. 2020 Jun 4;3(6):e204669. doi: 10.1001/jamanetworkopen.2020.4669

Table 3. CDPoRT Predictive Performance for the Full, Simple, and Parsimonious Versions for Women and Men, by Development and Validation Settings^a.

CDPoRT version	Measures of predictive performance	Development (70%)	Bootstrap validation (development)^b	Split sample validation (30%)	Ontario final model	Bootstrap validation (Ontario final model)^b	Manitoba external validation	Sensitivity analysis
CDPoRT version	Measures of predictive performance	Development (70%)	Bootstrap validation (development)^b	Split sample validation (30%)	Ontario final model	Bootstrap validation (Ontario final model)^b	Manitoba external validation	No chronic disease history	Additional chronic diseases
Full version for women, df = 30	Nagelkerke R²	0.160	0.159	NA	0.160	0.159	NA	NA	NA
	Brier score	0.087	NA	0.119	0.087	NA	0.128	0.090	0.322
	Harrell C index	0.779	0.778	0.777	0.767	0.779	0.752	0.774	0.596
	Discrimination slope	0.109	NA	0.096	0.089	NA	0.103	0.105	0.035
	Calibration in the large	0	0.021	−0.006	0	0.013	0.032	−0.018	0.284
	Calibration slope	1	0.994	0.994	1	0.996	0.893	0.952	0.317
Simple version for women, df = 14	Nagelkerke R²	0.144	0.143	NA	0.143	0.143	NA	NA	NA
	Brier score	0.088	NA	0.122	0.089	NA	0.13	0.092	0.319
	Harrell C index	0.767	0.767	0.766	0.767	0.766	0.739	0.764	0.590
	Discrimination slope	0.092	NA	0.080	0.089	NA	0.081	0.087	0.033
	Calibration in the large	0	0.010	−0.005	0	0.007	0.033	−0.016	0.280
	Calibration slope	1	0.997	1.010	1	0.998	0.894	0.972	0.293
Parsimonious version for women, df = 24	Nagelkerke R²	0.160	0.158	NA	0.159	0.158	NA	NA	NA
	Brier score	0.087	NA	0.119	0.088	NA	0.128	0.090	0.322
	Harrell C index	0.779	0.778	0.776	0.779	0.778	0.752	0.774	0.596
	Discrimination slope	0.109	NA	0.095	0.104	NA	0.102	0.104	0.035
	Calibration in the large	0	0.017	−0.005	0	0.012	0.03	−0.018	0.284
	Calibration slope	1	0.995	0.994	1	0.997	0.885	0.951	0.316
Full version for men df = 30	Nagelkerke R²	0.178	0.176	NA	0.169	0.168	NA	NA	NA
	Brier score	0.091	NA	0.091	0.091	NA	0.129	0.092	0.297
	Harrell C index	0.783	0.782	0.769	0.780	0.779	0.775	0.775	0.618
	Discrimination slope	0.112	NA	0.103	0.106	NA	0.121	0.106	0.049
	Calibration in the large	0	0.021	−0.008	0	0.015	0.013	−0.024	0.244
	Calibration slope	1	0.994	0.865	1	0.996	1.007	0.946	0.331
Simple version for men df = 15	Nagelkerke R²	0.159	0.176	NA	0.155	0.154	NA	NA	NA
	Brier score	0.092	NA	0.091	0.092	NA	0.129	0.093	0.294
	Harrell C index	0.773	0.773	0.765	0.771	0.771	0.766	0.768	0.614
	Discrimination slope	0.093	NA	0.095	0.092	NA.	0.099	0.093	0.045
	Calibration in the large	0	0.010	−0.003	0	0.006	0.021	−0.023	0.242
	Calibration slope	1	0.997	0.924	1	0.998	1.004	0.971	0.319
Parsimonious version for men df = 25	Nagelkerke R²	0.177	0.176	NA	0.168	0.167	NA	NA	NA
	Brier score	0.091	NA	0.091	0.091	NA	0.129	0.092	0.297
	Harrell C index	0.783	0.782	0.769	0.780	0.779	0.775	0.775	0.618
	Discrimination slope	0.111	NA	0.103	0.105	NA	0.120	0.105	0.049
	Calibration in the large	0	0.018	−0.003	0	0.011	0.017	−0.024	0.244
	Calibration slope	1	0.995	0.866	1	0.997	1.006	0.948	0.332

^{^a}

As a reference, the predictive performance measures of the female model with all predictors were Nagelkerke R² = 0.154, Brier score = 0.088, Harrell C index = 0.775, and discrimination slope = 0.106. The predictive performance measures of the male model with all predictors were Nagelkerke R² = 0.169, Brier score = 0.093, Harrell C index = 0.779, and discrimination slope = 0.107.

^{^b}

Some statistics were not calculated during bootstrap validation (rms package).

Table 3. CDPoRT Predictive Performance for the Full, Simple, and Parsimonious Versions for Women and Men, by Development and Validation Settingsa.

Table 3. CDPoRT Predictive Performance for the Full, Simple, and Parsimonious Versions for Women and Men, by Development and Validation Settings^a.