JMIR mHealth and uHealth

BackgroundVoice-based systems such as Amazon Alexa may be useful to collect self-reported information in realtime from participants of epidemiology studies, using verbal input. We demonstrate the technical feasibility of using Alexa, investigate participant acceptability, and provide an initial evaluation of the validity of the collected data. We use food and drink information as an exemplar. MethodsWe recruited 45 staff and students at the University of Bristol (UK). Participants were asked to tell Alexa what they ate or drank for 7 days, and also to submit this information using a web form. Questionnaires asked for basic demographic information and about their experience during the study and acceptability of using Alexa. ResultsOf the 37 participants with valid data, most were 20-39 years old (N=30; 81%) and 23 (62%) were female. Across 29 participants with Alexa and web entries corresponding to the same intake event, 357 Alexa entries (61%) contained the same food/drink information as the corresponding web entry. Participants often reported that Alexa interjected, and this was worse when entering the food and drink information compared with the event date and time. The majority said they would be happy to use a voice-controlled system for future research. ConclusionsWhile usability of our skill was poor, largely due to the conversational nature and because Alexa interjected if there was a pause in speech, participants were mostly open to participating in future research studies using Alexa. Many more studies are needed, in particular, to trial less conversational interfaces. KEY MESSAGESO_LIOver the last few years voice-controlled smart systems have emerged giving the possibility of collecting self-reported data using a voice-based approach. C_LIO_LIWe successfully collected epidemiology food and drink information in real-time, demonstrating that voice-based collection of self-reported data is technically feasible. C_LIO_LIThe conversational design of our skill meant that usability was poor, for example, most participants (86%) reported that Alexa either occasionally, often or always interjected during use, and the majority of participants who had previously used a paper diary or my fitness pal did not find Alexa as efficient to use compared with these approaches. C_LIO_LIAfter participating in this study, the majority of participants would be happy to use Alexa again, either at home or on a wearable device. C_LIO_LIOur results highlight that further work is needed to evaluate use of voice-based systems, including comparing Amazon Alexa with the Google Assistant, and trialling less conversational interfaces. C_LI

PurposeThe SmartSleep Study is established to comprehensively assess the impact of night-time smartphone use on sleep patterns and health. An innovative combination of large-scale repeated survey information, high-resolution sensor-driven smartphone data, in-depth clinical examination and registry linkage allow for detailed investigations into multisystem physiological dysregulation and long-term health consequences associated with night-time smartphone use and sleep impairment. ParticipantsThe SmartSleep Study consists of three interconnected data samples, which combined include 30,673 individuals with information on smartphone use, sleep and health. Subsamples of the study population also include high-resolution tracking data (n=5,927) collected via a customized app and deep clinical phenotypic data (n=245). A total of 7,208 participants will be followed in nationwide health registries with full data coverage and long-term follow-up. Findings to dateWe highlight previous findings on the relation between smartphone use and sleep in the SmartSleep Study, and we evaluate the interventional potential of the citizen science approach used in one of the data samples. We also present new results from an analysis in which we utilize 803,000 data-points from the high-resolution tracking data to identify clusters of temporal trajectories of night-time smartphone use that characterize distinct use patterns. Based on these objective tracking data, we characterize four clusters of night-time smartphone use. Future plansThe unprecedented size and coverage of the SmartSleep Study allow for a comprehensive documentation of smartphone activity during the entire sleep span. The study will be expanded by linkage to nationwide registers, which will allow for further investigations into the long-term health and social consequences of night-time smartphone use. We also plan new rounds of data collection in the coming years. STRENGTHS and LIMITATIONS of this studyO_LIThe unprecedented size and coverage of the SmartSleep Study allow for a comprehensive objective and subjective documentation of smartphone activity during the entire sleep span. C_LIO_LIThe data in the SmartSleep Study are sampled by three different strategies, which allow us to test robustness and validate findings across samples. This aligns with the principles of triangulation, which aims at obtaining more reliable answers to complex research questions through the integration of results from different approaches with different sources of bias. C_LIO_LIThe SmartSleep Study is readily available for research projects: the data sources have already been linked, the data have been cleaned and prepared for future analyses. C_LIO_LIThe SmartSleep Study is not fully representative of the general population due to the sampling procedures, and we are currently creating weights that can be used in the statistical analysis to compensate for this imbalance. C_LI

