Characterising spoken interactions of healthy ageing adults with CognoSpeak, a web‐based cognitive assessment tool

Abstract

AbstractBackgroundAutomatic conversation analysis presents a useful aid for diagnosis and management of people with cognitive impairments, having already successfully differentiated between people with Alzheimer’s disease (AD) and healthy controls. Current clinical brief cognitive tests such as Addenbrooke’s Cognitive Examination are influenced by age, education and practice effects but it is unknown how these factors influence automated analysis of speech. This study characterised the effects of age, education and repetition on measures of speech and language of cognitively healthy ageing adults.MethodAudio recordings of memory related questions and verbal fluency tasks from 37 cognitively healthy adults’ conversations with CognoSpeak were analysed (mean age 73.11 ± 7.97). Linguistic analysis comprised of pause to speech ratio (pause length/speech length), age of acquisition of words, phonemes per minute, words per minute and total word count. Spearman’s correlations assessed relationships with age and education. 6 participants repeated testing at an average of 4.05 and 12.25 months later. Friedman’s ANOVA and post‐hoc related‐samples Wilcoxon signed ranks tests assessed longitudinal changes.ResultPause to speech ratio was significantly higher for recent memory questions than remote memory. Age of acquisition was significantly higher for phonemic fluency than semantic fluency. A moderate negative correlation was found for age with ACE‐R score and phonemic fluency. A strong positive correlation was found with education and ACE‐R score, and a moderate positive correlation with education and words per minute. No further relationships for age or education were found. Semantic fluency showed practice effects, whereas linguistic measures did not. Further follow‐up data will be available by July 2021.ConclusionThese results support the theory of semantic memory preservation with age. It is indicated that age, education and time do not have large effects on linguistic measures, giving conversation analysis a practical advantage over the ACE‐R for longitudinal monitoring of people with Mild Cognitive Impairment or Subjective Cognitive Impairment.