Table 3: Discriminant indices of Savonix test scores, to detect MCI from Normal. Summary of logistic regression for logistic regressions with the predictor tests adjusted education years and ages.
Max/Mix were determined by the AUC values of logistic regression models of 1000 randomly down-sampled 30 Normal subjects (out of 265) and 28 MCI subjects. AUC: Area Under the Receiver Operating Characteristic Curve. CI: Confidence Interval. PPV: Positive Predictive Values. NPV: Negative Predictive Values. B: regression coefficients. SD: Standard Deviations for the regression coefficients. Wald Chisq: Wald Chi square statistics. P value: P values for Wald Chi square statistics. Selected 3 subtests: Trail Making 2; Maze 2: and Complex Figure.
Savonix Test composites AUC (95% CI) Sensitivity Specificity PPV NPV B SD Wald Chisq P value
11 Subtests total score
Max 0.933 0.857 0.933 0.923 0.875 Test score -2.08 0.691 9.066 p <0.01
(0.872-0.994) Education 0.135 0.151 0.792 p = 0.374 n.s.
Age 0.123 0.063 3.858 p < 0.05
Min 0.758 0.821 0.633 0.676 0.792 Test score -0.894 0.430 4.312 p<0.05
(0.633-0.883) Education 0.041 0.098 0.177 p = 0.674 n.s.
Age 0.050 0.048 1.046 p = 0.306 n.s.
Selected 3 subtests total score
Max 0.917 0.929 0.8 0.813 0.923 Test score -4.175 1.437 8,453 p < 0.01
(0.847-0.987) Education 0.17 0.145 1.36 p = 0.244 n.s.
Age 0.128 0.062 4.219 p < 0.05
Min 0.745 0.679 0.733 0.704 0.710 Test score -1.659 0.809 4.197 p < 0.05
(0.618-0.873) Education -0.014 0.098 0.021 p = 0.884 n.s.
Age 0.073 0.048 2.312 p = 0.128 n.s.