INTRODUCTION We investigated whether a multi-day learning curve (MDLC), derived from remote digital cognitive assessments over 1 week, adds prognostic value to plasma phosphorylated tau at threonine 217 (p-tau217) for predicting cognitive decline in cognitively unimpaired (CU) older adults. METHODS A total of 183 CU participants (mean age 73.3; 70% female) completed multi-day digital cognitive tests and provided plasma p-tau217. Linear mixed-effects models evaluated the independent and combined effects of baseline p-tau217 and MDLC on longitudinal cognitive decline over 2.5 years. RESULTS Adding MDLC to p-tau217 significantly explained longitudinal cognitive trajectory beyond p-tau217 alone. Although higher p-tau217 and lower MDLC were independently associated with steeper decline, individuals with both risk factors declined the fastest. Notably, those with elevated p-tau217 level but preserved MDLC remained stable, distinguishing them from those at imminent risk. DISCUSSION Combining digital learning metrics with plasma p-tau217 enables precise risk stratification. This approach identifies individuals most vulnerable to imminent decline, potentially enhancing screening for early preventive interventions.

OBJECTIVE:Self-appraisal of cognitive performance, a potentially useful marker of brain functioning, is typically assessed at a single time point where tests are naïve to what constitutes "good" or "bad" performance. Here, we determine whether familiarizing individuals with self-appraisal with daily memory testing for 7 days provide a more accurate estimate of cognitive functioning and mood. METHOD:Two hundred twenty-five participants (Mage ± SD: 74.1 ± 8.3 years; 66% female; median education 16.0 years) completed the online Boston Remote Assessment for NeuroCognitive Health, which included two associative memory tasks, for seven consecutive days. Each day, participants self-appraised their performance. At baseline, they completed various cognitive and mood measures. We computed Pearson's correlations between task performance and self-appraisal on Days 1 and 7 and used linear models to examine the relationship between self-appraisal scores and clinical measures. RESULTS:Accuracy (Day 1: 0.44 ± 0.12; Day 7: 0.81 ± 0.16) and self-appraisal (Day 1: 0.36 ± 0.15; Day 7: 0.70 ± 0.21) increased, as did the association between accuracy and self-appraisal, Day 1: correlation coefficient (r) = 0.22, 95% confidence interval (95% CI) [0.09, 0.34], p = .001; Day 7: r = 0.69, 95% CI [0.62, 0.76], p < .001. Self-appraisal scores on Day 7, but not Day 1, showed significant relationships with in-clinic measures. CONCLUSIONS:Repeated remote cognitive assessments may help elucidate individuals' capacities to refine their self-perception of cognitive performance during multiday learning. The weak association between accuracy and test-naïve self-appraisal warrants caution about using this metric cross-sectionally. Experienced self-appraisal could be especially relevant at the early stages of neurodegenerative diseases when subtle learning difficulties emerge and could improve our capacity to detect early meta-cognitive changes. (PsycInfo Database Record (c) 2025 APA, all rights reserved).

Timely identification of individuals at-risk for Alzheimer's disease (AD) is pivotal for secondary prevention and requires innovative approaches. Combining remote smartphone-based cognitive assessments with blood-based biomarkers holds promise for both sensitive and scalable detection of early AD-related cognitive changes. Here, we aimed to investigate whether the Boston Remote Cognitive Assessment of NeuroCognitive Health (BRANCH) captures cognitive changes associated with early AD pathophysiology as measured by plasma p -tau217. N  = 254 cognitively unimpaired older adults (age=74.5±8.6, 68% female, 16.6±2.4 years of education) from four well-characterized cohorts completed multi-day BRANCH on their personal device. Multiday BRANCH includes two associative memory tests (Face Name and Groceries Prices) and a processing speed test with an associative memory component (Digit Signs) with identical stimuli repeated for seven consecutive days. For each test, an MDLC score was computed using an area under the curve method combining day 1 performance with a non-linear learning trajectory over the subsequent six days. MDLCs for each individual test were averaged into a BRANCH Composite MDLC. All cohorts had standardized in-clinic cognitive test data available, from which a Preclinical Alzheimer's Cognitive Composite (PACC-5) score was derived. Concentrations of plasma p -tau217 were measured using the Meso Scale Discovery platform. Linear regression models adjusting for age, sex, years of education and study cohort were used to investigate the association between p -tau217 (log-transformed values) and BRANCH MDLC scores. For comparison, similar analyses were run with PACC-5 scores and p -tau217. Lower BRANCH Composite MDLC scores were associated with higher p -tau217 levels (corrected std. β = -0.23, 95%CI [-0.44 – -0.02], p  = 0.031) (Figure 1), which was primarily driven by the Digit Signs test (corrected std. β = -0.22, 95%CI [-0.42 – -0.02], p  = 0.031). In contrast, we did not find an association between the PACC-5 and p -tau217 (corrected std. β = -0.15, 95%CI [-0.37 – 0.07], p  = 0.187). These results complement our previous work that multi-day BRANCH may improve the detection of very subtle memory deficits that are associated with early AD pathophysiology. Combining a remote and sensitive cognitive paradigm like BRANCH with plasma biomarkers may facilitate scalable detection of those at risk for AD-related cognitive decline.

