Remote and unsupervised digital memory assessments can reliably detect cognitive impairment in Alzheimer's disease

Abstract

Introduction: Remote unsupervised cognitive assessments have the potential to complement and facilitate cognitive assessment in clinical and research settings.

Methods: Here, we evaluate the usability, validity, and reliability of unsupervised remote memory assessments via mobile devices in individuals without dementia from the Swedish BioFINDER-2 study and explore their prognostic utility regarding future cognitive decline.

Results: Usability was rated positively; remote memory assessments showed good construct validity with traditional neuropsychological assessments and were significantly associated with tau-positron emission tomography and downstream magnetic resonance imaging measures. Memory performance at baseline was associated with future cognitive decline and prediction of future cognitive decline was further improved by combining remote digital memory assessments with plasma p-tau217. Finally, retest reliability was moderate for a single assessment and good for an aggregate of two sessions.

Discussion: Our results demonstrate that unsupervised digital memory assessments might be used for diagnosis and prognosis in Alzheimer's disease, potentially in combination with plasma biomarkers.

Highlights: Remote and unsupervised digital memory assessments are feasible in older adults and individuals in early stages of Alzheimer's disease. Digital memory assessments are associated with neuropsychological in-clinic assessments, tau-positron emission tomography and magnetic resonance imaging measures. Combination of digital memory assessments with plasma p-tau217 holds promise for prognosis of future cognitive decline. Future validation in further independent, larger, and more diverse cohorts is needed to inform clinical implementation.

Keywords: Alzheimer's disease; ambulatory assessments; blood‐based biomarkers; digital cognitive markers; ecological momentary assessments; mHealth; memory; plasma marker; smartphone‐based unsupervised assessments.

FIGURE 1

Timeline of the study protocol. Participants enlisted for a 12‐month study of biweekly remote and unsupervised memory assessments of the MDT‐OS and ORR. In the initial in‐clinic session, they gave consent, got a brief introduction, answered a short questionnaire on their phone usage, and completed the first task. Every 2nd week, they received a short training session, followed by phase 1 of their respective task: encoding for ORR, and 1‐back task for MDT‐OS. After finishing phase 1, they were notified when the next phase was available, and could perform it straightaway or postpone if inconvenient; that is, there was a minimum delay of 24 h for the MDT‐OS and a minimum delay of 60 minutes for the ORR, but it was often extended by the participants. Phase 2 consisted of retrieval for ORR, and 2‐back task for MDT‐OS. It was followed by ratings regarding concentration, distraction throughout and subjective difficulty of the task. MDT‐OS; Mnemonic Discrimination Task for Objects and Scenes; ORR, objects‐in‐room‐recall task.

FIGURE 2

Smartphone‐based memory tasks. (A) MDT‐OS and (B) ORR test. MDT‐OS; Mnemonic Discrimination Test for Objects and Scenes; ORR, Objects‐In‐Room Recall.

FIGURE 3

Acceptability and user experience. Results from telephone‐based interviews with 38 participants focusing on their overall experience with the remote and unsupervised study.

FIGURE 4

Construct validity of the MDT‐OS and the ORR. Scatter plots showing relationships of the MDT‐OS Corrected hit rate with the (A) modified Preclinical Alzheimer's Cognitive Composite, (B) errors in the delayed 10‐word list recall test from ADAS‐cog, (C) Symbol Digit Modalities Test, (D) animal fluency, (E) Mini‐Mental State Examination as well as the (F) Mnemonic Discrimination Task for Objects and Scenes Corrected Hit Rate from an on‐site based task version that was performed during an MRI scan. Furthermore, scatter plots show relationships of the ORR‐DR with the (G) modified Preclinical Alzheimer's Cognitive Composite, (H) errors in the delayed 10‐word list recall test from ADAS‐cog, (I) Symbol Digit Modalities Test, (J) animal fluency, and the (K) Mini‐Mental State Examination. ADAS, Alzheimer's Disease Assessment Scale; ADAS‐cog, Alzheimer's Disease Assessment Scale—Cognitive Subscale; MDT‐OS; mnemonic discrimination test for objects and scenes; ORR, Objects‐In‐Room Recall; ORR‐DR, ORR delayed recall score.

FIGURE 5

Criterion validity of the MDT‐OS and the ORR‐DR. Scatter plots showing relationships between the MDT‐O and tau‐PET uptake in (A) area 35, (B) the anterior and (C) posterior hippocampus, as well as cortical thickness in area 35 (D) and anterior (E) and posterior hippocampal volume (F). Panel G shows the relationship of the MDT‐S with anterior hippocampal volume. Finally, panel (H), (I), and (J) illustrate relationships of the ORR‐DR and tau‐PET uptake in area 35 and the anterior hippocampus as well as the volume of the posterior hippocampus. A35, area 35; aHC, anterior hippocampus; pHC, posterior hippocampus; MDT‐OS; Mnemonic Discrimination Task for Objects and Scenes; ORR‐DR, Object‐In‐Room Recall–delayed recall; PET, positron emission tomography.