, Antonucci E, Grifoni E, Abbate R, Gensini GF, Prisco D.
Bleeding risk during oral anticoagulation in atrial fibrillation patients
older than 80 years. J Am Coll Cardiol 2009;54:999e1002.
16. Poli D, Antonucci E, Testa S, Tosetto A, Ageno W, Palareti G.
Bleeding risk in very old patients on vitamin K antagonist treatment:
results of a prospective collaborative study on elderly patients fol-
lowed by Italian Centres for Anticoagulation. Circulation 2011;124:
824e829.
1016 The American Journal of Cardiology (www.ajconline.org)
17. Saloheimo P, Juvela S, Hillbom M. Use of aspirin, epistaxis, and un-
treated hypertension as risk factors for primary intracerebral hemor-
rhage in middle-aged and elderly people. Stroke 2001;32:399e404.
18. Gage BF, Birman-Deych E, Kerzner R, Radford MJ, Nilasena DS,
Rich MW. Incidence of intracranial hemorrhage in patients with atrial
fibrillation who are prone to fall. Am J Med 2005;118:612e617.
19. Hart RG, Tonarelli SB, Pearce LA. Avoiding central nervous system
bleeding during antithrombotic therapy: recent data and ideas. Stroke
2005;36:1588e1593.
20. Poli D, Antonucci E, Dentali F, Erba N, Testa S, Tiraferri E, Palareti G.
Recurrence of ICH after resumption of anticoagulation with VK an-
tagonists: CHIRONE study. Neurology 2014;82:1020e1026.
21. Beyth RJ, Quinn LM, Landefeld CS. Prospective evaluation of an in-
dex for predicting the risk of major bleeding in outpatients treated with
warfarin. Am J Med 1998;105:91e99.
22. Gage BF, Yan Y, Milligan PE, Waterman AD, Culverhouse R,
Rich MW, Radford MJ. Clinical classification schemes for predicting
hemorrhage: results from the National Registry of Atrial Fibrillation
(NRAF). Am Heart J 2006;151:713e719.
23. Shireman TI, Mahnken JD, Howard PA, Kresowik TF, Hou Q,
Ellerbeck EF. Development of a contemporary bleeding risk model for
elderly warfarin recipients. Chest 2006;130:1390e1396.
24. Fang MC, Go AS, Chang Y, Borowsky LH, Pomernacki NK,
Udaltsova N, Singer DE. A new risk scheme to predict warfarin-
associated hemorrhage: the ATRIA (Anticoagulation and Risk Fac-
tors in Atrial Fibrillation) Study. J Am Coll Cardiol 2011;58:
395e401.
25. Lip GY, Laroche C, Popescu MI, Rasmussen LH, Vitali-Serdoz L,
Dan GA, Kalarus Z, Crijns HJ, Oliveira MM, Tavazzi L, Maggioni AP,
Boriani G. Improved outcomes with European Society of Cardiology
guideline-adherent antithrombotic treatment in high-risk patients with
atrial fibrillation: a report from the EORP-AF General Pilot Registry.
Europace 2015;17:1777e1786.
26. Kirchhof P, Benussi S, Kotecha D, Ahlsson A, Atar D,
Casadei B, Castella M, Diener HC, Heidbuchel H, Hendriks J,
Hindricks G, Manolis AS, Oldgren J, Popescu BA, Schotten U,
Van Putte B, Vardas P; Authors/Task Force Members. 2016 ESC
guidelines for the management of atrial fibrillation developed in
collaboration with EACTS: the Task Force for the management
of atrial fibrillation of the European Society of Cardiology (ESC)
developed with the special contribution of the European Heart
Rhythm Association (EHRA) of the ESC Endorsed by the Eu-
ropean Stroke Organisation (ESO). Eur Heart J 2016;37:
2893e2962.