Individuals with preclinical Alzheimer’s disease (AD) show reduced practice effects on annually repeated neuropsychological testing, suggesting a decreased ability to learn over repeated exposures. Remote, digital testing enables the assessment of learning over more frequent time intervals, thereby facilitating a more rapid detection of those early learning deficits. We previously showed that multi-day learning on the Boston Remote Assessment for Neurocognitive Health (BRANCH) was indeed diminished in Αβ+ cognitively unimpaired (CU) older adults. Here, we further investigated the impact of tau pathology on BRANCH multi-day learning curves (MDLCs). N = 136 CU older adults (age = 73.4±7.6, 66% female, 16.6±2.4 years education) from three well-characterized cohorts completed multi-day BRANCH on their personal device. The assessment includes two associative memory tests (Face Name and Groceries Prices) and a processing speed test with an associative memory component (Digit Signs) with identical stimuli repeated for seven consecutive days. An MDLC score is computed using an area under the curve method allowing for the combination of day 1 performance with a non-linear learning trajectory over the subsequent six days. All participants had [11C]Pittsburgh compound-B and [18F]flortaucipir PET within 0.7±0.5 years of BRANCH and were classified as A ± (global amyloid burden, DVR cutoff 1.14) and T ± (inferior-temporal tau SUVr, cutoff 1.30), resulting in n = 91 A-T-, n = 29 A+T- and n = 16 A+T+. Linear regression analyses adjusting for age, sex, and education were used to examine differences in BRANCH day 1 and MDLC scores across A/T groups. No A/T group differences were detected using day 1 scores. However, MDLC scores increasingly diminished across groups, with the A+T- group performing marginally worse (ß = -0.04,95%CI[-0.08–0.01], p = 0.11) and the A+T+ group significantly worse (ß = -0.06,95%CI[-0.11–0.01], p = 0.03) than A-/T- (Figure 1). A+ status regardless of T-status was associated with diminished Digit Signs MDLCs, whereas being T+ drove worse performance on Face Name and Groceries Prices MDLCs (Table 1). Subtle differences in learning among CU older adults with different A/T biomarker profiles are observable using MDLCs. These findings further support the notion that a multi-day learning paradigm can provide unique information about cognition that is not captured using a single time-point assessment and is particularly relevant in preclinical AD.

INTRODUCTION:Accelerated long-term forgetting (LTF) might be an early marker of subtle memory changes in older adults at risk for Alzheimer's disease (AD). We leveraged remote, multi-day digital testing to characterize LTF in older adults and investigated its association with initial learning and AD imaging biomarkers. METHODS:One hundred four cognitively unimpaired older adults completed a face-name memory task for seven consecutive days and were asked to recognize face-name pairs 1 week later. LTF was computed as the number of correctly identified stimuli divided by a participant's maximum performance during learning. RESULTS:Better learning was associated with less LTF (β = 0.52, 95% confidence interval [CI]:0.34-0.71, p < 0.001). Accelerated LTF was associated with cortical thinning in AD-signature regions (β = 0.33, 95% CI: 0.13-0.52, p = 0.001), but associations with regional tau were more subtle. DISCUSSION:Remote, multi-day testing may facilitate the assessment of LTF as an early cognitive marker of preclinical AD, but further replication is needed. HIGHLIGHTS:Using digital, remote assessments, we evaluated long-term forgetting in cognitively unimpaired older adults. We found a potential association between long-term forgetting and tau in Alzheimer's disease (AD)-related regions. Assessing long-term forgetting may facilitate early detection of AD-related cognitive decline.

