Development and validation of a Screening Tool to Evaluate and Warrant Anticoagulation Treatment prior to Discharge in inpatients with Atrial Fibrillation (STEWARD-AF).

Objective: Anticoagulation is highly effective for stroke prevention in atrial fibrillation (AF), reducing the risk by about 64%. Despite overwhelming evidence in support of anticoagulation, up to 40% of AF patients remain untreated. We aimed to develop and validate STEWARD-AF: a Screening Tool to Evaluate and Warrant Anticoagulation Treatment prior to Discharge in Atrial Fibrillation.

Materials And Methods: STEWARD-AF was developed by integrating information extracted from the electronic health record (EHR). A stepwise decision process was applied, based on AF diagnosis, estimated CHADS-VASc-score and anticoagulant use. A priority score was assigned accordingly, ranging from 0 (no risk) to 5 (highest risk of undertreatment). A cross-sectional study was performed to assess the accuracy of STEWARD-AF. Criterion and tool validity were ascertained by determining sensitivity and specificity, compared to a manual check of the EHR in an inpatient sample (n = 800). Consistency regarding the priority score was determined by estimating Cohen's kappa.

Results: A tool to screen for un(der)treated AF was developed and embedded into the EHR. Sensitivity and specificity for AF diagnosis were 98.4% and 87.6%, respectively. Overall sensitivity and specificity for identification of a CHADS-VASc-score ≥ 2 was 97.7% and 72.7%. Sensitivity and specificity to determine the presence of anticoagulant treatment was at least 87.8% and 97.1% There was good agreement for the priority score (κ 0.74 (unweighted); 0.66 (weighted)).

Conclusions: STEWARD-AF was able to identify untreated AF inpatients reliably and with a high sensitivity. Nearly no patients were missed. We will now implement this AF-screening tool in clinical practice to improve the use of anticoagulation and reduce the risk of stroke.