Comparison of ATRIA, CHADS2, and CHA2DS2-VASc Stroke Risk Scores
What are the predictive abilities of the ATRIA, CHADS2, and CHA2DS2-VASc risk scores for ischemic stroke in atrial fibrillation (AF) among a national Swedish population?
Using a national registry of all Swedish hospitalized and outpatient treatment episodes between July 2005 and December 2010, patients with a diagnosis of AF were identified. ATRIA, CHADS2, and CHA2DS2-VASc risk scores were calculated based on ICD-10 codes. The predictive ability of each score was compared using the C statistic (area under the receiver operator curve) and the net reclassification index (NRI).
Among 152,153 AF patients who did not receive warfarin anticoagulation, 11,053 ischemic stroke events were observed (3.2%/year) over a mean of 2.23 years. Using the entire point system for each risk score, ATRIA had a C statistic of 0.708 (95% confidence interval [CI], 0.704-0.713), which was significantly better than CHADS2 (0.690; 95% CI, 0.685-0.695) and CHA2DS2-VASc (0.694; 95% CI, 0.690-0.700). Using standard cut points for low/moderate/high ischemic stroke risk, the C statistic for ATRIA was 0.668 (95% CI, 0.664-0.595), which was better than CHADS2 (0.663; 95% CI, 0.658-0.668) and CHA2DS2-VASc (0.593; 95% CI, 0.591-0.595). NRI favored the use of ATRIA (0.16; 95% CI, 0.14-0.17) over CHADS2 and over CHA2DS2-VASc (0.21; 95% CI, 0.20-0.23).
The authors concluded that the ATRIA score predicted ischemic stroke risk better than the CHADS2 or CHA2DS2-VASc score in this Swedish population of AF patients without anticoagulation.
Despite frequent use by clinicians, this study highlights the limited predictive ability for the various stroke risk scores used to risk stratify AF patients. While demonstrating improved predictive ability based on both the C statistic and NRI, the ATRIA score remains more complex to calculate than the CHADS2 and CHA2DS2-VASc scores, thereby limiting its use. Additionally, given that guidelines have endorsed the use of CHA2DS2-VASc scores (and previously CHADS2 scores) to determine the need for systemic anticoagulation, most practitioners will likely continue to use those scoring systems routinely. However, this study suggests that when a clinician or patient’s goal is to estimate an individual’s stroke risk with AF, use of the ATRIA stroke risk score (if appropriately optimized to the local population) might be the most accurate approach.
< Back to Listings