Validation of connected speech as a marker for cognitive change in late‐middle aged African American adults at risk for Alzheimer's disease

Abstract

AbstractBackgroundConnected speech‐language (CSL) (linguistic, acoustic, and temporal measures from spoken language) is a promising performance‐based measure of cognition for studies researching Alzheimer’s disease (AD). CSL represents everyday communication and thus may be less influenced by sociocultural factors, but research investigating CSL in underrepresented groups is scarce. Picture description tasks are common for eliciting CSL, typically the Cookie Theft (CT). However, CT is outdated, it does not represent the US’s cultural diversity, and subsequently may not be effective for all groups. Our aims were 1) characterize CSL from late‐middle‐aged African Americans (AA) by sex and APOE; 2) compare AA CSL change vs. standardized language tests across cognitive statuses 3) determine if AA and non‐Hispanic whites (NHW) respond differently to the CT prompt.MethodCSL measures (Figure 1) were automatically extracted from picture description transcripts from WRAP’s Milwaukee site (AA n=48; NHW n=30). Aims 1 and 3 statistics included t‐tests, chi‐square tests, and ANCOVA. Aim 2 used linear mixed‐effects models in AA with >1‐timepoint (n=30), examining whether longitudinal CSL varied by cognitive status.ResultTable 1 shows speech metrics and demographics of AA participants by sex. AA and NHW did not differ on most baseline CSL measures including total words spoken (Table 2,3, Figure 2). Changes in CSL (Table 4) and BNT were associated with cognitive status in AA participants, but verbal fluency was not. Literacy was not related to CSL outcomes. CSL had higher effect sizes than BNT (b=.41 vs. .20).ConclusionIn one of the first studies to characterize prospective CSL from an AD‐risk‐enriched sample of AA adults, we found no sex/APOE carrier status differences in AA CSL, and no differences between AA and NHW on most CSL measures. AA and NHW did not differ in the amount of talking in response to CT. CSL change among cognitively‐declining AAs was similar to that seen previously in a NHW sample. This combined with predominantly weak associations between CSL measures and WRAT‐3 suggest that CSL may reflect a socioculturally unbiased cognitive activity. Future directions include developing an ethnoculturally relevant stimulus picture, analyses with increased time points, sample sizes and AD biomarkers.