Historical collection · First published Nov 6, 2024; added to EM Weekly Oct 27, 2026. This retrospective summary may be superseded by later evidence or guidance.
← All articlesDecision Support Intervention and Anticoagulation for Emergency Department Atrial Fibrillation: The O’CAFÉ Stepped-Wedge Cluster Randomized Clinical Trial
Summary
O’CAFÉ was a pragmatic stepped-wedge cluster randomized trial across 13 community EDs in 9 clusters testing physician education, facility-level audit and feedback, and optional atrial-fibrillation decision support among 1,203 anticoagulation-initiation-eligible adults discharged home with primary AF or flutter. Anticoagulation was initiated by ED discharge or within 30 days in 63.0% of control encounters versus 68.4% during intervention; the adjusted OR was 1.33 (95% CI 0.75-2.35; P=.13), so the intervention did not significantly improve the primary outcome.
Why it matters
Simply making low-intensity decision support available did not reliably change ED anticoagulation initiation. Systems trying to improve stroke prevention after ED AF evaluation may need stronger workflow integration or adoption strategies; the higher initiation rate when clinicians actually used the tool was a nonrandomized per-protocol association and should not be treated as causal evidence.
Study at a glance
- Design
- Randomized trial
- Population
- Adults discharged home from 13 community emergency departments with primary atrial fibrillation or flutter, elevated stroke risk, no recent anticoagulation, and eligibility for oral anticoagulation initiation.
- Sample size
- 1203
- Intervention
- Physician education, facility-specific audit and feedback, and optional RISTRA-AF clinical decision support encouraging shared decision-making, anticoagulation when appropriate, and timely follow-up.
- Comparator
- Usual care before cluster crossover to the intervention.
- Primary endpoint
- Oral anticoagulation initiation at emergency department discharge or within 30 days.
Main findings
- Primary outcome: 244/387 (63.0%) control versus 558/816 (68.4%) intervention; adjusted OR 1.33 (95% CI 0.75-2.35), P=.13.
- Decision support was used in 217/816 intervention encounters (26.6%).
- Per-protocol anticoagulation initiation was 75.6% with decision-support use versus 65.8% without use, P=.008, but this was not a randomized comparison.
- The study did not report a patient-centered long-term stroke or bleeding endpoint; the primary endpoint was anticoagulation initiation.
Limitations
Decision-support use occurred in only 26.6% of intervention encounters, baseline anticoagulation rates were high, and the integrated health-system setting may not generalize to less connected EDs. The trial measured prescribing behavior rather than filled prescriptions, bleeding, ischemic stroke, or other long-term patient outcomes, and per-protocol comparisons are susceptible to selection bias.
Read the primary source
Full text: Open-access full text available.
Original sourceDOI: 10.1001/jamanetworkopen.2024.43097
PMID: 39504024
Material reviewed: full text.
Source retrieved Oct 27, 2026. Historical backfill. JAMA Network Open and PMC open-access full text were reviewed. The publisher and PMC explicitly state published online 2024-11-06; PubMed's bibliographic display shows Nov 4, so the primary publisher's online-publication date was used for eligibility. A targeted DOI-specific search found no correction or retraction.
Corrections & review history
No corrections recorded. Human review: not yet human-reviewed.