A simple and easily implemented risk model to predict 1-year ischemic stroke and systemic embolism in Chinese patients with atrial fibrillation

Atrial fibrillation (AF)—the most common irregular heart rhythm—puts millions at risk of devastating stroke. Anticoagulants (blood thinners) are life-saving for high-risk patients, but they also carry serious bleeding risks. The problem? Current guidelines recommend anticoagulants for 90% of AF patients, even though only 6–8% die from stroke. For the other 90%, is there a way to avoid unnecessary meds without raising stroke risk?

A 2021 study from the China Atrial Fibrillation (China-AF) Registry offers a solution: a simplified risk score that identifies more low-risk patients while still catching most strokes. Led by researchers from Beijing Anzhen Hospital, Capital Medical University, and Ping An Health Technology, the study focused on 6,601 Chinese AF patients not on anticoagulants or catheter ablation at baseline. The goal? Create an easy-to-use tool to predict 1-year risk of ischemic stroke or systemic embolism (blood clots that block blood flow).

The Problem with Current Risk Scores

Doctors rely on the CHA2DS2-VASc score—a guideline-recommended tool—to assess stroke risk in AF. It assigns points for factors like heart failure, hypertension, diabetes, age, and prior stroke. But here’s the catch: only fewer than 10% of AF patients qualify as “low-risk” (score 0 for men, 1 for women). That means 90% are told to take anticoagulants—even though many may face more harm than benefit from bleeding.

The China-AF team wanted a better way. Using extreme gradient boosting (XGBoost)—a machine learning algorithm that sorts data to find the most powerful predictors—they analyzed 44 clinical variables to see which truly drove stroke risk.

The CAS Model: 3 Simple Factors That Matter Most

The algorithm zeroed in on three key risk factors that accounted for 73% of stroke risk:

Congestive heart failure or left ventricular ejection fraction (LVEF) <55% (a measure of heart pump function)
Age >65 years
Prior stroke

Together, these make up the CAS score (named for Congestive heart failure, Age, Prior stroke). Points are assigned as follows:

1 point for heart failure/LVEF <55%
1 point for age >65
2 points for prior stroke

Add them up: a CAS score of 0 means low risk, while scores ≥1 mean higher risk.

How Well Does CAS Work?

Over 1 year of follow-up, 163 patients (2.5%) had a stroke or systemic embolism. Here’s how CAS performed:

Low-risk group (CAS 0): 30.8% of patients (2,033 total) with a 1-year stroke risk of 0.81% (95% confidence interval: 0.41–1.19%).
CHA2DS2-VA comparison: The standard score (without sex) only labeled 15.2% of patients as low-risk—half as many as CAS. And even those “low-risk” patients had a slightly higher stroke risk (1.01%).

CAS also outperformed CHA2DS2-VA in discrimination (how well it distinguishes high- vs. low-risk patients): its C-statistic (a measure of prediction accuracy) was 0.69 vs. 0.66 for CHA2DS2-VA. That means CAS is better at spotting who will (and won’t) have a stroke.

Perhaps most importantly, CAS improved net reclassification by 12.2%. In plain terms: it moved more patients into the right risk group—so fewer low-risk people get unnecessary anticoagulants, and more high-risk people get the meds they need.

Why These Factors?

The CAS model prioritizes factors with the strongest links to stroke:

Prior stroke: The biggest red flag—patients who’ve already had a stroke are 2.5x more likely to have another.
Heart failure: Even in young AF patients, a weak heart creates a “hypercoagulable” state (thicker blood) that fuels blood clots.
Age >65: Stroke risk rises sharply with age, especially in AF.

Notably, hypertension, diabetes, and vascular disease—key factors in CHA2DS2-VASc—didn’t make the top 10. The researchers suspect this is because control of these conditions (e.g., well-managed blood sugar or pressure) matters more than just having the diagnosis.

What Does This Mean for Patients and Doctors?

The CAS score’s biggest win is precision: by focusing on three high-impact factors, it identifies 30% of AF patients as truly low-risk—twice as many as current scores. For these patients, skipping anticoagulants is safer: their 1-year stroke risk is under 1%, which is lower than the bleeding risk of many blood thinners.

For doctors, CAS is a quick, easy tool—no complicated calculations, no need to track dozens of factors. It helps balance the benefit of stroke prevention against the harm of bleeding, a key challenge in AF care.

Limitations to Keep in Mind

Like all studies, this one has caveats:

Population-specific: The cohort included only Chinese patients. More research is needed to confirm CAS works in other ethnic groups.
Missing data: The study didn’t collect info on AF burden (how often the heart rhythm is irregular) or biomarkers (like troponin) that might add predictive value.
Calibration: The team didn’t test how well CAS “matches” predicted vs. actual risk in a split sample—though bootstrapping (1000 repeats) helped estimate confidence intervals.

The Bottom Line

The CAS model is a simpler, more accurate alternative to current stroke risk scores for Chinese AF patients. By focusing on the factors that truly drive stroke risk, it helps doctors make smarter decisions about anticoagulants—sparing thousands from unnecessary bleeding while still protecting those at highest risk.

For patients, this means fewer trips to the pharmacy, less worry about bleeding, and more trust that their treatment is tailored to their risk—not a one-size-fits-all score.

Chao Jiang, Tian-Ge Chen, Xin Du, and colleagues conducted this research at Beijing Anzhen Hospital, Capital Medical University; Ping An Health Technology; and the Heart Health Research Center. The study followed the Declaration of Helsinki, was approved by the Beijing Anzhen Hospital Ethics Committee, and is registered at the Chinese Clinical Trial Registry (ChiCTR-OCH-13003729).

doi.org/10.1097/CM9.0000000000001515

Was this helpful?

0 / 0