m-TVIT | m-AHTST | m-TVIT and m-AHTST combined | |
ROC p value against ROC=0.5* | p<0.0001 | p<0.0001 | p<0.0001 |
AUC† | 0.880 | 0.835 | 0.921 |
P value using m-AHTST as reference | p=0.1287 | Reference | p=0.0002 |
P value using m-TVIT as reference | Reference | p=0.1287 | p<0.0001 |
Best cut-off‡ | 6 | 1 | 6 |
Youden’s J§ | 0.5816 | 0.5860 | 0.6782 |
*ROC analysis for each and combined screeners are shown in figure 2. ROC=0.5 signifies no discrimination.
†The area under the curve for each ROC varies from 0.5 (random) to 1.0 (perfect classifier). 0.9–1.0 is considered outstanding; 0.8–0.9 is considered excellent.
‡The best cut-off is the number of screener questions involved in achieving the maximum sensitivity (true positive rate) and specificity (true negative rate) against the partial PRIF criteria within each screener or combined.
§Youden’s J index is the height from the random AUC line (ROC=0.5) to the coordinates of the best cut-off or, sensitivity+specificity=1. Varies from 0 to 1.0.
AUC, area under the curve; PRIF, Prevalence Reduction Innovation Forum; ROC, receiver operating characteristics.