*3.3. Diagnostic Accuracy of Dementia and AD*

Table 4 shows the accuracy (diagnostic accuracy) and the safety and validity indicators of the multivariate logistic regression models for dementia and AD. For both variables, Model 3 adjusted for age, depression, dyslipidemia, and T2DM showed the highest goodness of fit (Nagelkerke r<sup>2</sup> = 0.48), i.e., the variables included accounted for 48% of the value of the result variable.


**Table 4.** Diagnostic accuracy of logistic regression models adjusted for dementia and AD.

Model 3 achieved for dementia an area under the curve of 86% (ROC curves in Figure 1), a sensitivity of 77.1%, and a specificity of 81.6%, which resulted in a Youden index of 0.59, with a positive predictive value (PPV) of 76.3% and a negative predictive value (NPV) of 82.3%. The percentage of correctly classified persons (validity index) was 79.6%.

Model 3 obtained an area under the curve of 85.8% and a diagnostic validity index of 76.9% for AD. Sensitivity was 66.2%, and specificity was 83.2%, with a Youden index of 0.48. PPV and NPV were 70% and 80.6%, respectively.
