OpenNeedle Ask your own

Question explored with the scientific record

Answered on this page ↓

Are there better stroke-risk tools than CHA2DS2-VASc for AF?

Sep 7, 2026 · 6 sources used · OpenNeedle synthesis
The short version: CHA2DS2-VASc is the standard tool, but its discrimination is barely better than chance, and several alternatives outperform it for specific populations.

The CHA2DS2-VASc score has a C-statistic of only 0.67 in the large Swedish AF cohort study [4], meaning it discriminates stroke risk only slightly better than a coin flip. In the AFIRE trial subanalysis, the C-statistic was even lower at 0.637 [28]. The score was designed for broad use, not for precision.

The HELT-E2S2 score, validated in a 2025 Japanese study of 362 patients with cardiac implantable devices, outperformed CHA2DS2-VASc with a C-statistic of 0.719 versus 0.700 [8]. The adapted diabetes complications severity index (DCSI) change score, tested in over 81,000 type 2 diabetes patients, achieved an AUROC of 0.79 for predicting AF itself, compared to 0.63 for CHA2DS2-VASc change [15]. For patients with acute ischemic stroke, a composite score combining CHADS2, CHA2DS2-VASc, and neuroimaging features improved the C-statistic from 0.58 to 0.75 [6].

ScorePopulationC-statisticCompared to CHA2DS2-VASc
HELT-E2S2Japanese AF with CIEDs0.7190.700 [8]
Adapted DCSI changeType 2 diabetes0.79 (for AF)0.63 [15]
Composite (CHADS2 + CHA2DS2-VASc + imaging)Acute ischemic stroke0.750.58 [6]
Cognitive tests (MoCA, MMSE)Cardiovascular patientsSuperior for IS predictionNon-inferior for AF [1]

A 2024 study found that simple cognitive tests (MoCA and MMSE) were non-inferior to CHA2DS2-VASc for predicting AF and superior for predicting ischemic stroke, with AUC differences of 0.128 and 0.13 [1]. This suggests that cognitive function captures stroke risk that the standard score misses.

My call: CHA2DS2-VASc is a reasonable starting point but not the best tool for any specific patient. For a fit 69-year-old with prior AF, the HELT-E2S2 score or a cognitive assessment may give a truer picture. Confidence: moderate, because most alternatives have been validated in single populations and need broader replication.

Keep digging

Sources used 6

  1. Usefulness of cognitive assessment in the prediction of documented atrial fibrillation and ischaemic stroke: a cross-sectional study European Journal of Preventive Cardiology (2024) Thin

    A cross-sectional retrospective study of 469 cardiovascular patients evaluating whether cognitive assessments (MoCA and MMSE) can predict documented atrial fibrillation and ischemic stroke, finding MoCA ≤26 and MMSE ≤24–28 associated with AF/IS and showing non-inferiority to CHA…

    DOI: 10.1093/eurjpc/zwae175.100
  2. Evaluation of risk stratification schemes for ischaemic stroke and bleeding in 182 678 patients with atrial fibrillation: the Swedish Atrial Fibrillation cohort study Thin

    This study evaluates the effectiveness of the CHA2DS2-VASc and HAS-BLED risk stratification schemes for predicting ischaemic stroke and bleeding in a large cohort of 182,678 patients with atrial fibrillation in Sweden, revealing significant associations between various risk fact…

    DOI: 10.1093/eurheartj/ehr488
  3. Novel composite scoring system to predict unknown atrial fibrillation in acute ischemic stroke patients Brain Research (2017) Thin

    This study introduces a novel composite scoring system that combines the CHADS2 and CHA2DS2-VASc scores with neuroimaging features to better predict unknown atrial fibrillation in patients with acute ischemic stroke, demonstrating improved predictive performance over traditional…

    DOI: 10.1016/j.brainres.2017.08.005
  4. Validity of HELT-E<sub>2</sub>S<sub>2</sub> Score in Predicting Ischemic Stroke in Japanese Patients With Non-Valvular Atrial Fibrillation With Cardiac Implantable Electronic Devices Circulation Journal (2025) Thin

    This study validates the HELT-E2S2 score as a reliable tool for predicting ischemic stroke in Japanese patients with non-valvular atrial fibrillation who have cardiac implantable electronic devices, demonstrating its superiority over traditional scoring systems like CHADS2 and C…

    DOI: 10.1253/circj.CJ-24-0715
  5. Comparisons of changes in the adapted diabetes complications severity index and CHA2DS2-VASc score for atrial fibrillation risk stratification in patients with type 2 diabetes mellitus: A nationwide cohort study International Journal of Cardiology (2018) primary study Strong

    Adapted diabetes complications severity index (DCSI) change significantly outperforms CHA2DS2-VASc score change for predicting atrial fibrillation in type 2 diabetes patients.

    DOI: 10.1016/j.ijcard.2018.07.010
  6. Rivaroxaban monotherapy versus combination therapy according to patient risk of stroke and bleeding in atrial fibrillation and stable coronary disease: AFIRE trial subanalysis American Heart Journal (2021) Thin

    Rivaroxaban monotherapy reduces major bleeding while maintaining noninferior thromboembolism outcomes compared with rivaroxaban plus antiplatelet therapy in atrial fibrillation with stable coronary disease, with consistent effects across stroke and bleeding risk categories.

    DOI: 10.1016/j.ahj.2021.02.021

Your question next

What do you want to know?

No question is too uncomfortable for the evidence. Bring yours.

Ask your question