Skip to main content

Table 4 A comparison of model performance across models 2–5

From: Using the kidney failure risk equation to predict end-stage kidney disease in CKD patients of South Asian ethnicity: an external validation study

Performance measure

Model 2a

Model 2b

Model 3a

Model 3b

Model 4

Model 5

White (95% CI)

South Asian (95% CI)

White (95% CI)

South Asian (95% CI)

White (95% CI)

South Asian (95% CI)

White (95% CI)

South Asian (95% CI)

Discrimination

C-index (up to 5 years)

0.908 (0.886, 0.928)

0.954 (0.933, 0.968)

0.908 (0.886, 0.928)

0.954 (0.933, 0.968)

0.908 (0.886, 0.928)

0.954 (0.933, 0.968)

0.919 (0.898, 0.939)

0.957 (0.936, 0.974)

Calibration

O/E ratio

0.938 (0.821, 1.06)

1.00 (0.815, 1.19)

0.946 (0.828, 1.06)

0.967 (0.778, 1.16)

0.938 (0.821, 1.06)

1.01 (0.820, 1.20)

1.12 (1.00, 1.24)

0.972 (0.782, 1.16)

Calibration intercept

0.094 (−0.067, 0.255)

0.146 (−0.120, 0.413)

0.062 (−0.101, 0.226)

0.151 (−0.103, 0.404)

0.096 (−0.065, 0.257)

0.153 (−0.113, 0.419)

0.227 (0.046, 0.409)

0.156 (−0.107, 0.419)

Calibration slope

1.05 (0.920, 1.18)

0.843 (0.690, 0.995)

1.02 (0.894, 1.15)

0.917 (0.751, 1.08)

1.05 (0.921, 1.18)

0.843 (0.691, 0.996)

0.905 (0.798, 1.01)

0.950 (0.752, 1.15)

Overall fit

Brier score

0.011 (0.009, 0.012)

0.029 (0.023, 0.035)

0.011 (0.009, 0.012)

0.029 (0.023, 0.034)

0.011 (0.009, 0.012)

0.029 (0.023, 0.035)

0.010 (0.009, 0.012)

0.025 (0.019, 0.030)

Scaled Brier score (%)

25.4 (20.2, 29.3)

34.6 (20.7, 43.7)

25.6 (21.3, 29.4)

35.0 (23.6, 45.3)

25.4 (19.7, 29.8)

34.6 (21.8, 44.1)

27.4 (22.7, 31.6)

43.8 (31.9, 52.8)

  1. Models 2a and 2b adjusted the baseline risk by ethnic group, models 3a and 3b rescaled the linear predictor and updated baseline risk for each group, model 4 included ethnicity as a predictor, and model 5 included ethnicity as a predictor and re-estimated the model coefficients. 95% confidence intervals were found via bootstrapping with 1000 simulations for the c-index and the scaled Brier score. CI confidence interval, O/E ratio observed/expected ratio