Table 4 Calibration and discrimination statistics comparing the QRISK and Framingham equations for predicted 10-year risk of cardiovascular disease in the QRESEARCH validation cohort and the THIN cohort
StatisticsQRISK equationFramingham equation
In QRESEARCH validation cohortIn THIN validation cohortIn QRESEARCH validation cohortIn THIN validation cohort
Women
ROC statistic0.7879 (0.7842 to 0.7915)0.7888 (0.7858 to 0.7918)0.7759 (0.7723 to 0.7796)0.7595 (0.7564 to 0.7626)
D statistic1.535 (1.508 to 1.562)1.556 (1.532 to 1.580)1.397 (1.370 to 1.424)1.378 (1.354 to 1.402)
R2 statistic (%)36.01 (35.16 to 36.86)36.64 (35.94 to 37.34)31.79 (30.92 to 32.66)31.18 (30.45 to 31.91)
Brier score0.03700.03210.03820.0338
Predicted/observed (women)1.000.901.191.10
Men
ROC statistic0.7700 (0.7667 to 0.7733)0.7619 (0.7592 to 0.7646)0.7619 (0.7586 to 0.7653)0.7365 (0.7336 to 0.7394)
D statistic1.425 (1.401 to 1.449)1.394 (1.374 to 1.414)1.317 (1.293 to 1.341)1.255 (1.235 to 1.275)
R2 statistic (%)32.64 (31.88 to 33.39)31.70 (31.09 to 32.31)29.30 (28.54 to 30.05)27.31 (26.69 to 27.93)
Brier score0.05040.04700.05690.055
Predicted/observed (men)0.970.871.491.32
Predicted/observed (overall)0.990.881.361.23
  • Results are shown as mean (95% CI) unless stated otherwise.

  • Note that higher values for the ROC, D and R2 indicate better discrimination. Lower values of the Brier score indicate better calibration.

  • ROC, receiver operator curve.