Background/ObjectivesMobility is a meaningful aspect of an individuals health whose quantification can provide clinical insights. Wearable sensor technology can quantify walking behaviors (a key aspect of mobility) through continuous passive monitoring. Our objective was to characterize the accuracy and reliability performance of a suite of digital measures of walking behaviors, as critical aspects in the practical implementation of digital measures into clinical studies. MethodsWe collected data from a wrist-worn device (the Verily Study Watch) worn for multiple days by a cohort of volunteer participants without history of gait/walking impairment in a real world setting. Based on step measurements computed in 10-second epochs from sensor data, we generated individual daily aggregates (participant-days) to derive a suite of measures of walking: step count, walking bout duration, number of total walking bouts, number of long walking bouts, number of short walking bouts, peak 30-minute walking cadence, peak 30-minute walking pace. To characterize accuracy of the measures, we examined agreement with truth labels generated by a concurrent, ankle-worn, reference device (Modus StepWatch 4) with known low error, calculating the following metrics: Intraclass Correlation Coefficient (ICC), Pearson R, Mean Error (ME), Mean Absolute Error (MAE). To characterize the reliability, we developed a novel approach to identify the time to reach a reliable readout (time-to-reliability) for each measure. This was accomplished by computing mean values over aggregation scopes ranging from 1-30 days, and analyzing test-retest reliability based on ICCs between adjacent (non-overlapping) time windows for each measure. ResultsIn the accuracy characterization, we collected data for a total of 162 participant-days from a testing cohort (N=35 participants; median observation time, 5 days). Agreement with the reference device-based readouts in the testing subcohort (n=35) for the eight measurements under evaluation, as reflected by ICCs, ranged between 0.7-0.9; Pearson R values were all greater than 0.75. For the time-to-reliability characterization, we collected data for a total of 15,120 participant days (overall cohort N=234; median observation time, 119 days). Here, all digital measures achieved an ICC between adjacent readouts > 0.75 by 16 days of wear time. ConclusionsWe characterized accuracy and reliability of a suite of digital measures that provides comprehensive information about walking behaviors in real-world settings. These results, which report the level of agreement with high-accuracy reference labels and the time duration required to establish reliable measure readouts, can guide practical implementation of these measures into clinical studies. Well-characterized tools to quantify walking behaviors in research contexts can provide valuable clinical information about general population cohorts and patients with specific conditions.

Sleep health is essential for older adults. However, validity of wrist- and waist-worn devices for assessing sleep under free-living conditions remains unclear. This study evaluated the accuracy of a wrist-worn smartwatch (Silmee W22) and a waist-worn activity monitor (MTN-221) in measuring key sleep parameters, using portable electroencephalography (EEG; Insomnograf K2) as the reference. Healthy older adults wore all devices simultaneously for at least three nights. Total sleep time, sleep onset latency, wake after sleep onset, and sleep efficiency were analyzed using Bland- Altman plots, multilevel models, and intraclass correlation coefficients (ICCs). Fifty-five participants completed the study, yielding valid EEG-paired data for 49 participants with Silmee W22 (238 nights) and 53 with MTN-221 (265 nights). Silmee W22 overestimated total sleep time by 35 min and sleep efficiency by 8.1%, whereas MTN-221 overestimated it by 3 min and sleep efficiency by 1.0%. Both devices underestimated sleep onset latency and wake after sleep onset, with greater discrepancies observed as the estimated values increased. ICCs for total sleep time were 0.60-0.75 for Silmee W22 and 0.66-0.79 for MTN-221, while agreement for sleep onset latency and wake after sleep onset remained lower. While Silmee W22 did not provide sufficiently accurate estimates of total sleep time, MTN-221 yielded estimates that may offer practical benefits for large-scale sleep monitoring in older adults. In both devices, estimates of sleep onset latency, wake after sleep onset, and sleep efficiency should be interpreted with caution due to misclassification of quiet wakefulness. Further algorithm refinement is warranted.

Mobile health applications, which employ wireless technology for healthcare, can aid behaviour change and subsequently improve health outcomes. Mobile health applications have been developed to increase physical activity, but are rarely grounded on behavioural theory and employ simple techniques for personalisation, which has been proven effective in promoting behaviour change. In this work, we propose a theoretically driven and personalised behavioural intervention delivered through an adaptive knowledge-based system. The behavioural system design is guided by the Behavioural Change Wheel and the Capability-Opportunity-Motivation behavioural model. The system exploits the ever-increasing availability of health data from wearable devices, point-of-care tests and consumer genetic tests to issue highly personalised physical activity and sedentary behaviour recommendations. To provide the personalised recommendations, the system firstly classifies the user into one of four diabetes clusters based on their cardiometabolic profile. Secondly, it recommends activity levels based on their genotype and past activity history, and finally, it presents the user with their current risk of developing cardiovascular disease. In addition, leptin, a hormone involved in metabolism, is included as a feedback biosignal to personalise the recommendations further. As a case study, we designed and demonstrated the system on people with type 2 diabetes, since it is a chronic condition often managed through lifestyle changes, such as physical activity increase and sedentary behaviour reduction. We trained and simulated the system using data from diabetic participants of the UK Biobank, a large-scale clinical database, and demonstrate that the system could help increase activity over time. These results warrant a real-life implementation of the system, which we aim to evaluate through human intervention.

