Skip to main content
. 2016 Apr 14;7:48. doi: 10.3389/fpsyt.2016.00048

Table 6.

Fivefold cross-validation results.

Fold No. Obs Population prevalence parameters % Sensitivity % Specificity ROC area Odds ratio % PPV* % NPV*
1 28 0.35 78.6 92.9 0.857 47.7 3.7 99.9
30 78.6 92.9 0.857 47.7 82.5 91.0
2 26 0.35 84.6 92.3 0.855 66 3.7 99.9
30 84.6 92.3 0.885 66 82.5 93.3
3 28 0.35 85.7 64.3 0.750 10.8 0.8 99.9
30 85.7 64.3 0.750 10.8 50.7 91.3
4 26 0.35 92.3 92.3 0.923 144 4.0 100.0
30 92.3 92.3 0.923 144 83.7 96.6
5 26 0.35 84.6 92.3 0.885 66 3.7 99.9
30 84.6 92.3 0.885 66 82.5 93.3
Scored 15 biomarker set 134 0.35 85.1 86.6 0.858 36.7 2.2 99.9
30 85.1 86.6 0.858 36.7 73.1 93.1

Results are reported across prevalence values that may be clinically encountered, 0.35% prevalence = mean prevalence for schizophrenia (applicable to screening) within the Australian population, where 12 monthly MEAN schizophrenia prevalence is estimated as equivalent to 0.35/100 (0.35%), 30% prevalence = estimated prevalence rate among a clinical population (where expected conversion rate to schizophrenia is 30–35% at 1-year follow-up) (89).

*Percent positive predictive value and negative predictive value.