A new model to predict major bleeding in patients with atrial fibrillation using warfarin or direct oral anticoagulants

J'neka S. Claxton, Richard F. MacLehose, Pamela L. Lutsey, Faye L. Norby, Lin Y. Chen, Wesley T. O’Neal, Alanna Chamberlain, Lindsay G.S. Bengtson, Alvaro Alonso

Research output: Contribution to journalArticle

Abstract

Background No scores presently exist to predict bleeding in atrial fibrillation (AF) populations using direct oral anticoagulants (DOACs). We used data from two independent healthcare claims databases to develop and validate a predictive model of major bleeding in a contemporary AF population. Methods Patients with non-valvular AF initiating oral anticoagulation were identified in the MarketScan databases from 2007–2014. Using Cox regression models in 1000 bootstrapped samples, we developed a model that selected variables predicting major bleeding in the first year after anticoagulant initiation. The final model was validated in patients with non-valvular AF in the Optum Clinformatics database in the period 2009–2015. The discriminative ability of existing bleeding scores were individually evaluated and compared with the new bleeding model termed Anticoagulation-specific Bleeding Score (ABS) in both MarketScan and Optum. Results Among 119,083 patients with AF initiating oral anticoagulation in the derivation cohort, 4,030 experienced a bleeding event. The variable selection model identified 15 variables (including individual type of oral anticoagulant) associated with major bleeding. Discrimination of the model was modest [c-statistic 0.68, 95% confidence interval (CI) 0.67–0.69]. The model was subsequently applied to 81,285 AF patients in the validation data set (3,238 bleeding events), showing similar discrimination (c-statistic 0.68, 95% CI 0.67–0.69). In both cohorts, the predictive performance of the ABS was better than the existing models for bleeding prediction in AF. Conclusions We developed a model that uses administrative healthcare data for the identification of AF patients at higher risk of bleeding after initiation of oral anticoagulation, taking into account the lower bleeding risk in DOAC compared to warfarin users.

Original languageEnglish (US)
Article numbere0203599
JournalPLoS One
Volume13
Issue number9
DOIs
StatePublished - Sep 1 2018

Fingerprint

warfarin
anticoagulants
Warfarin
Anticoagulants
Atrial Fibrillation
hemorrhage
mouth
Hemorrhage
Databases
atrial fibrillation
Statistics
health services
confidence interval
statistics
Confidence Intervals
Delivery of Health Care
Identification (control systems)
Proportional Hazards Models
Population

ASJC Scopus subject areas

  • Biochemistry, Genetics and Molecular Biology(all)
  • Agricultural and Biological Sciences(all)

Cite this

Claxton, J. S., MacLehose, R. F., Lutsey, P. L., Norby, F. L., Chen, L. Y., O’Neal, W. T., ... Alonso, A. (2018). A new model to predict major bleeding in patients with atrial fibrillation using warfarin or direct oral anticoagulants. PLoS One, 13(9), [e0203599]. https://doi.org/10.1371/journal.pone.0203599

A new model to predict major bleeding in patients with atrial fibrillation using warfarin or direct oral anticoagulants. / Claxton, J'neka S.; MacLehose, Richard F.; Lutsey, Pamela L.; Norby, Faye L.; Chen, Lin Y.; O’Neal, Wesley T.; Chamberlain, Alanna; Bengtson, Lindsay G.S.; Alonso, Alvaro.

In: PLoS One, Vol. 13, No. 9, e0203599, 01.09.2018.

Research output: Contribution to journalArticle

Claxton, JS, MacLehose, RF, Lutsey, PL, Norby, FL, Chen, LY, O’Neal, WT, Chamberlain, A, Bengtson, LGS & Alonso, A 2018, 'A new model to predict major bleeding in patients with atrial fibrillation using warfarin or direct oral anticoagulants', PLoS One, vol. 13, no. 9, e0203599. https://doi.org/10.1371/journal.pone.0203599
Claxton, J'neka S. ; MacLehose, Richard F. ; Lutsey, Pamela L. ; Norby, Faye L. ; Chen, Lin Y. ; O’Neal, Wesley T. ; Chamberlain, Alanna ; Bengtson, Lindsay G.S. ; Alonso, Alvaro. / A new model to predict major bleeding in patients with atrial fibrillation using warfarin or direct oral anticoagulants. In: PLoS One. 2018 ; Vol. 13, No. 9.
@article{b25b3a8afbca4bdda32d2d6d476bf166,
title = "A new model to predict major bleeding in patients with atrial fibrillation using warfarin or direct oral anticoagulants",
abstract = "Background No scores presently exist to predict bleeding in atrial fibrillation (AF) populations using direct oral anticoagulants (DOACs). We used data from two independent healthcare claims databases to develop and validate a predictive model of major bleeding in a contemporary AF population. Methods Patients with non-valvular AF initiating oral anticoagulation were identified in the MarketScan databases from 2007–2014. Using Cox regression models in 1000 bootstrapped samples, we developed a model that selected variables predicting major bleeding in the first year after anticoagulant initiation. The final model was validated in patients with non-valvular AF in the Optum Clinformatics database in the period 2009–2015. The discriminative ability of existing bleeding scores were individually evaluated and compared with the new bleeding model termed Anticoagulation-specific Bleeding Score (ABS) in both MarketScan and Optum. Results Among 119,083 patients with AF initiating oral anticoagulation in the derivation cohort, 4,030 experienced a bleeding event. The variable selection model identified 15 variables (including individual type of oral anticoagulant) associated with major bleeding. Discrimination of the model was modest [c-statistic 0.68, 95{\%} confidence interval (CI) 0.67–0.69]. The model was subsequently applied to 81,285 AF patients in the validation data set (3,238 bleeding events), showing similar discrimination (c-statistic 0.68, 95{\%} CI 0.67–0.69). In both cohorts, the predictive performance of the ABS was better than the existing models for bleeding prediction in AF. Conclusions We developed a model that uses administrative healthcare data for the identification of AF patients at higher risk of bleeding after initiation of oral anticoagulation, taking into account the lower bleeding risk in DOAC compared to warfarin users.",
author = "Claxton, {J'neka S.} and MacLehose, {Richard F.} and Lutsey, {Pamela L.} and Norby, {Faye L.} and Chen, {Lin Y.} and O’Neal, {Wesley T.} and Alanna Chamberlain and Bengtson, {Lindsay G.S.} and Alvaro Alonso",
year = "2018",
month = "9",
day = "1",
doi = "10.1371/journal.pone.0203599",
language = "English (US)",
volume = "13",
journal = "PLoS One",
issn = "1932-6203",
publisher = "Public Library of Science",
number = "9",