Remote, digital cognitive testing on an individual’s own device provides the opportunity to deploy previously understudied but promising cognitive paradigms in preclinical Alzheimer’s disease (AD). The Boston Remote Assessment for NeuroCognitive Health (BRANCH) captures a personalized learning curve for the same information presented over seven consecutive days. Here, we examined BRANCH multi-day learning curves (MDLCs) in 167 cognitively unimpaired older adults (age = 74.3 ± 7.5, 63% female) with different amyloid-β (A) and tau (T) biomarker profiles on positron emission tomography. MDLC scores decreased across ascending biomarker groups, with the A + T- group performing numerically worse (β = –0.24, 95%CI[–0.55,0.07], p = 0.128) and the A + T+ group performing significantly worse (β = –0.58, 95%CI[–1.06,–0.10], p = 0.018) than the A-T- group. Further, lower MDLC scores were associated with greater cortical thinning (β = 0.18, 95%CI[0.04,0.34], p = 0.013). Our results suggest that diminished MDLCs track with advanced AD pathophysiology, and demonstrate how a digital multi-day learning paradigm can provide novel insights about cognitive decline during preclinical AD.

Response time (RT), traditionally linked to age-related processing speed, may also reflect memory consolidation—stabilizing and reorganizing information for faster retrieval. Alzheimer's disease (AD) pathology, particularly in hippocampal-cortical networks, disrupts consolidation, potentially prolonging retrieval latency before overt memory decline. We examined whether memory recall RT for correct trials evolves with repeated learning and correlates with amyloid and tau burden in the brain. A total of 175 cognitively unimpaired older adults (Age=74.21±8.33, Education = 16.55±2.50, Female=66.9%) completed seven days of daily associative memory tests with the same stimuli each day using the Boston Remote Assessment for Neurocognitive Health (BRANCH). Because memory consolidation is expected to strengthen over repeated exposures, we considered Day 1 a “early-consolidation” phase and Days 5–7 a “late-consolidation” phase. Then, we compared RT-accuracy correlation across these phases. In a PET-imaged subsample ( n  = 150), late-consolidation RT (averaged RT across Days 5–7) was regressed on cortical amyloid, medial temporal tau, and neocortical tau, adjusting for age, sex, and education. Given age's known impact on processing speed, a Johnson–Neyman analysis was utilized to identify a potential age-specific association between RT and AD-related pathology. Recall RT shortened significantly from the early-consolidation phase (Day1: M=3.52s, SD=1.26) to the late-consolidation phase (Days 5–7: M=2.68s, SD=0.66), reflecting improved recall efficiency. The RT-accuracy correlation strengthened in the late-consolidation phase (Day1: ρ=-0.15, p  = 0.047, Days 5–7: ρ=-0.52, p  <0.001), supporting that RT reflects memory consolidation. Slower late-consolidation RT correlated with higher medial temporal tau (β=0.24, p  = 0.002) but not with neocortical tau (β=0.12, p  = 0.137) and amyloid (β=-0.11, p  = 0.527). An age-specific relationship emerged only for neocortical tau and late-consolidation RT among individuals <74 (β=0.29, p  = 0.023), whereas RT among those ≥74 was explained solely by age (β=0.36, p  <0.001). Even among correct responses, recall speed—a novel and underexplored metric—revealed critical insights: RT tracks memory consolidation and functions as a pathology-specific marker, complementing traditional accuracy measures susceptible to ceiling effects in individuals without overt cognitive decline. Furthermore, we found that tau burden concurrently affects RT alongside age effect, with pronounced effects in individuals under 74—highlighting its potential to detect early Alzheimer's disease pathology.

INTRODUCTION:We investigated whether memory recall response time (RT) on remote digital testing reflects decrements in memory function linked to Alzheimer's disease (AD) biomarkers. METHODS:One hundred seventy-five cognitively unimpaired participants (age = 74.21 ± 8.33; 66.9% female) completed daily associative memory tests over 7 days. We examined how RT changed with repeated learning and how this related to accuracy. In a positron emission tomography-imaged subsample (n = 150), linear regression evaluated associations between mean RT across Days 1 to 3 and Days 5 to 7 and cortical amyloid and medial temporal and neocortical tau. RESULTS:RT for correct items progressively decreased, with a stronger RT-accuracy association observed on later learning days than at baseline. Moreover, even after accuracy plateaued, RT continued improving. Slower RT was associated with higher medial temporal tau but not with amyloid burden. DISCUSSION:Recall RT provides additional insight into AD-related learning decrements beyond accuracy alone. Slower RT in the memory recall task possibly reflects tau burden. HIGHLIGHTS:We examined recall response time (RT) changes over 7 days of repeated digital learning tasks. RT reveals learning-related changes even after accuracy plateaus. RT correlates with medial temporal tau and is moderated by age for neocortical tau. Digital RT data provide insights beyond accuracy, detecting subtle cognitive changes. RT may serve as a novel digital marker for preclinical Alzheimer's disease.

