Smartphone-based cognitive assessments can provide unique information about cognition that is difficult or impossible with traditional cognitive assessments. Using high-frequency measurement “burst” designs, we have shown that older adults are capable and willing to participate in smartphone-based research, that this method dramatically improves between-subject reliability compared to traditional methods and demonstrates extraordinary test-retest reliabilities, and that high-frequency measurement can reveal time of day effects that are increased in those with elevated Alzheimer’s disease biomarkers. In this symposium session, we will provide an overview of our current work in older adults at risk for AD and highlight new analyses on the interaction between day to day variability in sleep and cognition. We will also cover our approach for measuring smartphone latencies, a critical aspect of bring-your-own-device (BYOD) studies.

The Ambulatory Research in Cognition (ARC) smartphone application for iOS and Android administers custom-designed tests of associate memory, processing speed, and spatial working memory. ARC uses a measurement burst design in which very brief (typically 60s or less) tests are completed at random times several times per day for up to one week. Measurement burst designs rely on principles from ecological momentary assessment, and can be described with a simple formula: 1. Test often and everywhere, 2. Keep assessments brief, and 3. Combine the data across sessions to increase reliability. At the Knight Alzheimer’s Disease Research Center at Washington University in St Louis, we have enrolled over 400 participants (ages 60-99 years) at risk for AD in ARC studies. These participants are comprehensively assessed with traditional cognitive tests, clinical examinations, neuroimaging, and fluid biomarkers. ARC also assesses sleep with the Pittsburgh Sleep Quality Index that captures essential sleep parameters, which are assessed daily during each 7-day measurement burst. Analyses of sleep and cognition focused on parameters including total sleep time, number of awakenings, sleep quality ratings, and an extremes analysis comparing cognition after nights with more sleep and after nights with less sleep.

Overall, participants reporting less total sleep time and more awakenings had lower memory and processing speed scores. This remained significant after modeling covariates including age, self-reported gender, education, and APOE ε4 status. Compared to nights with the most sleep, memory was worse after the nights with the poorest sleep.

When considering AD biomarkers in these analyses, participants with elevated AD biomarkers, including neurofilament light chain (NfL) and phosphorylated-tau181 (p-tau181), demonstrated more impacts of poor sleep on cogntion, such that the nights with the least sleep tended to impact cognition more than in those with normal biomarker levels, suggesting an important role for sleep in maintaining cognition in preclinical and early symptomatic AD. Interestingly, self-reported sleep quality was not associated with ARC cognitive tests, consistent with studies emphasizing the need for more quantitative assessments of sleep quality. In addition to these sleep data, we will review publications using the ARC platform, including a recently accepted manuscript in JINS (Nicosia et al., 2022).