Effect of Clinical Decision Support Alerts on Anticoagulation Management in Atrial Fibrillation

Anticoagulation decisions in atrial fibrillation (AF) depend on balancing stroke and bleeding risk, often guided by CHA2DS2-VASc (a validated clinical score used to estimate stroke risk in patients with atrial fibrillation) and HAS-BLED (a validated clinical score used to estimate bleeding risk in patients treated with anticoagulation) scores. Manual calculation of these scores can be time-consuming and inconsistently performed. This study evaluated whether implementing real-time, electronic health record (EHR)-integrated alerts in a rural academic primary care clinic would influence physician and non-physician provider (NPP) behavior around anticoagulation management. A single-arm observational study was conducted from March 2024 to September 2025 at a West Virginia University (WVU) Family Medicine Clinic. A rules-based engine in Epic calculated risk scores using 1 year of structured data and displayed them within a non-interruptive “Our Practice Advisory” alert. Physician or NPP interaction—defined as initiation of anticoagulation, documentation of rationale, or adding exclusion diagnosis to problem list—was analyzed using chi-square testing. Among 313 patients triggering the alert, 53 (16.9%) were newly started on anticoagulation, 112 (35.8%) had a documented rationale for not initiating therapy, and 2 had the exclusion diagnosis added to their chart. In total, 50.5% of patients had a clinically meaningful interaction with the tool (χ2 = 9.82, p = 0.0017). Across 2,447 encounters, the overall alert success rate was 19.8%, reflecting encounter-level engagement. Common acknowledgment reasons included corrective measures completed, high bleeding risk, recent procedures, and patient refusal. Physician and NPP comments informed iterative refinement, leading to expanded acknowledgment options. Real-time alerts displaying stroke and bleeding risk scores were associated with meaningful physician and NPP engagement, particularly for initiating anticoagulation in high-risk patients. While most interactions reflected review rather than treatment change, the tool appeared to support point-of-care decision-making. These findings support further investigation of EHR-based advisories to improve anticoagulation management in AF.