Background The multi-day Boston Remote Assessment of Neurocognitive Health (BRANCH) is a remote, web-based assessment designed to capture the earliest cognitive changes in the preclinical stage of Alzheimer's disease (AD). It has been validated in unimpaired older adults, but as individuals progress on the AD continuum, assessments need to remain feasible and valid at different clinical stages. The focus of this study was to assess feasibility and validity of multi-day BRANCH in participants with and without cognitive impairment. Methods For seven days participants completed the BRANCH paradigm to capture a muti-day learning curve score. Participants also completed the mini-mental-status-exam (MMSE) and the Quick Dementia Rating Scale (QDRS). The primary cohort included 81 older adults: 38 with cognitive impairment (CI) and 43 cognitively-unimpaired (CU). A complementary replication cohort included 16 participants with consensus-defined mild cognitive impairment (MCI) and 47 demographically-matched cognitively unimpaired participants. Results Multi-day BRANCH was feasibile with 92 % or participants completing all seven days of testing. More CI than CU reported nervousness and found tasks slightly less enjoyable on Day 1, but ratings increased at a similar rate in both groups. Convergent validity was confirmed by a positive association between BRANCH and total MMSE and QDRS scores. There was a large effect size of group status on BRANCH (CI vs. CU; Cohen's d = 0.83) and per logistic regression, BRANCH significantly predicted group status (β = -1.49, p < 0.001); even more so between MCI and CU in the replication cohort. Conclusions Findings suggest that a remotely administered web-based assessment of multi-day learning is feasible and valid in participants with and without cognitive impairment.

Identifying clinically unimpaired individuals at greatest risk for short-term cognitive decline related to Alzheimer's disease is critical for early intervention. Doing so with a combination of remote digital cognitive testing and blood-based AD biomarkers would be an efficient and cost-effective approach. Here, we examined whether a digitally-collected multi-day learning curve (MDLC)—a sensitive digital measure of memory consolidation—combined with p -tau217 could predict cognitive decline over approximately two years. Two hundred cognitively unimpaired older adults, aged 74±8.2 years, completed seven days of daily remote cognitive testing via the Boston Remote Assessment for NeuroCognitive Health (BRANCH), from which MDLCs were derived. Baseline measures included the Preclinical Alzheimer's Cognitive Composite (PACC-5) and p -tau217 (analyzed with the Meso Scale Discovery platform). Annual follow-up of PACC-5 was conducted over an average of 2.3±0.78 years (range=1–5). “Cognitive decliners” were defined as those whose longitudinal PACC-5 slope was at most -0.1 standard deviations per year. First, linear mixed-effect models controlling for covariates assessed whether baseline MDLCs and p -tau217 each predicted changes in PACC-5 over time. Second, receiver operating characteristic (ROC) analyses first tested p -tau217 alone to identify cognitive decliners and then examined whether adding MDLCs improved predictive performance. Longitudinal cognitive decline was associated with both higher baseline p -tau217 (β = -0.032, 95%CI [-0.058, -0.007], p  = 0.015) and lower baseline MDLCs (β = 0.046, 95%CI [0.019, 0.073], p  = 0.001). In predicting cognitive decliners ( n  = 15, 8%), p -tau217 showed an AUC of 0.63 (95%CI: 0.46–0.80), while adding the MDLCs increased the discriminative accuracy to an AUC of 0.82 (95%CI: 0.70–0.94), a statistically significant difference (DeLong's one-sided test = -1.828, p  = 0.034). A remote, web-based cognitive assessment of memory consolidation explains unique variance in cognitive decline over approximately 2 years when paired with p -tau217 amongst clinically unimpaired older adults. While p -tau217 alone could predict who might experience cognitive decline, adding an MDLC, which detects deficits related to AD pathology in cognitively normal individuals, further increased the identification accuracy. These results confirm the utility of pairing sensitive digital cognitive assessments with plasma markers to better identify individuals at greatest risk for imminent cognitive decline.

