Diagnostic performance of verbal fluency measures: a cross-sectional study in the stages of cognitive continuum

TL;DR

This study shows that semantic verbal fluency and its discrepancy with phonemic fluency can help detect early cognitive decline, but accuracy is moderate.

Contribution

The study introduces the use of VF discrepancy scores for detecting subjective cognitive decline, which was previously underexplored.

Findings

SCD individuals showed significantly lower semantic fluency scores compared to healthy controls.

Dementia was most accurately identified using semantic fluency with 81% balanced accuracy.

The discrepancy score detected SCD with 61% balanced accuracy, outperforming semantic fluency alone.

Abstract

Verbal fluency (VF) measures are sensitive markers of advanced cognitive decline; however, the utility of the discrepancy score remains underexplored in the early stages of cognitive decline, such as subjective cognitive decline (SCD). This study evaluated semantic fluency (SF) and phonemic fluency (PF), as well as discrepancy score sensitivity, in clinical populations with SCD, mild cognitive impairment (MCI), and dementia.

In this cross-sectional study, 193 older adults (72 healthy controls, 67 with SCD, 16 with MCI, and 38 with dementia) were consecutively recruited from the Nyírő Gyula National Institute of Psychiatry and Addictology, Hungary. Each participant underwent a comprehensive neurological interview and neuropsychological assessments. Semantic and phonemic fluency, along with their discrepancy, served as the primary outcome measures, defined as the number of correct words generated in one minute for each fluency type and the difference between them. Group differences were assessed using one-way ANCOVAs that controlled for age and education, and diagnostic classification performance was evaluated using a multinomial logistic regression model-based metric.

Relative to healthy controls, SCD showed significantly lower SF scores (β = − 2.267, p = .009) but no difference in PF (p = .493). Both fluency types were reduced in MCI and dementia, with semantic declines being especially pronounced. Multinomial logistic regression model identified dementia most accurately using SF (balanced accuracy = 0.81) and SCD using the discrepancy score (balanced accuracy = 0.61), while MCI classification was poor, likely due to a small sample size. These results underscore the potential of SF and its discrepancy with PF for early detection and differentiation of cognitive decline.

This study demonstrates that individuals with SCD have deficits in semantic—but not phonemic—fluency, and that VF discrepancy scores detect SCD more effectively than semantic fluency alone. However, accuracy is only moderate and insufficient for use as a standalone clinical test.

Figures