Wearable-derived physiological features have been associated with disease risk, but most current studies focus on single conditions, limiting understanding of cross-disease patterns. This study adopts a trans-diagnostic approach to examine whether wearable data capture shared and condition-specific physiological signatures across multiple chronic conditions spanning physical and mental health, and then evaluates the utility of these features for disease classification. A total of 9,301 patients with at least 21 days of consecutive FitBit data from the All of Us Controlled Tier Dataset version 8 were analyzed. Disease subcohorts included cardiovascular disease (CVD), diabetes, obstructive sleep apnea (OSA), major depressive disorder (MDD), anxiety, bipolar disorder, and attention-deficit/ hyperactivity disorder (ADHD), chosen based on prevalence and relevance. Logistic regression and XGBoost models were fitted for each disease subcohort versus the control cohort. We found that compared to using just baseline demographic and lifestyle features, incorporating wearable-derived features enabled improved classification performance in all subcohorts for both models, except for ADHD where improvement was mainly observed for ROC-AUC in logistic regression model likely due to the smaller sample size in ADHD subcohort. The largest performance gains were observed in MDD (increase in ROC-AUC of 0.077 for Logistic regression, 0.071 for XGBoost; p < 0.001) and anxiety (increase in ROC-AUC of 0.077 for logistic regression, 0.108 for XGBoost; p < 0.001). This study provides one of the first comprehensive transdiagnostic evaluations of wearable-derived features for disease classification, highlighting their potential to enhance risk stratification in the real-world setting as a practical complement to clinical assessments and providing a foundation to explore more fine-grained wearable data. Author summaryWearable devices such as fitness trackers and smartwatches are becoming increasingly popular and affordable, providing continuous measurements of heart rate, physical activity, and sleep. Alongside the growing digitization of health records, this creates new opportunities for large-scale, real-world health studies. In this study, we analyzed wearable-derived physiological patterns across a range of chronic conditions spanning both physical and mental health to better understand how these signals relate to disease risk. We found that incorporating wearable-derived heart rate, activity and sleep features improved disease risk classification across several conditions, with particularly strong gains for major depressive disorder and anxiety. By examining how individual features contributed to model predictions, we also identified meaningful associations between physiological signals and disease risk. For example, both duration and day-to-day variation of deep and rapid eye movement (REM) sleep were associated with increased risk in certain conditions. Our study supports the development of real-time, automated tools to assess disease risk alongside clinical care.

The World Health Organization identifies unhealthy behaviors, such as smoking, as significant risk factors contributing to mortality and morbidity, underscoring the necessity to adopt healthier habits. The increasing prevalence of health applications (apps) presents opportunities for promoting healthier lifestyles. Notably, personalized mobile health (mHealth) interventions can enhance user engagement and their effectiveness. Our scoping review aims to contribute to guide the personalization of mHealth interventions for health behavior change by defining which mechanisms should be favored for a given user profile. Online databases were searched to identify articles published between 2008 and 2024 describing the topic of personalization, behavior change apps and mobile app mechanisms. Of 1806 articles identified, 18 articles were retained. We then categorized the mechanisms and user profiles described in the selected articles into existing taxonomies. Finally, the relationship between the user profiles and mechanisms were reported. The four user profiles identified included personality and gamer profiles. Twenty-one mechanisms extracted from the articles were categorized as behavioral change techniques, gamification or mobile app mechanisms, with limited numbers of preference relations between mechanisms and user profiles. The relation matrix was not complete and covered only 51% of possible relations: game mechanisms, 30%; behavioral change techniques, 16%; and app mechanisms, 5%. Two user profiles, the Big Five (18%) and Hexad scale (20%), covered 38% of relations, whereas the two remaining user profiles contributed to the remaining 13%. Social mechanisms, including competition, cooperation and social comparison, exhibit strong connections to user profiles and are pivotal in persuasive system design. Self-efficacy theory links mechanisms such as self-monitoring, social persuasion and rewards to behavior change. However, only 51% of potential relationships between profiles and mechanisms were identified. Adapting mHealth content based on user profiles requires reliable personality assessments and privacy-conscious data collection to enable personalized, profile-specific interventions for improved outcomes. Author SummaryThe promotion of healthy behavior, as well as addressing health risk factors that contribute to mortality, such as sedentary lifestyles, has led to a proliferation of mHealth apps. These apps have the potential to facilitate behavior change and offer a variety of features, including reminders, progress tracking and personalized interventions, which have been demonstrated to enhance user engagement and adherence. Personalization is of critical importance in the process of adapting interventions to align with the specific characteristics and needs of individual user profiles. The use of tailored messages and feedback has been demonstrated to be more effective than the use of generic ones, particularly in the context of promoting physical activity and weight loss. The incorporation of game design elements is also a prevalent feature in health apps, with evidence suggesting that it positively impacts on user engagement and motivation. However, there is a lack of comprehensive frameworks that provide guidance on their implementation in mHealth interventions. Here, we aim to optimize the effectiveness of interventions designed to facilitate health behavior change by defining game mechanisms, behavior change techniques and app mechanisms employed to personalize apps based on user profiles.

