Back

The Language & Memory Test: International feasibility study of an unsupervised digital test of cognition for neurologic populations

Leavitt, V. M.; Simani, L.; Koch, M. W.; Morrow, S. A.; Heuer, L. B.; Roozbeh, M.; Roozbeh, M.; Traynor, S. P.

2025-01-29 neurology
10.1101/2025.01.28.25321275 medRxiv
Show abstract

BackgroundCurrent cognitive outcomes are insufficient to monitor subtle change over time in neurologic populations. Challenging cognitive tests (i.e., cognitive "stress test") unaffected by practice effects and learning are ideal for monitoring and tracking change. Here we describe the Language & Memory Test (LMT), an unsupervised digital test we developed to measure rapid naming, visuospatial memory, and fine motor dexterity, that takes approximately 4 minutes to complete. ObjectiveConduct feasibility study of the LMT in adults recruited across three countries; in English and non-English speaking participants; employing in-person, remote, and unsupervised administration modalities. MethodsWe investigated retest reliability, usability / tolerability, and equivalence across centers, administration modalities, and English / non-English speakers. We also evaluated performance across adult age groups: 18-40, 41-60, and 61+ years. ResultsThe LMT was administered 555 times to 440 individuals ages 18-84 years in Canada, Iran, and the United States. Our sample was geographically and racially/ethnically diverse. Participants included a control sample of adults (n=115) and adults with a diagnosed neurologic disease (n=325). Of those invited to complete the LMT, 95.3% agreed; 99% of those who began the test completed it. Retest reliability was acceptable for finger tapping (Cronbachs =0.942) and language (=0.872), and marginally acceptable for memory 1 (=0.604). Equivalence varied across administration modalities, centers/countries, settings, and English / non-English speakers. Expected age-related performance decrements were shown for all three subtests. ConclusionsThe LMT shows promise as a culturally adaptable tool for use in diverse neurologic populations.

Matching journals

The top 6 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.