Reading the CHA₂DS₂-VASc score

CHA₂DS₂-VASc is a stroke-risk assessment for patients with atrial fibrillation. It estimates the annual risk of ischaemic stroke in people with non-valvular AF who are not taking an anticoagulant, and it runs from 0 to 9. It was refined by Lip and colleagues in 2010 from the earlier CHADS₂ score to better separate genuinely low-risk patients from the rest.¹

The score is a points model built from routinely available clinical facts, so it is quick to apply and easy to reproduce. It is worth holding one distinction in mind from the start: a points score sorts patients into risk tiers with published average rates. It is not a precise prediction for the individual in front of you, and the guidelines are explicit that the observed rate at any given score varies from one population to another.²³

Each letter in the name maps to a criterion with a fixed point value. The table lists them with the definitions used in Curie Calc, which follow the Lip 2010 criteria as reproduced in the 2023 ACC/AHA guideline. Count one age band, and count the stroke/TIA category once.

The two age rows are alternatives, not additions: a patient is under 65 (0), 65 to 74 (1), or 75 or older (2). The maximum total is 9. In Curie Calc, stroke and TIA/embolism are entered as separate questions, because a bleeding-risk score such as HAS-BLED counts neither TIA nor embolism; the tool still adds only a single 2 points for the combined stroke/TIA category.¹³

These are the points where the same patient can be scored two ways depending on how a criterion is read. None of them is exotic; they simply reward a second look.

Does controlled hypertension still count?

Yes. History matters, even when today's blood pressure is controlled. The hypertension criterion is met by a documented history of readings over 140/90 mmHg on at least two occasions, or by being on antihypertensive treatment. A single elevated reading in clinic is not the same as a documented history, and a normal reading today does not remove a history that is already established.¹³

What qualifies for the heart-failure point?

The heart-failure point is met by clinical heart failure or by objective evidence of moderate-to-severe left-ventricular dysfunction. A reduced ejection fraction is not required, so heart failure with preserved ejection fraction can still meet the criterion when the clinical picture supports it. The point rests on the diagnosis, not on an ejection-fraction number in isolation.¹⁴

What counts as vascular disease?

Vascular disease covers coronary artery disease (prior myocardial infarction, angina, prior PCI, or CABG) and peripheral artery disease. Guideline definitions differ at the edges: the 2024 ESC guideline also counts complex aortic plaque on imaging (mobile, ulcerated, pedunculated, or 4 mm or thicker), and Curie Calc includes that item. This is one place where the definition you use can change the score for a plaque-only patient.³⁴

Do stroke and TIA contribute separate points?

No. They share a single category worth 2 points. A patient with both a prior ischaemic stroke and a prior TIA scores 2 for that category, not 4. Curie Calc asks about them separately for a different reason (a bleeding score counts neither TIA nor embolism), but adds only one 2-point value.¹³

What changes at ages 65 and 75?

Age is scored in bands. Under 65 adds nothing, 65 to 74 adds 1 point, and 75 or older adds 2. The boundaries are the interpretation trap: a patient the day before a 75th birthday sits in the 65-to-74 band, and a day later moves to the 2-point band.¹³

Two guideline bodies now describe the score differently, and the difference is not cosmetic. Name the score and the guideline framework together when you report a result.

The 2023 ACC/AHA guideline keeps the point for female sex and applies a sex-specific interpretation, describing the relevant thresholds separately for men and women. The 2024 ESC guideline moves to CHA₂DS₂-VA, which drops the sex criterion entirely and scores 0 to 8; the stated rationale is that sex acts as an age-dependent risk modifier rather than an independent risk factor across the range.³⁴

A CHA₂DS₂-VASc result therefore still includes the female-sex point by design; it has not been "adjusted" for sex. The practical consequence is that a number on its own is ambiguous. A "3" that includes a sex point and a CHA₂DS₂-VA "2" can describe the very same patient, so the framework has to travel with the number.³⁴

Guideline context

The published guidelines pair each score with an approach to risk rather than a single cutoff that fits every patient. The 2023 ACC/AHA guideline discusses annual stroke risk by score, notes an intermediate-risk group, and frames its thresholds separately for men and women. This article summarizes that named guidance and links to it directly; it does not restate it as a rule for a specific patient.³

Why annual stroke-risk percentages differ

Published stroke rates at the same score are not interchangeable, because they depend on the cohort studied, the endpoint counted, and whether patients were anticoagulated. A cohort-derived rate is not an exact individual forecast. Curie Calc displays rates from Friberg's non-anticoagulated subgroup (about 90,000 patients), which is why the figures describe untreated risk; that subgroup and the study's full cohort are different denominators.²

The 2023 ACC/AHA guideline makes the same point from the other direction: at a score of 2, it reports annual ischaemic-stroke rates across 15 validation cohorts that fall below 1% in four cohorts, between 1% and 2% in six, and above 2% in five. The single rate Curie Calc shows at that score (about 2.2 per 100 patient-years) is one cohort's point estimate, not a universal value.²³

Each case below is synthetic, and in each one all scoring criteria not mentioned are absent. The structure is the same throughout: case, contributing criteria, total, and the learning point.

These totals match the current Curie Calc scoring runtime. Check the point breakdown in Curie Calc.

The score has a defined scope, and reading it outside that scope is the most consequential error of all. It applies to non-valvular AF, and it should not be used to guide stroke prevention in moderate-to-severe rheumatic mitral stenosis or a mechanical heart valve. Both current guidelines also advise against basing these decisions on the score in hypertrophic cardiomyopathy or cardiac amyloidosis.³⁴

Several risk-relevant factors sit outside the score by design. It does not capture:

• obesity, or an enlarged left atrium;
• poorly controlled (as opposed to simply present) hypertension;
• reduced kidney function, such as an eGFR below 45, or proteinuria.

Because these factors change over time, and because AF risk factors accumulate, the score is worth reassessing as the clinical picture changes rather than treating a single result as permanent.³

CHA₂DS₂-VASc and HAS-BLED answer different questions. One describes stroke risk; the other describes bleeding risk. A bleeding-risk score is meant to flag and modify reversible bleeding risks, not to independently decide anticoagulation eligibility, and the two outputs should be read side by side rather than merged into a single verdict. You can work through bleeding-risk factors in the HAS-BLED calculator.³

Open Curie Calc's CHA₂DS₂-VASc calculator to enter the relevant criteria and review the result alongside its sources and limitations. Check the inputs and the scoring framework when you apply the reference to your own clinical context.

1. Enter the required criteria.
2. Review the contributing points.
3. Read the source and interpretation notes.