Physical activity (PA) is recognised for providing several health benefits in humans, mainly preventing and controlling chronic non-communicable diseases. However estimating PA is a challenging and expensive task. An alternative would be to devise a model to estimate PA using actigraphy devices calibrated from an initially validated model. This has been previously done to a number of devices, including ActiGraph(R) GT3X+. In this study we aimed at validating ActTrust(R) against the widely used GT3X+ and comparing activity counts to metabolic equivalents (METs) derived from indirect calorimetry during treadmill walking and running. Fifty-six young adults (34 men, 22 women) participated in controlled effort exercises including light, moderate, vigorous, and very vigorous activity intensities. We developed a general linear model to estimate energy expenditure (EE) from movement count of combinations of GT3X+ and ActTrust devices placed at hip or wrist. We then estimated cut-off points for each intensity range. Our results showed correlations between treadmill speed, METs, and movement counts across all devices and placements combinations. Our proposed model performed well with balanced accuracies above 0.77 for all intensity ranges and over 0.9 for light and moderate activity. This is the first study to model estimate and validate PA intensity thresholds on ActTrust(R) devices. Our findings support the use of ActTrust(R) devices in PA estimation as a low complexity and cost approach to allow 24-hour assessments of EE.

IntroductionWearable accelerometers are a valuable tool for monitoring sleep, sedentary behaviour, and physical activity patterns within 24h time-use in free-living environments. While wrist-worn accelerometers are favoured for monitoring sleep, they do not accurately distinguish between sitting and lying positions (Narayanan et al., 2020). This study aims to determine whether back or thigh-mounted accelerometers yield sleep metrics comparable to wrist-worn devices using an open-source algorithm originally validated for the wrist. MethodsData from 20 healthy sleepers were collected using Axivity AX3 accelerometers. Participants wore accelerometers on their right thigh, low-back, and wrist for one night of sleep in their own bed. Sleep metrics were calculated using the van Hees algorithm through the GGIR package in R. The primary outcomes were: Total Sleep Time (TST), Wake After Sleep Onset (WASO), Awakenings (AWK), Sleep Efficiency (SE), Sleep Interval (SI) and Sleep Onset Timestamp (SOT). Within-subject ANOVA with Tukeys post hoc, Pearson correlation coefficients, Bland-Altman plots, and Cohens d were used to assess the comparability of sleep metrics between the body placements. ResultsData analysis included all 20 participants. Mid-thigh accelerometers demonstrated a strong linear relationship with wrist accelerometers across all metrics (r = 0.86-0.98). Bland-Altman plots demonstrated a narrow 95% confidence interval suggesting that wrist and mid-thigh metrics are in good agreement, except for AWK which is slightly underestimated by the mid-thigh device. Conversely, low-back accelerometers demonstrated moderate linear relationship with the wrist (r = 0.63-0.98) and the Bland-Altman results showed wide limits of agreement with significant overestimations of TST, SE, SI and underestimations of WASO, AWK, SOT. Cohens d demonstrated small differences between mid-thigh and wrist devices, except for AWK (d= 0.42). Low-back values for WASO, SE, and AWK showed moderate differences. ConclusionsThis analysis demonstrates that the mid-thigh accelerometer yields comparable sleep metrics to wrist-worn devices when processed with the van Hees algorithm.

