Clinical background
The CHADS₂ score was developed for a concrete problem: to quantify in a single figure the risk of ischaemic stroke in non-valvular atrial fibrillation and thereby to guide the decision on thromboprophylaxis. Before CHADS₂ there were several competing schemes (Atrial Fibrillation Investigators, SPAF III) that all graded risk to some degree, but none had a structure that was simple to use in clinical practice while also discriminating adequately.
CHADS₂ combined the risk factors that had shown the strongest independent association with stroke in atrial fibrillation in earlier studies into an additive scoring system. An important design choice was that previous stroke or TIA scored double, since it is by far the strongest risk factor for recurrence. The score was intended to sort patients into low, intermediate and high risk, with anticoagulation reserved for the high-risk group.
The tool has largely been replaced by CHA₂DS₂-VASc, which builds on the same principles but adds female sex, vascular disease and a finer age stratification. The main reason was that CHADS₂ left too many patients in the intermediate category, in which the decision on anticoagulation remained unclear [3].
Calculating the CHADS₂ score
The score is calculated as a simple sum of five components:
where for heart failure (otherwise 0), for hypertension, for age years, for diabetes mellitus and for previous stroke or TIA. The maximum score is 6.
The derivation cohort consisted of 1,733 Medicare beneficiaries in the USA, aged 65 to 95 years, with non-rheumatic atrial fibrillation and not on warfarin at hospital discharge. The data came from peer review organisations in seven states, combined into the National Registry of Atrial Fibrillation (NRAF). Over 2,121 patient-years of follow-up there were 94 ischaemic strokes, corresponding to 4.4 per 100 patient-years [1]. The outcome was defined as hospital admission for ischaemic stroke as recorded by Medicare.
It is essential to note that the whole cohort was 65 years or older. The age threshold in the scoring system, 75 years, therefore fell within a population that was already years older than most patients with atrial fibrillation in current clinical practice.
Interpretation in practice
In the derivation cohort, the annual stroke risk rose by a factor of 1.5 (95% CI 1.3 to 1.7) per point [1]:
| CHADS₂ score | Strokes per 100 patient-years (95% CI) |
|---|---|
| 0 | 1.9 (1.2 to 3.0) |
| 1 | 2.8 (2.0 to 3.8) |
| 2 | 4.0 (3.1 to 5.1) |
| 3 | 5.9 (4.6 to 7.3) |
| 4 | 8.5 (6.3 to 11.1) |
| 5 | 12.5 (8.2 to 17.5) |
| 6 | 18.2 (10.5 to 27.4) |
A score of 0 was regarded as low risk and anticoagulation could be withheld, but even this group had an annual stroke risk of nearly 2%, which in a younger population can hardly be called negligible. A score of 1 served as an intermediate category in which the decision was unclear, and a score of 2 or higher routinely led to anticoagulation.
Clinical interpretation is complicated by the fact that the intermediate category, a score of 1, comprised a substantial proportion of patients in the original cohort. It was precisely this lack of clarity that motivated the development of CHA₂DS₂-VASc, which moves more patients into a clearly low or a clearly increased risk category [3].
Validation and performance
In the derivation cohort, CHADS₂ achieved a c-statistic of 0.82 (95% CI 0.80 to 0.84), better than both the AFI scheme (c = 0.68) and the SPAF III scheme (c = 0.74) [1]. This high discrimination has not, however, been reproduced in external cohorts.
In a validation study by Fang et al. of 13,559 adults with atrial fibrillation followed for a median of 6.0 years, the c-statistic for CHADS₂ was only 0.56 to 0.62, in the same range as the other risk schemes [3]. The proportion of patients classified as low risk varied between 11.7% and 37.1% depending on the scheme used, and the proportion at high risk varied between 16.4% and 80.4%.
A systematic review and meta-analysis by van der Endt et al., which included 6,267,728 patients with atrial fibrillation and 359,373 ischaemic strokes, calculated a pooled c-statistic for CHADS₂ of 0.658 (95% CI 0.644 to 0.672) [2]. For comparison, the corresponding value for CHA₂DS₂-VASc was 0.644 (95% CI 0.635 to 0.653). Calibration was reported in only 17 of the included studies and was judged reasonable but inadequately evaluated. The PROBAST assessment showed a high risk of methodological bias in all the included studies, above all in the outcome and analysis domains [2].
The discrepancy between the original c-statistic of 0.82 and external values of around 0.56 to 0.66 is probably explained by the narrow age distribution of the NRAF cohort (65 to 95 years only) and the high proportion of patients already at increased risk. In a broader population with a wider age range, the score loses some of its discriminatory power.
Limitations
CHADS₂ applies to non-valvular atrial fibrillation. With a mechanical valve prosthesis or significant mitral stenosis, anticoagulation is indicated whatever the score, and the tool must not be used to justify withholding it.
The derivation cohort consisted solely of Medicare beneficiaries aged 65 years and over, which makes the score unreliable in younger patients. It omits the age group 65 to 74 years as a separate risk factor, even though stroke incidence rises markedly from that age. Female sex is absent as a risk factor, and vascular disease (previous myocardial infarction, peripheral arterial disease, aortic plaque) is absent altogether, even though these have been shown to be independently associated with stroke in atrial fibrillation [3].
The principal criticism, supported both in the primary literature and in reviews, is that the intermediate category (a score of 1) is too broad [3, 4]. In a primary care study by Klein and Levine, CHADS₂ was shown not to be sensitive enough to separate patients clearly into high, intermediate and low risk groups, and the tool was therefore applied inconsistently in clinical practice [4]. In the meta-analysis, CHA₂DS₂-VASc, despite a slightly lower pooled c-statistic, showed a better ability to identify patients at genuinely low risk [2].
References
- Gage BF et al. Validation of clinical classification schemes for predicting stroke: results from the National Registry of Atrial Fibrillation. JAMA. 2001;285(22):2864–70. PMID: 11401607
- van der Endt VHW et al. Comprehensive comparison of stroke risk score performance: a systematic review and meta-analysis among 6 267 728 patients with atrial fibrillation. Europace. 2022;24(11):1739–1753. PMID: 35894866
- Fang MC et al. Comparison of risk stratification schemes to predict thromboembolism in people with nonvalvular atrial fibrillation. Journal of the American College of Cardiology. 2008;51(8):810–815. PMID: 18294564
- Klein D, Levine M. Are family physicians using the CHADS₂ score? Is it useful for assessing risk of stroke in patients with atrial fibrillation? Canadian Family Physician. 2011;57(8):e305–e309. PMID: 21841094
