Validation of the test for finding word retrieval deficits (WoFi) in detecting Alzheimer's disease in a naturalistic clinical setting

Abstract

Background Detecting impaired naming capacity contributes to the detection of mild (MildND) and major (MajorND) neurocognitive disorder due to Alzheimer’s disease (AD). The Test for Finding Word retrieval deficits (WoFi) is a new, 50-item, auditory stimuli-based instrument. Objective The study aimed to adapt WoFi to the Greek language, to develop a short version of WoFi (WoFi-brief), to compare the item frequency and the utility of both instruments with the naming subtest of the widely used Addenbrooke’s cognitive examination III (ACEIIINaming) in detecting MildND and MajorND due to AD. Methods This cross-sectional, validation study included 99 individuals without neurocognitive disorder, as well as 114 and 49 patients with MildND and MajorND due to AD, respectively. The analyses included categorical principal components analysis using Cramer’s V, assessment of the frequency of test items based on corpora of television subtitles, comparison analyses, Kernel Fisher discriminant analysis models, proportional odds logistic regression (POLR) models and stratified repeated random subsampling used to recursive partitioning to training and validation set (70/30 ratio). Results WoFi and WoFi-brief, which consists of 16 items, have comparable item frequency and utility and outperform ACEIIINaming. According to the results of the discriminant analysis, the misclassification error was 30.9%, 33.6% and 42.4% for WoFi, WoFi-brief and ACEIIINaming, respectively. In the validation regression model including WoFi the mean misclassification error was 33%, while in those including WoFi-brief and ACEIIINaming it was 31% and 34%, respectively. Conclusions WoFi and WoFi-brief are more effective in detecting MildND and MajorND due to AD than ACEIIINaming.

Comparative Analysis between Quality of Life and Human Labor in Countries Belonging to G7 and BRICS Blocks: Proposition of Discriminant Analysis Model

“…Some individuals in Group 1 are classified in Group 3, probably influenced by the variables of X7, X9, X10, X11, X12, X13 (Medium Quality Value), and 3 (High Quality Value), which are better classified, as this is not a fixed category, but rather a transitional one, in which countries move between one category and another or advance from one point to another. In a further analysis, we employed the KFD analysis model, which is a technique to find linear functions capable of two or more classes or groups (Georgiou et al 2023). Table 11 introduces the KFD classification function coefficients used to classify the observations in each group.…”

“…The KFD functions helped to classify a given case according to the highest value recorded for one of the three discriminant functions. In a further analysis, we employed the KFD analysis model, which is a technique to find linear functions capable of two or more classes or groups (Georgiou et al 2023). Table 11 introduces the KFD classification function coefficients used to classify the observations in each group.…”

“…This approach maps input data into a higher-dimensional "feature space", where linear classification can be effectively performed, which corresponds to a powerful non-linear decision function in the original input space (Mika et al 1999, p. 43). Similar to Georgiou et al (2023), Kernel Fisher discriminant analysis models were employed to find two or more correctly separated classes or groups. This included different steps: (1) mapping to a higher-dimensional space, (2) computing the kernel-matrix, (3) between-class and within-class scatter matrices, (4) maximizing Fisher's criterion, (5) extracting discriminant functions, (6) classification function coefficients, and (7) classifying the new observations, which were supported through the SPSS data analysis.…”