Table 3.
Discriminative ability and calibration of the different 6-year hypertension incident risk models for both genders in training and testing set, respectively
| Models | Cut-off | AUC (95% CI) | Calibration χ2 | P-value |
|---|---|---|---|---|
|
Training set: | ||||
|
Men | ||||
| M1 |
0.1926 |
0.765 (0.745, 0.784) |
4.91334 |
0.84180 |
| ANN |
0.2305 |
0.767 (0.747, 0.786) |
24.54347 |
0.00352 |
| NBC |
0.2205 |
0.751 (0.730, 0.770) |
105.88180 |
<0.00001 |
| CART |
0.0994 |
0.720 (0.699, 0.741) |
4.56824 |
0.10186 |
|
Women | ||||
| W1 |
0.1920 |
0.806 (0.791, 0.820) |
4.72712 |
0.31645 |
| W2 |
0.1922 |
0.806 (0.791, 0.820) |
1.18206 |
0.88104 |
| ANN |
0.2512 |
0.809 (0.795, 0.823) |
5.44370 |
0.24472 |
| NBC |
0.2588 |
0.796 (0.780, 0.810) |
193.18980 |
<0.00001 |
| CART |
0.0909 |
0.740 (0.724, 0.756) |
17.95192 |
0.00012 |
|
Testing set: | ||||
|
Men | ||||
| M1 |
0.1745 |
0.771 (0.750, 0.791) |
6.30570 |
0.70898 |
| ANN |
0.2799 |
0.773 (0.752, 0.793) |
29.27430 |
0.00058 |
| NBC |
0.2205 |
0.760 (0.738, 0.781) |
82.26996 |
<0.00001 |
| CART |
0.0994 |
0.722 (0.699, 0.743) |
5.249259 |
0.07247 |
|
Women | ||||
| W1 |
0.1798 |
0.765 (0.746, 0.783) |
6.78323 |
0.14780 |
| W2 |
0.1446 |
0.764 (0.746, 0.783) |
7.40462 |
0.11599 |
| ANN |
0.2022 |
0.756 (0.737, 0.775) |
4.74466 |
0.31451 |
| NBC |
0.1860 |
0.761 (0.742, 0.779) |
189.75400 |
<0.00001 |
| CART | 0.0909 | 0.698 (0.677, 0.717) | 19.73303 | 0.00005 |
AUC – area under the receiver operating characteristic curve, CI – confidence interval, M1 – men office-based model, ANN – Artificial Neural Network, NBC – Naive Bayes Classifier, CART – Classification and Regression Tree, W1 – women office-based model, W2 – women laboratory-based model