Accelerated long-term forgetting (LTF) is characterized by unimpaired retention of information after short-term delays (e.g., 20-30 minutes) with increased forgetting at longer intervals (e.g., weeks to months). Previous studies have suggested that assessing LTF may provide a useful marker of preclinical Alzheimer’s disease (AD). However, assessing LTF over longer intervals is challenging using standardized in-clinic paper-pencil cognitive tests. Here, we leverage remote, digital cognitive testing to investigate LTF at different timepoints and its association with AD biomarkers in cognitively unimpaired (CU) older adults. N = 61 CU older adults (age = 76.5±8.5, 67.2% female, 18% Aβ+) with amyloid (PiB) and tau (FTP) PET completed the Boston Remote Assessment for NeuroCognitive Health (BRANCH) at-home on a personal device for seven consecutive days, including a Face-Name Matching Task with identical stimuli each day. Learning across seven days was quantified using a previously validated multi-day learning curve (MDLC) metric. Participants were asked to recall previously learned face-name pairs after 1-week (Median = 8(IQR = 7-35) days) and after 6 months (Median = 6.67(IQR = 6.35,7.85) months). LTF was computed by dividing the percentage of correctly recalled face-name pairs by a participant’s maximum performance during the 7-day learning phase. We used linear regression models to examine the associations between LTF and initial MDLCs, global amyloid burden, entorhinal cortex (EC) and inferior-temporal (IT) tau deposition, correcting for age, sex, and education when needed (covariates with a p-value > 0.1 were excluded). Better initial MDLCs were associated with less accelerated LTF after 1 week (β = 0.55,95%CI[0.00-1.09], p = 0.048), but not after 6 months (β = 0.34,95%CI[-0.35–1.02], p = 0.329). There were no associations between LTF and global amyloid burden. However, higher EC tau was associated with accelerated 1-week LTF (β = -0.18,95%CI[-0.31—0.05], p = 0.009) but not with extended LTF(β = -0.14,95%CI[-0.35–0.07], p = 0.172). In contrast, higher IT tau was associated with accelerated LTF after 6 months (β = -0.38,95%CI[-0.75—0.01], p = 0.045), but not after 1 week (β = -0.15,95%CI[-0.43–0.13], p = 0.280) (Figures 1-2). We showed that 1-week LTF is associated with initial learning and EC tau, and LTF at an extended interval of 6 months was associated with IT tau. This suggests that accelerated LTF may be an early cognitive sign in preclinical AD, but that assessing LTF over different time intervals may reveal unique information.

Remote digital testing provides the opportunity to deploy memory paradigms that mimic learning in everyday life by exposing participants to repeated stimuli over frequent intervals. Here, we used the Boston Remote Assessment for Neurocognitive Health (BRANCH) multi-day learning curve (MDLC) paradigm and investigated whether repeating MDLCs over time could capture subtle cognitive changes in preclinical Alzheimer’s disease (AD). N = 223 cognitively unimpaired older adults (age = 74±8.1, 65% female, MMSE 29±1.4) with standardized cognitive testing and amyloid and tau PET from the Harvard Aging Brain Study completed a modified version of the Face-Name Association Examination (FNAME) at-home for seven consecutive days on a personal device. After 10.8±2.6 months a subsample (n = 54; age = 75±8.8, 61% female, MMSE = 29±1, 19% Aβ+, n = 51 with tau PET) completed a second version of the FNAME with new stimuli for seven consecutive days. A summary learning curve metric for each MDLC (baseline and follow-up) was computed using an Area Under the Curve (AUC) method allowing for the combination of Day 1 performance and a learning trajectory over the subsequent six days. We used linear mixed effect (LME) models on the MDLC AUCs to investigate change in MDLCs and compared this to change in Day 1 performance of the MDLCs. We also ran LME models adjusting for demographic factors to examine whether baseline Preclinical Alzheimer’s Cognitive Composite-5 (PACC5) performance and amyloid and tau burden were associated with change in MDLCs. Overall, MDLCs diminished over time (Time = -0.028, 95%CI[-0.056 – -0.001], p = 0.047) which was not detected by change in Day 1 performance of the MDLC (Time = -0.026, 95%CI[-0.063 – 0.012], p = 0.117). A lower MDLC at follow-up compared to baseline was associated with elevated amyloid (Aβ+ Group*Time = -.08, [95%CI = -0.15- -0.01], p = 0.042) (Figure 1) and worse PACC5 performance (PACC5*Time = 0.05, 95%CI = [0.01-0.09], p = 0.029) (Figure 2). Greater entorhinal tau burden was associated with a lower baseline MDLC (Tau = -0.18, 95%CI[-0.34 – -0.03], p = 0.025) but not with MDLC change over time. These results highlight how capturing multiple datapoints over days detects short-term cognitive changes in preclinical AD that are undetectable using standard single timepoint assessments. Paradigms such as BRANCH MDLC may thereby offer more sensitive cognitive outcome measures for AD secondary prevention trials.

