Skip to main content
. 2022 Jul 27;24(11):1739–1753. doi: 10.1093/europace/euac096

Table 3.

Results of the random-effects meta-analysis: pooled c-statistic ranging from largest to smallest validation cohort

Risk score characteristics Development Validation
Original sample size N of events Original c-statistic Model type Time frame N of studiesb
(N of validations)
Sample size N of events Pooled c-statistic
CHA2DS2-VASc 1084 25 0.61 Risk score 1 year 82 (n = 135) 3 229 267 169 199 0.644 (0.635–0.653)
CHADS2 1733 94 0.82 Risk score 1 year 46 (n = 68) 1 479 228 71 644 0.658 (0.644–0.672)
ATRIA 10 927 685 0.73 Risk score 1 year 11 (n = 24) 562 443 45 444 0.683 (0.658–0.708)
AFI 5955 208 Risk score 1 year 7 (n = 12) 153 530 6879 0.598 (0.558–0.636)
GARFIELD-AF 38 935 511 0.69 Formula 1 year 4 (n = 13) 149 848 5427 0.707 (0.676–0.737)
SPAF 2012 130 Risk score 1 year 5 (n = 5) 116 864 3778 0.650 (0.564–0.726)
Modified-CHADS2 51 807 5526 0.72 Risk score 5 years 5 (n = 10) 109 313 10 083 0.715 (0.674–0.754)
Framingham 705 83 0.66 Risk score 5 years 6 (n = 8) 95 145 2716 0.633 (0.602–0.662)
The ABC stroke risk score 27 929 a 391 0.68 Risk score 1 year 5 (n = 11) 40 340 1441 0.678 (0.658–0.697)
GARFIELD-AF II 52 032 957 0.68 Formula NA 1 (n = 2) NA NA 0.690 (0.672–0.707)

NA, data not available. Note that the number of validations slightly differs with Figure 3 since some validations needed to be excluded from the random effect meta-analysis due to studies omitting confidence intervals.

a

Person years.

b

Validation studies were included if discrimination measures (c-statistic) together with confidence intervals were available.