Dataset for the discrepancy between letter and semantic fluency as a neurocognitive marker

<b>Objectives. </b>The discrepancy between letter fluency and semantic fluency has been proposed as a useful marker in neurocognitive diagnostics. No studies to date have examined discrepancy across different cognitive disorders in real world clinical settings. We aimed to assess the utility of a discrepancy score between letter fluency and semantic fluency in accurately classifying between several neurocognitive disorders.<b>Methods. </b>479 patients diagnosed with a neurocognitive disorder who had undergone neuropsychological assessment were recruited retrospectively from four clinical services at a tertiary hospital in Melbourne, Australia. Letter and semantic fluency scores were re-normed based on a control group, and a scaled subtraction discrepancy score was computed. Differences across diagnoses were assessed using general linear models (GLMs). Classification accuracy was assessed using receiver operating characteristic (ROC) curves.<b>Results. </b>Across the entire sample, the most common pattern was of normal performance on both letter and semantic fluency (<i>n</i> = 257, 53.7%). An impairment on both fluency measures (<i>n </i>= 96, 20.1%) was similar in frequency to an impairment in one paradigm only (<i>n</i> = 126, 26.2%). No differences between the diagnostic groups were observed based on the discrepancy score, and whilst there was a very modest discrimination ability in dementia of the Alzheimer type, it did not prove useful in accurately classifying other disorders from controls.<b>Conclusion. </b>This study did not support the use of a psychometric discrepancy score as a useful diagnostic metric, particularly in the absence of clinical interpretation. Further research is needed to determine in which clinical contexts, if any, such a psychometric marker is diagnostically useful when interpreted in isolation.