OBJECTIVE:Unsupervised remote digital cognitive assessment makes frequent testing feasible and allows for measurement of learning over repeated evaluations on participants' own devices. This provides the opportunity to derive individual multiday learning curve scores over short intervals. Here, we report feasibility, reliability, and validity, of a 7-day cognitive battery from the Boston Remote Assessment for Neurocognitive Health (Multiday BRANCH), an unsupervised web-based assessment. METHOD:Multiday BRANCH was administered remotely to 181 cognitively unimpaired older adults using their own electronic devices. For 7 consecutive days, participants completed three tests with associative memory components (Face-Name, Groceries-Prices, Digit Signs), using the same stimuli, to capture multiday learning curves for each test. We assessed the feasibility of capturing learning curves across the 7 days. Additionally, we examined the reliability and associations of learning curves with demographics, and traditional cognitive and subjective report measures. RESULTS:Multiday BRANCH was feasible with 96% of participants completing all study assessments; there were no differences dependent on type of device used (t = 0.71, p = .48) or time of day completed (t = -0.08, p = .94). Psychometric properties of the learning curves were sound including good test-retest reliability of individuals' curves (intraclass correlation = 0.94). Learning curves were positively correlated with in-person cognitive tests and subjective report of cognitive complaints. CONCLUSIONS:Multiday BRANCH is a feasible, reliable, and valid cognitive measure that may be useful for identifying subtle changes in learning and memory processes in older adults. In the future, we will determine whether Multiday BRANCH is predictive of the presence of preclinical Alzheimer's disease. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Objective:Unsupervised remote digital cognitive assessment makes frequent testing feasible and allows for measurement of learning across days on participants’ own devices. More rapid detection of diminished learning may provide a potentially valuable metric that is sensitive to cognitive change over short intervals. In this study we examine feasibility and predictive validity of a novel digital assessment that measures learning of the same material over 7 days in older adults.Participants and Methods:The Boston Remote Assessment for Neurocognitive Health (BRANCH) (Papp et al., 2021) is a web-based assessment administered over 7 consecutive days repeating the same stimuli each day to capture multi-day-learning slopes. The assessment includes Face-Name (verbal-visual associative memory), Groceries-Prices (numeric-visual associative memory), and Digits-Signs (speeded processing of numeric-visual associations). Our sample consisted of200 cognitively unimpaired older adults enrolled in ongoing observational studies (mean age=74.5, 63% female, 87% Caucasian, mean education=16.6) who completed the tasks daily, at home, on their own digital devices. Participants had previously completed in-clinic paper-and-pencil tests to compute a Preclinical Alzheimer’s Cognitive Composite (PACC-5). Mixed-effects models controlling for age, sex, and education were used to observe the associations between PACC-5 scores and both initial performance and multi-day learning on the three BRANCH measures.Results:Adherence was high with 96% of participants completing all seven days of consecutive assessment; demographic factors were not associated with differences in adherence. Younger participants had higher Day 1 scores all three measures, and learning slopes on Digit-Sign. Female participants performed better on Face-Name (T=3.35, pConclusions:Seven days of remote, brief cognitive assessment was feasible in a sample of cognitively unimpaired older adults. Although various demographic factors were associated with initial performance on the tests, multi-day-learning slopes were largely unrelated to demographics, signaling the possibility of its utility in diverse samples. Both initial performance and learning scores on an associative memory and processing speed test were independently related to baseline cognition indicating that these tests’ initial performance and learning metrics are convergent but unique in their contributions. The findings signal the value of measuring differences in learning across days as a means towards sensitively identifying differences in cognitive function before signs of frank impairment are observed. Next steps will involve identifying the optimal way to model multi-day learning on these subtests to evaluate their potential associations with Alzheimer’s disease biomarkers.

Lower self-ratings of performance on single timepoint neuropsychological measures have been previously associated with elevated AD biomarkers, aligned with the notion of a period of heightened awareness of cognitive decline in preclinical AD. With the rise of digital tools, cognitive assessments can now be collected over multiple days to examine learning over repeated exposures, thought to be diminished in preclinical AD. However, it is unknown whether daily self-appraisal of performance would mimic reduced learning observed in individuals with elevated biomarkers. In this study, we used the Boston Remote Assessment for Neurocognitive Health (BRANCH) platform to capture multi-day learning curves and to analyze whether self-evaluation of performance was associated with amyloid and tau PET in a sample of cognitively unimpaired older individuals. Our sample consisted of 193 cognitively unimpaired individuals (mean age = 74.1, 64.8% female, 85.5% Caucasian) already enrolled in a longitudinal observational study who underwent PiB- and Flortaucipir-PET. Using the BRANCH platform remotely, participants completed two paired associative learning tests with repeated stimuli on their own digital devices for 7 consecutive days. A self-evaluation was prompted at the end of each assessment, asking participants to rate their performance from 0-100. Multi-day learning curve metrics for objective performance (obj-MDLC) and self-appraisal (subj-MDLC) were calculated for each participant. We examined correlations between obj-MDLC and subj-MDLC and using linear mixed-effects models investigated the relationship between MDLCs, global β-amyloid (Aβ) and tau burden in entorhinal (ET) cortex. Obj-MDLC was moderately correlated with the subj-MDLC (r = 0.61, p < 0.001) and showed that, on average, participants under-estimated their performance on the tasks. Both higher Aβ and higher ET tau were associated with a decreased obj-MDLC (Day*Aβ = -0.08, p = 0.015; Day*ET = -0.06, p = 0.020) and subj-MDLC (Day*Aβ = -0.09, p = 0.021; Day*ET = -0.09, p = 0.003). Participant’s self-appraisal of their performance on the BRANCH tasks across days showed significant associations with tau and amyloid that were consistent with those seen with actual performance on the BRANCH tasks. These findings demonstrate that utilizing daily self-appraisals, in addition to objective assessment, provides complementary information in early preclinical AD.

