To evaluate the feasibility, usability, and preliminary validity of a digital phenotyping protocol to capture everyday cognition and activities in vivo among older adults.

Eight participants (M age=69.1 + 2.6; M education=18.0 + 1.4; 50% female; 88% non-Hispanic White) with normal cognition or mild cognitive impairment used an open-source smartphone application (mindLAMP) to passively and continuously capture sensor data including global positioning system (GPS) trajectories for a 4-week study period. Baseline neuropsychological tests and measures of depression, self-reported cognitive decline and mobility patterns were collected as external validators for digital data. Participants downloaded mindLAMP onto their smartphones and resumed their daily routines for 4 weeks before removing mindLAMP and completing a debriefing questionnaire. A cognitive composite was derived by averaging T-scores across domains of attention, executive functioning, processing speed, memory, and language. GPS raw data were processed to generate monthly average and standard deviation mobility metrics for each participant, including time spent at home, distance travelled, radius of gyration, flight length, and circadian routine. Feasibility and usability findings are presented along with correlation coefficients >.4 between GPS metrics and external validators.

100% of enrolled participants completed the 4-week study without requesting to withdraw. Usability ratings ranged from poor to excellent. 75% of participants agreed that mindLAMP was easy to use, whereas only 1 participant enjoyed using mindLAMP. 100% of participants were satisfied with the study team’s explanation of procedures, privacy safeguards, data encryption methods and risks/benefits, reflected in an average score of 98.8% on the comprehension of consent quiz. No participants reported feeling uncomfortable, suspicious, or paranoid due to the study application running on their smartphone. No participants endorsed new problems using their smartphone, though 75% reported charging it more frequently during the study period. On average each day, participants spent 1121 + 227 minutes at home, travelled 38727 + 36210 geodesic units, and had 201 + 149 minutes of missing GPS data. Overall, greater amounts of activity (monthly average) and higher variability (monthly standard deviation) in GPS metrics were associated with better outcomes. Specifically, less time spent at home, greater distance travelled, larger radius of gyration, greater flight length, and greater variability in home time, distance travelled, radius of gyration and flight length were associated with less depression, less self-reported cognitive decline, better cognition, and greater self-reported mobility (.40< |r| <.69). On the other hand, greater circadian routine was associated with more self-reported cognitive decline (r=.66) and less self-reported mobility (r=-.43).

Smartphone digital phenotyping is a feasible and acceptable method to capture everyday activities in older adults. Continuous collection of data from personal devices warrants caution; however, participants denied privacy concerns and expressed an overall positive experience. High frequency GPS data collection impacts battery life and should be considered among relative risks and confounds to naturalistic assessment. Patterns of behavior from passive smartphone data show promise as an unobtrusive method to identify cognitive risk and resilience in older adults. Subsequent analyses will evaluate additional sensor metrics across a larger and more heterogeneous cohort.