AO_SCPLOWBSTRACTC_SCPLOWO_ST_ABSBackgroundC_ST_ABSSleep is essential for overall health and well-being, but assessing sleep architecture and quality is often costly and time-consuming, relying primarily on polysomnography (PSG). Wearable and nearable devices offer potential alternatives, but they regularly lack rigorous validation, especially in real-world settings. ObjectivesThis study evaluates the accuracy and reliability of Withings Sleep Analyzer (WSA), a contactless sleep monitoring device, compared to PSG in a home setting using a large and diverse cohort of healthy individuals. MethodsA total of 117 participants (69 women; 39.9 {+/-} 11.4 years, mean {+/-} std) underwent home-based polysomnography (PSG) and simultaneous WSA recording. Data analysis focused on evaluating sleep-wake distinction and sleep stage identification using standard classification metrics. ResultsWSA demonstrates high sensitivity (93%) for sleep detection and moderate sensitivity (73%) for wakefulness, achieving an overall accuracy of 87% for sleep-wake distinction. The device showed consistent performance across various demographic subgroups, including different age, BMI, mattress and sleep arrangements (with or without bed partner) categories. Challenges were noted in accurately classifying specific sleep stages, particularly in distinguishing between light and deep sleep, with a mean accuracy of 63% and a Cohens Kappa of 0.49. The WSA tended to overestimate total sleep time (+20 min) and light sleep (+1h21 min) while underestimating REM (-15 min) and deep sleep (-46 min) durations. Disagreements between expert reviewers, particularly between light and deep sleep stages, mirrored in part the WSAs misclassifications. Participants reported significantly altered perceived sleep quality during the night with the PSG, suggesting potential discomfort during sleep. ConclusionsWSA offers a promising approach to sleep monitoring in natural home environments. Being contactless and placed under the mattress, the WSAs allows for long-term monitoring of sleep measures. It shows competitive performance in sleep-wake and sleep stage identification compared to other consumer devices. Progress in wearable and nearable devices is necessary to enhance their accuracy to better support the monitoring of populations with strongly impaired sleep, although limited by an imperfect gold standard. This work also emphasizes the importance of using large, diverse, and challenging datasets, as well as the need for a standardized methodology for accurate sleep stage classification.

The intensity gradient (IG) quantifies the distribution of time spent across accelerometer-assessed physical activity intensity and is positively associated with health. It was developed using the Euclidean Norm Minus One (ENMO) intensity metric. This study aimed to enable generation of comparable IGs across other metrics (mean amplitude deviation (MAD), monitor independent movement summary units (MIMS), and counts), by addressing a key step in the IG algorithm of dividing physical activity intensity into incremental intensity bins. Two methods of creating analogous bins for MAD, MIMS and counts were explored: 1) linear scaling ("naive"); 2) non-linear modelling ("modelled"). Generated IGs were compared to the original IG (IG_ENMO) using limits of agreement (LoA) and intra-class correlation (ICC). 43 adults (age, median [IQR]: 23 (21, 26), 61% female) were included. Relative to IG_ENMO, the modelled approach led to lower IGs (bias: -0.43, -1.23, -0.91 for MAD, MIMS, and counts, respectively). In contrast, the naive approach led to higher IGs (+0.27, +0.39, +0.54, respectively). For MAD and counts, LoA were slightly wider for naive bins (95% LoA: {+/-}0.26, {+/-}0.34) vs modelled bins ({+/-}0.21, {+/-}0.28), but for MIMS were slightly wider for modelled bins (modelled: {+/-}0.35, naive: {+/-}0.31). ICCs were higher for the modelled approach with IG_MAD most consistent (ICC 95% confidence interval: 0.72-0.91) and IG_MIMS least consistent (0.59-0.86). For the naive approach, IG_MAD was most consistent (0.49-0.82) and IG_counts least consistent (0.09-0.61). Results indicate that consistency of the IG between metrics is improved with appropriate scaling to create analogous intensity bins, but agreement is limited.

BackgroundMany adults in high-income countries carry a device capable of measuring physical- activity behaviour. Thus, there is public health need to understand whether such data can enhance prediction of future health outcomes. We aimed to investigate whether device-measured daily-step count and walking cadence improve the prediction of mortality beyond traditional risk-factors. MethodsRisk models were developed to predict five-year all-cause mortality using data from the UK Biobank accelerometer sub-study, with external validation in the US 2011-2014 National Health and Nutrition Examination Survey (NHANES). Median daily-step count and peak one-minute walking cadence were derived using self-supervised machine learning models from seven-day wrist-worn accelerometer data. Cox models were used to develop a baseline model incorporating traditional risk- factors, and a baseline model plus accelerometer data (i.e. daily-steps and walking cadence). Changes in model performance were assessed using Harrells C-index, net reclassification index (NRI; 10% threshold), and the Nam-DAgostino calibration test. FindingsAmong 79,717 UK Biobank participants, 1,640 died within 5-years. Adding accelerometer data to the baseline model modestly improved risk discrimination and classification with a change in c- index of 0.008 (95% confidence interval [CI] 0.005-0.011) and 3.3% NRI (95%CI 2.1%-4.5%). Greatest improvements in prediction were observed in participants with prior disease at baseline, showing a change in c-index of 0.028 (95%CI 0.019-0.039) and 5.9% NRI (95%CI 3.1%-8.6%). In the NHANES external validation cohort (n=4,713; deaths=378), similar improvements in prediction were observed (change in c-index: 0.015, 95%CI 0.007-0.025; NRI: 4.0%, 95%CI 0.7%-7.4%). All models were well calibrated (Nam-DAgostino {chi}2 range: 6.8-13.2). InterpretationDevice-measured daily-step count and walking cadence consistently demonstrated modest improvements in predicting mortality risk beyond traditional risk-factors, with the most significant enhancements seen in individuals with prior disease. These findings suggest that incorporating information from wearables does provide important new ways to improve risk stratification for targeted intervention in high-risk individuals.