OBJECTIVE:This study was undertaken to determine whether assessing learning over days reveals Alzheimer disease (AD) biomarker-related declines in memory consolidation that are otherwise undetectable with single time point assessments. METHODS:Thirty-six (21.9%) cognitively unimpaired older adults (aged 60-91 years) were classified with elevated β-amyloid (Aβ+) and 128 (78%) were Aβ- using positron emission tomography with 11C Pittsburgh compound B. Participants completed the multiday Boston Remote Assessment for Neurocognitive Health (BRANCH) for 12 min/day on personal devices (ie, smartphones, laptops), which captures the trajectory of daily learning of the same content on 3 repeated tests (Digit Signs, Groceries-Prices, Face-Name). Learning is computed as a composite of accuracy across all 3 measures. Participants also completed standard in-clinic cognitive tests as part of the Preclinical Alzheimer's Cognitive Composite (PACC-5), with 123 participants undergoing PACC-5 follow-up after 1.07 (standard deviation = 0.25) years. RESULTS:At the cross-section, there were no statistically significant differences in performance between Aβ+/- participants on any standard in-clinic cognitive tests (eg, PACC-5) or on day 1 of multiday BRANCH. Aβ+ participants exhibited diminished 7-day learning curves on multiday BRANCH after 4 days of testing relative to Aβ- participants (Cohen d = 0.49, 95% confidence interval = 0.10-0.87). Diminished learning curves were associated with greater annual PACC-5 decline (r = 0.54, p < 0.001). INTERPRETATION:Very early Aβ-related memory declines can be revealed by assessing learning over days, suggesting that failures in memory consolidation predate other conventional amnestic deficits in AD. Repeated digital memory assessments, increasingly feasible and uniquely able to assess memory consolidation over short time periods, have the potential to be transformative for detecting the earliest cognitive changes in preclinical AD. ANN NEUROL 2024;95:507-517.

Estimating the pose of multiple animals is a challenging computer vision problem: frequent interactions cause occlusions and complicate the association of detected keypoints to the correct individuals, as well as having extremely similar looking animals that interact more closely than in typical multi-human scenarios. To take up this challenge, we build on DeepLabCut, a popular open source pose estimation toolbox, and provide high-performance animal assembly and tracking—features required for robust multi-animal scenarios. Furthermore, we integrate the ability to predict an animal’s identity directly to assist tracking (in case of occlusions). We illustrate the power of this framework with four datasets varying in complexity, which we release to serve as a benchmark for future algorithm development.

