Table 3

Validation statistics for each model in the THIN and QResearch validation cohorts

Women QResearch, mean (95% CI)Women THIN, mean (95% CI)Men QResearch, mean (95% CI)Men THIN mean, (95% CI)
Acute renal failure
R2 statistic62.86 (60.06 to 65.66)59.62 (57.15 to 62.10)58.5 (55.64 to 61.37)59.68 (57.44 to 61.91)
D statistic2.66 (2.50 to 2.82)2.49 (2.36 to 2.62)2.43 (2.29 to 2.57)2.49 (2.37 to 2.61)
ROC statistic0.852 (0.833 to 0.872)0.831 (0.816 to 0.846)0.834 (0.816 to 0.853)0.847 (0.835 to 0.859)
Cataract
R2 statistic58.70 (58.06 to 59.34)59.14 (58.49 to 59.78)60.55 (58.76 to 62.34)59.42 (57.71 to 61.12)
D statistic2.44 (2.41 to 2.47)2.46 (2.43 to 2.49)2.54 (2.44 to 2.63)2.48 (2.39 to 2.56)
ROC statistic0.880 (0.877 to 0.882)0.878 (0.876 to 0.881)0.881 (0.877 to 0.885)0.869 (0.870 to 0.877)
Myopathy
R2 statistic40.98 (40.95 to 41.01)42.15 (42.10 to 42.14)38.05 (38.02 to 38.08)35.98 (35.96 to 36.00)
D statistic1.71 (1.61 to 1.81)1.75 (1.68 to 1.82)1.6 (1.51 to 1.69)1.53 (1.47 to 1.59)
ROC statistic0.727 (0.691 to 0.764)0.739 (0.716 to 0.763)0.743 (0.714 to 0.773)0.717 (0.697 to 0.737)
Liver dysfunction
R2 statistic17.2 (15.38 to 19.02)15.55 (14.31 to 16.78)13.73 (11.75 to 15.72)10.83 (9.57 to 12.10)
D statistic0.93 (0.87 to 0.99)0.89 (0.84 to 0.92)0.82 (0.75 to 0.89)0.71 (0.67 to 0.76)
ROC statistic0.660 (0.651 to 0.670)0.646 (0.639 to 0.652)0.635 (0.626 to 0.644)0.612 (0.605 to 0.619)
  • Notes on understanding validation statistics:

    R2 statistic shows explained variation—higher values indicate that more variation is explained.

    ROC statistic is a measure of discrimination—higher values indicate better discrimination.

    D statistic is a measure of discrimination—higher values indicate better discrimination and an increase of ≥0.1 over other risk prediction models is a good marker of improved prognostic separation.

  • ROC, receiver operating characteristic.