Impact of an atrial fibrillation decision support tool on thromboprophylaxis for atrial fibrillation

Research output: Contribution to journalArticlepeer-review


  • Mark H. Eckman
  • Ruth E. Wise
  • Barbara Speer
  • Megan Sullivan
  • Nita Walker
  • Brett Kissela
  • Matthew L. Flaherty
  • Dawn Kleindorfer
  • Robert Ireton
  • Dave Hoskins
  • Brett M. Harnett
  • Carlos Aguilar
  • Anthony C. Leonard
  • Lora Arduser
  • Dylan Steen
  • Alexandru Costea
  • John Kues

Colleges, School and Institutes


Background: Appropriate thromboprophylaxis for patients with atrial fibrillation (AF) remains a national challenge.

Methods: We hypothesized that provision of decision support in the form of an Atrial Fibrillation Decision Support Tool (AFDST) would improve thromboprophylaxis for AF patients. We conducted a cluster randomized trial involving 15 primary care practices and 1,493 adults with nonvalvular AF in an integrated health care system between April 2014 and February 2015. Physicians in the intervention group received patient-level treatment recommendations made by the AFDST. Our primary outcome was the proportion of patients with antithrombotic therapy that was discordant from AFDST recommendation.

Results: Treatment was discordant in 42% of 801 patients in the intervention group. Physicians reviewed reports for 240 patients. Among these patients, thromboprophylaxis was discordant in 63%, decreasing to 59% 1 year later (P = .02). In nonstratified analyses, changes in discordant care were not significantly different between the intervention group and control groups. In multivariate regression models, assignment to the intervention group resulted in a nonsignificant trend toward decreased discordance (P = .29), and being a patient of a resident physician (P = .02) and a higher HAS-BLED score predicted decreased discordance (P = .03), whereas female gender (P = .01) and a higher CHADSVASc score (P = .10) predicted increased discordance.

Conclusions: Among patients whose physicians reviewed recommendations of the decision support tool discordant therapy decreased significantly over 1 year. However, in nonstratified analyses, the intervention did not result in significant improvements in discordant antithrombotic therapy.


Original languageEnglish
Pages (from-to)17-27
JournalAmerican Heart Journal
Early online date19 Feb 2016
Publication statusPublished - Jun 2016