Unsupervised remote digital cognitive assessment makes frequent testing feasible and allows for measurement of learning across days on participants’ own devices. More rapid detection of diminished learning may provide a potentially valuable metric for clinical trials seeking to capture change over shorter intervals. Here, we assess whether clinically normal (CN) participants differ in their learning of the same stimuli across daily exposures depending on their β-amyloid (Aβ) burden. The Boston Remote Assessment for Neurocognitive Health (BRANCH) (Papp et al., 2021) is a web-based assessment administered over 7 consecutive days repeating the same stimuli each day to capture multi-day-learning-curves (MDLC). The composite score includes accuracy on three cross-modal associative memory tasks: face-name, groceries-prices, digits-signs. Our sample consisted of 192 CN older adults enrolled in ongoing observational studies (mean age = 74.0, 63% female, 87% Caucasian, mean education = 16.6) who completed the tasks daily at home on their own digital devices. Participants had previously completed in-clinic paper-and-pencil tests to compute a Preclinical Alzheimer’s Cognitive Composite (PACC-5) as well as PET imaging with 11 C-Pittsburg Compound-B to estimate globalAβ burden and classify participants as Aβ+ and Aβ-. Mixed-effects models controlling for age, sex, and education were used to observe the associations between Aβ status and BRANCH-MDLC; and AUC analyses compared classification of amyloid using BRANCH-MDLC vs. PACC-5. Adherence was high with 95% of participants completing all seven days of consecutive assessment. Aβ+ status (n = 26) was associated with a lower BRANCH-MDLC across 7-days (t = -4.03, p<.001) compared with Aβ- status (n = 148). Amyloid status was not associated with difference in Day 1 BRANCH or PACC-5 performance. Furthermore, BRANCH-MDLC showed a significant ability to classify individuals as Aβ+ vs. Aβ- (AUC = 0.71, p = <.001) compared to a single-timepoint pencil-and-paper composite (PACC-5) (AUC = 0.52, p = 0.56). Diminished task learning across seven days was associated with greater amyloid pathology, whereas Day 1 of BRANCH or an in-person supervised one-time neuropsychological assessment were not associated with elevated amyloid. These findings signal the value of measuring differences in learning across days as a means towards capturing biomarker-associated memory decrements and identifying those in the preclinical stage of AD.

Digital cognitive testing completed on an individual’s own device, independently, and remotely is a highly appealing solution to sensitively capture early memory changes in preclinical AD. However, few digital assessments have been designed specifically for a preclinical AD population and validated against paper-and-pencil measures and relevant AD biomarkers. Here, we describe initial validation steps for the Boston Remote Assessment of Neurocognitive Health (BRANCH), web-based cognitive testing targeting cognitive domains with AD susceptibility (e.g., cross-modal associative memory, semantically facilitated learning and recall, and pattern separation) and using task stimuli relevant to everyday life. To determine the validity of BRANCH, we explored correlations between remote BRANCH to in-clinic paper and pencil measures and PET amyloid burden. A link to BRANCH was either texted or emailed to 128 clinically normal (CN) older adults participating in the Harvard Aging Brain Study. Participants completed the 4 BRANCH measures (modified Face Name Test, groceries test, categories test, signs test) over a mean 19 minutes. Participants had previously completed in-clinic paper and pencil tests to compute a Preclinical Alzheimer’s Cognitive Composite (PACC-5) as well as PET imaging with 11C Pittsburg Compound-B to estimate global amyloid burden. A composite of accuracy across BRANCH tasks was computed. Participants had a mean age of 74.38(Range 51-89), were 82% Caucasian, and 60.7% female. They completed BRANCH on smartphones (26%), tablets (18%), laptops (26%), and desktops(29%). Only 2.59% of participants reported difficulty completing BRANCH on a post-test survey. Lower BRANCH composite performance was associated with worse performance on in-person paper and pencil measures (Figure 1; PACC; r=0.637, p<0.001). Lower BRANCH performance was associated with greater amyloid burden (r=-0.275, p=0.003). A digital memory assessment with ecologically-valid tasks and stimuli is feasible for CN older adults to complete independently on their own devices. The relatively strong correlation observed between BRANCH and PACC suggests that BRANCH captures valid information about cognitive performance despite being collected remotely on an individual’s own device. The significant association between BRANCH and amyloid burden suggests that these remotely captured tasks may be promising tools to detect and track AD-specific cognitive decrements on a larger scale.

Abstract Introduction Unsupervised digital cognitive testing is an appealing means to capture subtle cognitive decline in preclinical Alzheimer's disease (AD). Here, we describe development, feasibility, and validity of the Boston Remote Assessment for Neurocognitive Health (BRANCH) against in‐person cognitive testing and amyloid/tau burden. Methods BRANCH is web‐based, self‐guided, and assesses memory processes vulnerable in AD. Clinically normal participants (n = 234; aged 50–89) completed BRANCH; a subset underwent in‐person cognitive testing and positron emission tomography imaging. Mean accuracy across BRANCH tests (Categories, Face‐Name‐Occupation, Groceries, Signs) was calculated. Results BRANCH was feasible to complete on participants’ own devices (primarily smartphones). Technical difficulties and invalid/unusable data were infrequent. BRANCH psychometric properties were sound, including good retest reliability. BRANCH was correlated with in‐person cognitive testing (r = 0.617, P < .001). Lower BRANCH score was associated with greater amyloid (r = –0.205, P = .007) and entorhinal tau (r = –0.178, P = .026). Discussion BRANCH reliably captures meaningful cognitive information remotely, suggesting promise as a digital cognitive marker sensitive early in the AD trajectory.