Skip to main content

Table 4 Discrimination, MAPE and calibration slopes of included models in patients with complete follow-up data (all models and risk scores) and in the full validation cohort (Cox proportional hazards regression models only)

From: An external validation of models to predict the onset of chronic kidney disease using population-based electronic health records from Salford, UK

  Study Patients with complete follow-up (n = 162,653) Full validation cohort (n = 178,399)
AUC (95 % CI) MAPE (SD)a Calibration slope (CI) c-index (95 % CI) MAPE (SD)a
Models Bang et al. [54] 0.899 (0.895–0.903) 0.063 (0.162) 0.97 (0.96–0.98) NA NA
Chien et al. [51]b 0.898 (0.895–0.901) 0.081 (0.162) 0.65 (0.64–0.65) 0.888 (0.885–0.892) 0.085 (0.166)
QKidney® [36]b 0.910 (0.907–0.913) 0.05 (0.166) 1.02 (1.01–1.04) 0.900 (0.897–0.903) 0.052 (0.165)
Kshirsagar et al. [53] 0.896 (0.892–0.900) 0.068 (0.164) 1.74 (1.72–1.76) NA NA
Kwon et al. [55] 0.899 (0.895–0.902) 0.086 (0.158) 0.68 (0.67–0.69) NA NA
O’Seaghdha et al. [52] 0.907 (0.904–0.911) 0.089 (0.169) 0.53 (0.52–0.53) NA NA
Thakkinstian et al. [56] 0.892 (0.888–0.985) 0.179 (0.161) 0.44 (0.43–0.45) NA NA
Simplified Scores Bang et al. [54] 0.895 (0.891–0.899) NA NA NA NA
Chien et al. [51] 0.880 (0.876–0.883) NA NA NA NA
Kshirsagar et al. [53] 0.891 (0.887–0.895) NA NA NA NA
Kwon et al. [55] 0.895 (0.891–0.898) NA NA NA NA
Thakkinstian et al. [56] 0.869 (0.864–0.873) NA NA NA NA
  1. AUC, area under receiver operating characteristic curve; eGFR, estimated glomerular filtration rate; NA, not applicable; SD, standard deviation; CI, confidence interval.
  2. aCalculated as mean difference between observed and predicted CKD cases
  3. bCox proportional hazard regression model
\