What Is the Best Way to Predict Stroke Risk in General Practice?

For general practitioners, predicting stroke risk is most salient in patients with atrial fibrillation (AF), because risk stratification guides decisions about prescribing anticoagulant drugs. Stroke risk prediction may also identify AF-free patients who would benefit from intensive intervention to modify other stroke risk factors such as hypertension.

To develop a clinical stroke prediction score for patients with and without AF, investigators prospectively examined electronic data from 676 U.K. primary care practices. They derived the QStroke score from 3.5 million patients not receiving anticoagulant drugs and free of stroke or transient ischemic attack. The investigators validated the score in 1.9 million similar patients, and compared its performance to the Framingham stroke score in patients without AF and to the CHADS₂ and CHA₂DS₂VASc scores in those with AF.

The 10-year absolute risk for stroke ranged from <1% in those in the lowest QStroke decile to about 15% in those in the highest decile. In patients without AF, the QStroke outperformed the Framingham score. In those with AF, the QStroke score clearly outperformed CHADS₂ and was somewhat better than CHA₂DS₂VASc, especially in identifying truly low-risk patients. Cutoffs could be found for both QStroke (score ≥20th percentile; 80% classified as high-risk) and CHA₂DS₂VASc (score ≥1; 85% classified as high-risk) that had >95% sensitivity for predicting stroke during follow-up, but with low specificity (QStroke, 22.1%; CHA₂DS₂VASc, 16.4%).