Prediction model for high glycated hemoglobin concentration among ethnic Chinese in Taiwan

Table 4 Summary statistics for different prediction models based on covariates in the Cambridge, clinical, and biochemical model algorithms on the validation data

	Area under ROC curve	Brier score*	2*Forecast Outcome Covariance	Hosmer Lemeshow chi-square**	Hosmer Lemeshow P value**
Cambridge	0.691	0.0219	0.0004	14.6	0.07
Clinical, coefficient-based	0.712	0.0217	0.0007	12.8	0.12
Clinical, points-based	0.723	0.0220	0.0003	18.7	0.03
Biochemical, coefficient-based	0.773	0.0213	0.0013	8.8	0.36
Biochemical, points-based	0.770	0.0219	0.0003	3.8	0.87

*A low Brier score indicated a goodness-of-fit. ** Low Hosmer-Lemeshow chi-square and high P values indicated a goodness-of-fit
A higher area under the ROC area as well as 2*Forecast-outcome-covariance represented better performance. A lower Hosmer-Lemeshow chi-square value represented a goodness-of-fit model.

ISSN: 1475-2840