International Journal of Geriatric Psychiatry

Dementia tests differed in false alarms for UK ethnic minorities

The Montreal Cognitive Assessment misclassified 47.5% of healthy participants, versus 3.4% with the Rowland Universal Dementia Assessment Scale and 2% with the Multicultural Cognitive Examination.

Our read

Biomarkers
Studied:
Cross-sectional study of 200 cognitively healthy UK ethnic minority adults
What changed:
False alarms affected 47.5% of healthy ethnic minority participants with MoCA, against 3.4% with RUDAS and 2% with MCE
The catch:
English-proficiency requirements excluded less-fluent speakers and limited recruitment of older adults
In the record:
One of 3 papers on cognospeak on record: 2 with results in people.

Editor of record: Boris Djordjevic · AgingResearch.News

Follow @agingbrief

Two hundred cognitively healthy adults over 18 from Somali, South Asian and Chinese communities in Yorkshire took part in an observational study. Recruited through four community centers, they completed three pen-and-paper tests of thinking and memory and CognoSpeak, an automated speech assessment. All testing was in English.

The authors concluded that the Rowland and Multicultural tests may suit multilingual ethnic minority populations better than the Montreal test. Lower English proficiency was linked to lower Montreal test scores, but no clear link emerged for the other two tests. CognoSpeak’s model based on speech sounds achieved 83% accuracy in the ethnic minority group, compared with 86% in monolingual participants. Models using language or text showed more bias.

Why it matters

Dementia detection in multilingual communities raises the question of whether a low test score reflects cognitive difficulties or language differences. The authors said automated tools showed promise but needed further work on bias before clinical use.

Caveats

All 200 participants from the minority communities were cognitively healthy, so the comparison did not assess how well the tools detected dementia in those communities. Testing excluded adults unable to complete the assessments in English.

From the paper

Participants

205 participants were recruited; however, 5 participants failed to complete one or more of the cognitive assessments, meaning 200 were included (see Table ).

Methods, Participants

Limitation

Firstly, the average age of our cohort is lower than that of most patients seeking a memory diagnosis.

Limitations and Future Directions

Quoted word for word from the full text on PubMed Central.

Comments

Be the first to comment.