Supplementary Material for: Why Almost Always Animals? Ranking Fluency Tasks for the Detection of Dementia Based on Receiver Operating Characteristic (ROC) and Quality ROC Analyses

<p><b><i>Background/Aims:</i></b> Category fluency tasks have been widely used to assess cognitive functioning in both clinical and experimental environments as an index of cognitive and psycholinguistic dysfunctions in dementia. Typically, a reduced group of semantic categories has been selected for neuropsychological assessment (e.g., <i>animals</i>, <i>fruits</i> or <i>vegetables</i>), although empirical support for the prevalence of one category among others is absent in the literature. <b><i>Methods:</i></b> We provide an empirical evaluation of the ability of 14 category fluency tasks to discriminate between subjects with dementia of the Alzheimer type and healthy elderly participants. As a novelty, we used both receiver operating characteristic (ROC) curves and quality ROC calibrated analyses to characterize the interplay of sensitivity and specificity of every category fluency task performance as a screening tool. The use of calibrated measures provided us with a useful tool for comparing the diagnostic ability of the different categories, as well as making rankings of categories based on the quality indices of efficiency, sensitivity, and specificity. <b><i>Results:</i></b> The habitually used category of <i>animals</i> is far from being the most efficient one in terms of its diagnostic power to evaluate dementia. <b><i>Conclusion:</i></b> Our study might guide the selection of suitable category fluency tasks according to the diagnostic purposes in dementia.</p>