ObjectivesTo describe physical behaviours following hospital admission for COVID-19 including associations with acute illness severity and ongoing symptoms. Methods1077 patients with COVID-19 discharged from hospital between March and November 2020 were recruited. Using a 14-day wear protocol, wrist-worn accelerometers were sent to participants after a five-month follow-up assessment. Acute illness severity was assessed by the WHO clinical progression scale, and the severity of ongoing symptoms was assessed using four previously reported data-driven clinical recovery clusters. Two existing control populations of office workers and type 2 diabetes were comparators. ResultsValid accelerometer data from 253 women and 462 men were included. Women engaged in a mean{+/-}SD of 14.9{+/-}14.7 minutes/day of moderate-to-vigorous physical activity (MVPA), with 725.6{+/-}104.9 minutes/day spent inactive and 7.22{+/-}1.08 hours/day asleep. The values for men were 21.0{+/-}22.3 and 755.5{+/-}102.8 minutes/day and 6.94{+/-}1.14 hours/day, respectively. Over 60% of women and men did not have any days containing a 30-minute bout of MVPA. Variability in sleep timing was approximately 2 hours in men and women. More severe acute illness was associated with lower total activity and MVPA in recovery. The very severe recovery cluster was associated with fewer days/week containing continuous bouts of MVPA, longer sleep duration, and higher variability in sleep timing. Patients post-hospitalisation with COVID-19 had lower levels of physical activity, greater sleep variability, and lower sleep efficiency than a similarly aged cohort of office workers or those with type 2 diabetes. ConclusionsPhysical activity and regulating sleep patterns are potential treatable traits for COVID-19 recovery programmes.

Study objectives Consumer sleep technologies (CSTs) enable low-burden longitudinal sleep monitoring, and their output measures are often interpreted as equivalent to polysomnography (PSG) measures. We applied a measurement reliability-aware approach to determine whether CST-derived 'sleep' measures (1) are interchangeable or device-specific, (2) can reliably assess trait-like sleep characteristics of an individual, (3) can be reduced to latent principal components of sleep, and (4) can be used for classification and biomarker discovery. Methods Data from 74 older adults (20 people living with dementia [PLWD]) were collected at-home (upto 14 nights; Total=752nights) using four tools simultaneously: research-grade actigraphy (Axivity), a wearable (Withings Watch), a nearable (Withings Sleep Analyzer) and Sleep Diary, followed by one in-lab PSG assessment. We used repeated-measures correlation analyses, intraclass correlation coefficients (ICC), principal component analysis (PCA) and binary classification models to address our objectives. Results Single-night between-device correlations and correlations with PSG were moderate (0.3[&le;]r<0.7) for some duration- and timing-related measures, but other associations were weak (r<0.3). Seventy-one percent of sleep measures reached acceptable reliability (ICC[&ge;]0.7) within seven nights of aggregation, but the required aggregation window varied across measures, tools and between PLWD and Controls. Reliability-filtered PCA yielded stable and interpretable principal components, but Duration was the only component showing moderate between-device association. Principal components were successfully used to classify PLWD vs Controls but feature importance varied across devices. Conclusions Aggregation of CST derived measures across 7-14 nights, yielded reliable measures, most of which were device-specific, with duration being the only essential aspect transferable between devices.

Wearable devices can collect changes in human behaviors related to mental health including depression and anxiety. Here, we examined whether and how digital metrics from a consumer-grade wearable smart ring (Oura Ring) differed by severity of depression and anxiety symptoms using data from a large-scale population-based sample of young adults (n=1,290, age range: 33-35). Participants wore the ring for two weeks, assessing sleep architecture, nocturnal heart rate (HR), heart rate variability (HRV), and movement intensity. Mental health symptoms were assessed using the Generalized Anxiety Disorder 7-item and Hopkins Symptom Checklist-25 scales. On average, participants with higher depression and/or anxiety symptoms had lower levels of rapid eye movement and had higher levels of deep and light sleep, elevated nocturnal HR, reduced HRV, and lower daytime movement compared to non-symptom individuals. Findings suggest that symptoms of depression and anxiety may manifest in physiological and behavioral metrics collected by consumer-grade wearable devices.