}

TY - JOUR

T1 - A new model to predict major bleeding in patients with atrial fibrillation using warfarin or direct oral anticoagulants

AU - Claxton, J'neka S.

AU - MacLehose, Richard F.

AU - Lutsey, Pamela L.

AU - Norby, Faye L.

AU - Chen, Lin Y.

AU - O’Neal, Wesley T.

AU - Chamberlain, Alanna

AU - Bengtson, Lindsay G.S.

AU - Alonso, Alvaro

PY - 2018/9/1

Y1 - 2018/9/1

N2 - Background No scores presently exist to predict bleeding in atrial fibrillation (AF) populations using direct oral anticoagulants (DOACs). We used data from two independent healthcare claims databases to develop and validate a predictive model of major bleeding in a contemporary AF population. Methods Patients with non-valvular AF initiating oral anticoagulation were identified in the MarketScan databases from 2007–2014. Using Cox regression models in 1000 bootstrapped samples, we developed a model that selected variables predicting major bleeding in the first year after anticoagulant initiation. The final model was validated in patients with non-valvular AF in the Optum Clinformatics database in the period 2009–2015. The discriminative ability of existing bleeding scores were individually evaluated and compared with the new bleeding model termed Anticoagulation-specific Bleeding Score (ABS) in both MarketScan and Optum. Results Among 119,083 patients with AF initiating oral anticoagulation in the derivation cohort, 4,030 experienced a bleeding event. The variable selection model identified 15 variables (including individual type of oral anticoagulant) associated with major bleeding. Discrimination of the model was modest [c-statistic 0.68, 95% confidence interval (CI) 0.67–0.69]. The model was subsequently applied to 81,285 AF patients in the validation data set (3,238 bleeding events), showing similar discrimination (c-statistic 0.68, 95% CI 0.67–0.69). In both cohorts, the predictive performance of the ABS was better than the existing models for bleeding prediction in AF. Conclusions We developed a model that uses administrative healthcare data for the identification of AF patients at higher risk of bleeding after initiation of oral anticoagulation, taking into account the lower bleeding risk in DOAC compared to warfarin users.

AB - Background No scores presently exist to predict bleeding in atrial fibrillation (AF) populations using direct oral anticoagulants (DOACs). We used data from two independent healthcare claims databases to develop and validate a predictive model of major bleeding in a contemporary AF population. Methods Patients with non-valvular AF initiating oral anticoagulation were identified in the MarketScan databases from 2007–2014. Using Cox regression models in 1000 bootstrapped samples, we developed a model that selected variables predicting major bleeding in the first year after anticoagulant initiation. The final model was validated in patients with non-valvular AF in the Optum Clinformatics database in the period 2009–2015. The discriminative ability of existing bleeding scores were individually evaluated and compared with the new bleeding model termed Anticoagulation-specific Bleeding Score (ABS) in both MarketScan and Optum. Results Among 119,083 patients with AF initiating oral anticoagulation in the derivation cohort, 4,030 experienced a bleeding event. The variable selection model identified 15 variables (including individual type of oral anticoagulant) associated with major bleeding. Discrimination of the model was modest [c-statistic 0.68, 95% confidence interval (CI) 0.67–0.69]. The model was subsequently applied to 81,285 AF patients in the validation data set (3,238 bleeding events), showing similar discrimination (c-statistic 0.68, 95% CI 0.67–0.69). In both cohorts, the predictive performance of the ABS was better than the existing models for bleeding prediction in AF. Conclusions We developed a model that uses administrative healthcare data for the identification of AF patients at higher risk of bleeding after initiation of oral anticoagulation, taking into account the lower bleeding risk in DOAC compared to warfarin users.

UR - http://www.scopus.com/inward/record.url?scp=85053105323&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85053105323&partnerID=8YFLogxK

U2 - 10.1371/journal.pone.0203599

DO - 10.1371/journal.pone.0203599

M3 - Article

C2 - 30199542

AN - SCOPUS:85053105323

VL - 13

JO - PLoS One

JF - PLoS One

SN - 1932-6203

IS - 9

M1 - e0203599

ER -