Maintaining attention underlies many aspects of cognition and becomes compromised early in neurodegenerative diseases like Alzheimer’s disease (AD). The consistency of maintaining attention can be measured with reaction time (RT) variability. Previous work has focused on measuring such fluctuations during in-clinic testing, but recent developments in remote, smartphone-based cognitive assessments can allow one to test if these fluctuations in attention are evident in naturalistic settings and if they are sensitive to traditional clinical and cognitive markers of AD.

Three hundred and seventy older adults (aged 75.8 +/− 5.8 years) completed a week of remote daily testing on the Ambulatory Research in Cognition (ARC) smartphone platform and also completed clinical, genetic, and conventional in-clinic cognitive assessments. RT variability was assessed in a brief (20-40 seconds) processing speed task using two different measures of variability, the Coefficient of Variation (CoV) and the Root Mean Squared Successive Difference (RMSSD) of RTs on correct trials.

Symptomatic participants showed greater variability compared to cognitively normal participants. When restricted to cognitively normal participants, APOE ε4 carriers exhibited greater variability than noncarriers. Both CoV and RMSSD showed significant, and similar, correlations with several in-clinic cognitive composites. Finally, both RT variability measures significantly mediated the relationship between APOE ε4 status and several in-clinic cognition composites.

Attentional fluctuations over 20–40 seconds assessed in daily life, are sensitive to clinical status and genetic risk for AD. RT variability appears to be an important predictor of cognitive deficits during the preclinical disease stage.

Smartphones have the potential for capturing subtle changes in cognition that characterize preclinical Alzheimer’s disease (AD) in older adults. The Ambulatory Research in Cognition (ARC) smartphone application is based on principles from ecological momentary assessment (EMA) and administers brief tests of associative memory, processing speed, and working memory up to 4 times per day over 7 consecutive days. ARC was designed to be administered unsupervised using participants’ personal devices in their everyday environments.

We evaluated the reliability and validity of ARC in a sample of 268 cognitively normal older adults (ages 65–97 years) and 22 individuals with very mild dementia (ages 61–88 years). Participants completed at least one 7-day cycle of ARC testing and conventional cognitive assessments; most also completed cerebrospinal fluid, amyloid and tau positron emission tomography, and structural magnetic resonance imaging studies.

First, ARC tasks were reliable as between-person reliability across the 7-day cycle and test-retest reliabilities at 6-month and 1-year follow-ups all exceeded 0.85. Second, ARC demonstrated construct validity as evidenced by correlations with conventional cognitive measures (r = 0.53 between composite scores). Third, ARC measures correlated with AD biomarker burden at baseline to a similar degree as conventional cognitive measures. Finally, the intensive 7-day cycle indicated that ARC was feasible (86.50% approached chose to enroll), well tolerated (80.42% adherence, 4.83% dropout), and was rated favorably by older adult participants.

Overall, the results suggest that ARC is reliable and valid and represents a feasible tool for assessing cognitive changes associated with the earliest stages of AD.