QuestionWhat is the current evidence base for the association between digital biomarkers from wrist-worn wearables, loneliness and social isolation in adults? Study selection and analysisWe systematically searched six databases from inception to 24th September, 2024. We narratively synthesised findings and pooled effect sizes using random-effects meta-analyses where possible. FindingsWe included 14 studies from 12 articles (12 assessing loneliness, two assessing social isolation). Eight studies used sleep metrics, four used physical activity metrics, and two studies used machine learning approaches. Three meta-analyses were conducted: worse sleep efficiency (SE), but not total sleep time or sleep onset latency, was associated with higher loneliness (Fishers z = -0.20, 95% CI -0.34 to -0.06, p = 0.006). Two studies examined wake after sleep onset (WASO), and found longer periods of WASO were associated with higher loneliness. These findings on loneliness were echoed in the study examining social isolation. One study found that lower total physical activity was associated with higher levels of loneliness and social isolation, while other activity intensities showed mixed evidence. Machine learning studies demonstrated high accuracy in predicting loneliness, though models using digital biomarkers from smartphones provided better accuracy. ConclusionsWorse SE, more WASO, and lower total physical activity were associated with loneliness and social isolation, particularly in middle- and older-age. Digital biomarker-based machine learning studies are sparse but show potential in predicting loneliness. Leveraging digital biomarkers as proxy markers of loneliness and social isolation could facilitate early detection of these conditions. Key messages of the articlesO_ST_ABSWhat is already known on this topicC_ST_ABSLoneliness and social isolation are linked to negative health outcomes, including increased dementia risk. Digital biomarkers have shown potential in detecting mental health conditions and symptoms, but no systematic review has explored their association with loneliness and social isolation. What this study addsThis review identified 14 studies examining the association between digital biomarkers and loneliness or social isolation. Worse sleep efficiency, more wake after sleep onset, and lower physical activity were associated with higher levels of loneliness and social isolation. How this study might affect research, practice or policyDigital biomarkers could support the early detection of loneliness and social isolation, enabling timely intervention.

BackgroundAccurately detecting the Sleep Period Time (SPT) window in the daily life is essential for understanding habitual sleep and health. Although actigraphy devices (accelerometers) placement varies across studies, most SPT-detection algorithms are developed for wrist data. Open-source algorithms support reproducibility and transparency in estimating the SPT. AimsTo optimise and evaluate two open-source algorithms, HDCZA and HorAngle, for estimating the SPT window using hip-worn accelerometer data. MethodsA total of 109 children and 194 adults wore wrist and hip accelerometers for six nights and completed sleep diaries. An established algorithm combining wrist and diary data served as the reference. HDCZA and HorAngle parameters were optimised using Bayesian optimisation on 60% of the sample and evaluated in the remaining 40%. ResultsMean differences for sleep onset and wake-up were -3 and 4 minutes for HDCZA (limits of agreement [LoA]: -221,215 and -185,194; root-mean square error [RMSE]=111 and 97) and 0 and -4 minutes for HorAngle (LoA: -199,199 and -223,214; RMSE=111 and 112). For SPT duration, mean differences were 7 minutes (LoA: -252,266; RMSE=132) for HDCZA and -4 minutes (LoA: -254,246; RMSE=128). No significant differences in SPT duration were found (P=0.774; P=0.237). Both algorithms showed moderate agreement with the reference in ranking sleep duration ({kappa} {approx} 0.56-0.58). Differences were unrelated to age or sex but linked to non-wear time. ConclusionsBoth open-source algorithms demonstrated value for estimating the SPT window from hip data. While HDCZA requires no additional sensor-specific parameters, HorAngle depends on accurate axis identification. Statement of SignificanceAccurately estimating the sleep period time (SPT) window from hip-worn accelerometers is essential for studies assessing sleep in free-living conditions. However, most available algorithms were developed for wrist-worn data. This study optimised and validated two open-source algorithms, HDCZA and HorAngle, for hip-worn accelerometer data in children and adults. Both algorithms performed comparably to a wrist-based reference using sleep diaries, showing consistent agreement across age and sex. These methods enable researchers to estimate habitual sleep without additional sensors or diaries, improving reproducibility and scalability in observational research. The algorithms are openly implemented in the GGIR R package, offering accessible and standardised tools for analysing hip-based accelerometer data.

