A Receiver Operating Characteristic (ROC) curve analysis offers a means of comparing the sensitivity/specificity of two tests applied to the same population. The areas under the ROC curves resulting from comparing the specificity and sensitivity of the CANA with the MMSE using the age and education adjusted total scores were 0.963 (standard error = 0.0152, p<0.001) and 0.758 (standard error = 0.0412, p<0.001) for the CANA and the MMSE respectively. Therefore, each test was significantly better than chance with respect to specificity and sensitivity. In addition, the ROC curve analyses provide the optimum cut-off score for the best combination of specificity and sensitivity for a given test. Therefore, for the CANA Test, the cut-off score to optimize sensitivity/specificity is -1.22 (i.e., a T-score of 38), which yields a sensitivity of 94.2% and a specificity of 91.4%. Both indices are in the excellent range (Tape, 2001).

Reliability Coefficients

For the CANA Test, test-retest was not a viable option; therefore, the split-half analysis was performed. There were 49 individual test items that were divided into two half-tests using the odd/even method. This provided two nearly parallel forms with approximately equal variances. Then Spearman Brown Split-halves Reliability coefficient formulas were used in the analysis. The reliability coefficient for the total normative sample (N = 148) was r = .77 which is an acceptable level of reliability (Pedhazur & Schmelkin, 1991).

To read more extensive details, click here