Aging & Longevity

Detection of Age-Related Cognitive Decline Over Short Time Frames Using Digital Multiday Learning Assessments: Longitudinal Study.

Blotenberg I, Kim H, et al. • JMIR aging • 2026

TL;DR

A digital multiday learning paradigm combined with computational modeling of response times sensitively detects age-related cognitive decline over short timescales (<3 years) in cognitively unimpaired older adults, and identifies process-based markers associated with aging, including slower performance, reduced stability, and diminished learning across testing intervals.

Key Findings

Results

Peak performance (a practice-adjusted marker of best expected response time) slowed over time in cognitively unimpaired older adults, indicating age-related cognitive decline over under 3 years.

Peak performance slowed by 0.04 seconds per year (95% credible interval 0.02-0.06), approximately 2% per year.
The sample was 209 cognitively unimpaired older adults (mean age 73.6, SD 8.2 years; 65.1% female) from the Harvard Aging Brain Study and affiliated studies.
Participants were followed across up to 4 testing intervals over a mean of 2.8 (SD 0.4) years.
The outcome was daily median response time for correct responses on the face-name task of the Boston Remote Assessment for Neurocognitive Health (BRANCH).

Results

Older age was associated with slower peak performance at baseline.

A 1-SD higher age was associated with +0.19 seconds slower peak performance (95% credible interval 0.12-0.26).
Estimates came from a computational model with embedded linear regressions for age, sex, years of education, and device type.

Results

Older age was associated with steeper annual slowing in peak performance.

Higher age (+1 SD) was associated with an additional +0.036 seconds per year of slowing in peak performance.
This indicates faster decline over the roughly 2.8-year follow-up in older participants.

Results

Older age was associated with greater within-person variability (reduced stability) in response times.

Higher age (+1 SD) was associated with +0.043 seconds greater within-person variability.
Stability was one of several process-based markers derived from computational modeling.

Results

Older age was associated with reduced learning across testing intervals and with more between-interval forgetting.

Learning within and across testing intervals and between-interval forgetting were modeled separately from peak performance.
Higher age was associated with reduced learning across testing intervals and more forgetting.

Conclusions

The digital multiday learning paradigm with computational modeling captures age-related decline over short timescales that are difficult to detect with conventional designs relying on infrequent assessments.

Decline was detected over a mean of 2.8 years (under 3 years).
The approach separates multiple process-based cognitive markers: peak performance, stability, learning, and forgetting.

What This Means

This study asked whether short, repeated, smartphone- or computer-based memory tests could pick up the small cognitive changes that come with aging over just a few years. In 209 older adults without cognitive impairment (average age about 74), researchers used a face-name learning task, in which people learn to pair faces with names, given over several days in up to four rounds across roughly 2.8 years. They used computer modeling of response times to separate a person's best possible speed from the effects of practice. The best-performance speed slowed by about 0.04 seconds per year, or roughly 2% a year, showing measurable decline over under three years. Older participants were slower to begin with, slowed down faster, were more variable from day to day, learned less across testing rounds, and forgot more between rounds. This research suggests that frequent digital testing combined with modeling may detect subtle age-related changes over short periods, when traditional studies using infrequent testing might miss them. It also suggests that different aspects of thinking, such as speed, consistency, learning and forgetting, can be tracked separately. This may be useful for future research on aging and early cognitive change, though this study does not itself test diagnosis or treatment.

Have a question about this study?

Citation

Blotenberg I, Kim H, Soberanes D, Dubbelman M, Minto L, Udeogu O, et al.. (2026). Detection of Age-Related Cognitive Decline Over Short Time Frames Using Digital Multiday Learning Assessments: Longitudinal Study.. JMIR aging. https://doi.org/10.2196/100917

Key Findings

What This Means

Have a question about this study?

Related Research

Citation