Background/AimsPhysical inactivity, sedentary behaviour (SB) and inadequate sleep are key behavioural risk factors of cardiometabolic diseases; each is mainly considered in isolation. The study aim was to investigate associations of five movement behaviour compositions with adiposity and cardiometabolic biomarkers. MethodsCross-sectional data from 15,246 participants from the Prospective Physical Activity, Sitting and Sleep consortium (ProPASS) were analysed. Time spent in sleep, SB, standing, light-intensity physical activity (LIPA) and moderate-vigorous physical activity (MVPA) made up the composition. Outcomes included BMI, waist circumference, HDL cholesterol, total:HDL cholesterol ratio, triglycerides and HbA1c. Compositional linear regression examined associations between compositions and each outcome, including modelling reallocation of time between behaviours. ResultsThe average daily composition of the sample(age:53.7{+/-}9.7years; 54.7%female) was 7.7hrs sleeping,10.4hrs sedentary,3.1hrs standing,1.5hrs LIPA and 1.3hrs MVPA. A greater proportion of MVPA time and smaller proportion of SB time was associated with better outcomes. Reallocating time from SB,standing,LIPA or sleep into MVPA had the largest theoretical improvement across all outcomes. For example, replacing 30min of SB, sleep, standing or LIPA with MVPA was associated with -0.63 (95%CI -0.48,-0.79), -0.43 (-0.25,-0.59), -0.40 (-0.25,-0.56) and -0.15 (0.05,-0.34)kg/m2 lower BMI, respectively. A larger proportion of standing time was beneficial for outcomes; sleep had a detrimental association when replacing LIPA or MVPA and positive association when replacing SB. The minimal displacement into MVPA for improved cardiometabolic health ranged from 3.8 (HbA1c) to 12.7 (triglycerides) min/day. ConclusionsCompositional data analyses revealed a distinct hierarchy of behaviours. MVPA demonstrated the strongest, most time-efficient protective associations with cardiometabolic outcomes. Theoretical benefits from reallocating SB into sleep, standing or LIPA required substantial changes in daily activity.

BackgroundStep counts are increasingly used in public health and clinical research to assess wellbeing, lifestyle, and health status. However, estimating step counts using commercial activity trackers has several limitations, including a lack of reproducibility, generalizability, and scalability. Smartphones are a potentially promising alternative, but their step-counting algorithms require robust validation that accounts for temporal sensor body location, individual gait characteristics, and heterogeneous health states. ObjectiveOur goal was to evaluate an open-source step-counting method for smartphones under various measurement conditions against step counts estimated from data collected simultaneously from different body locations ("internal" validation), manually ascertained ground truth ("manual" validation), and step counts from a commercial activity tracker (Fitbit Charge 2) in patients with advanced cancer ("wearable" validation). MethodsWe used eight independent datasets collected in controlled, semi-controlled, and free-living environments with different devices (primarily Android smartphones and wearable accelerometers) carried at typical body locations. Five datasets (N=103) were used for internal validation, two datasets (N=107) for manual validation, and one dataset (N=45) used for wearable validation. In each scenario, step counts were estimated using a previously published step-counting method for smartphones that uses raw sub-second level accelerometer data. We calculated mean bias and limits of agreement (LoA) between step count estimates and validation criteria using Bland-Altman analysis. ResultsIn the internal validation datasets, participants performed 751.7{+/-}581.2 (mean{+/-}SD) steps, and the mean bias was -7.2 steps (LoA -47.6, 33.3) or -0.5%. In the manual validation datasets, the ground truth step count was 367.4{+/-}359.4 steps while the mean bias was -0.4 steps (LoA -75.2, 74.3) or 0.1 %. In the wearable validation dataset, Fitbit devices indicated mean step counts of 1931.2{+/-}2338.4, while the calculated bias was equal to -67.1 steps (LoA -603.8, 469.7) or a difference of 0.3 %. ConclusionsThis study demonstrates that our open-source step counting method for smartphone data provides reliable step counts across sensor locations, measurement scenarios, and populations, including healthy adults and patients with cancer.

IntroductionWearables with proprietary scoring protocols are typically used to assess sleep and physical activity, but the field is shifting to wearables with raw sensor data accessible and open-source scoring methods to enhance rigor and reproducibility. The data infrastructure to process raw sensor data in clinical research is underdeveloped; we therefore developed the Modular Actigraphy Platform (MAP). MethodsMAP is a cloud-based computational platform that processes high-resolution time series sensor data to derive sleep and physical activity metrics. It was engineered to be modular, providing flexibility in data processing and enabling the seamless integration of open-source sleep and physical activity scoring methodologies as they become available (currently, GGIR and MIMS processing algorithms have been integrated). A structured Software Development Life Cycle (SDLC) approach was used to guide the development of MAP, with a multi-level testing framework consisting of unit testing (verification of modules), integration testing (interaction among modules), and system testing (validating specifications). Following these foundational tests, we then completed user acceptance testing in two phases - alpha (17 test files) and beta (686 files from 4 pediatric cohorts) to assess processing performance. ResultsFor beta testing, MAP leveraged up to 60 CPU cores and 500 GiB of memory. The pre-processing module was the most computationally demanding and was more efficient in MAP compared to offline processing (up to 8 CPU cores and 23.2 GiB of memory). For example, the preprocessing GGIR part 1 container was completed at a speed of 0.29-0.49 minutes per file (1.6-2.9 times faster than offline processing) and the pre-processing MIMS container was completed at a speed of 0.49-4.66 minutes per file (2.4 to 14.0 times faster than offline processing). ConclusionMAP is an efficient computational platform that integrates open-source scoring algorithms to efficiently process raw sensor data for wearable sleep and physical activity estimation in clinical research and is available through the Childrens Hospital of Philadelphias Research Institute.