Developing a predictive tool for psychological well-being among Chinese adolescents in the presence of missing data

Table 3 Characteristics of SIF Predictive Tool

	Training Samples		Validation Samples
Measure	Estimate (SE)	95% CI	Estimate (SE)	95% CI
AUC	0.758 (0.017)	0.722 to 0.794	0.755 (0.014)	0.726 to 0.783
Calibration Slopeª	0.993 (0.076)	0.838 to 1.147	0.976 (0.059)	0.858 to 1.094
Sensitivity^b	0.478 (0.030)	0.417 to 0.540	0.480 (0.025)	0.428 to 0.531
Specificity^b	0.862 (0.008)	0.848 to 0.877	0.860 (0.007)	0.845 to 0.874
Positive Predictive Value^b	0.441 (0.022)	0.397 to 0.485	0.431 (0.022)	0.388 to 0.473
Negative Predictive Value^b	0.879 (0.011)	0.856 to 0.903	0.882 (0.008)	0.865 to 0.898
Likelihood Ratio of positive test^b	3.478	2.879 to 4.200	3.414	2.918 to 3.993
Likelihood Ratio of negative test^b	0.604	0.534 to 0.684	0.605	0.547 to 0.670
Diagnostic Odds Ratio^{b, c}	5.755	4.241 to 7.810	5.638	4.390 to 7.241

ªCalibration slope is defined as the regression coefficient in the logistic regression of psychosocial maladjustment on the SIF Predictive Tool.
^bSensitivity, specificity, positive and negative predictive values, likelihood ratios of a positive and negative test, and the diagnostic odds ratio assume that a negative test is defined as when the SIF Predictive Tool score exceeds 1.
^cDiagnostic odds ratio equals the likelihood ratio of a positive test divided by the likelihood ratio of a negative test.

ISSN: 1471-2288