Self-Supervised Machine Learning to Characterise Step Counts from Wrist-Worn Accelerometers in the UK Biobank.

PURPOSE: Step count is an intuitive measure of physical activity frequently quantified in health-related studies; however, accurate step counting is difficult in the free-living environment, with error routinely above 20% in wrist-worn devices against camera-annotated ground truth. This study aims to describe the development and validation of step count derived from a wrist-worn accelerometer and assess its association with cardiovascular and all-cause mortality in a large prospective cohort. METHODS: We developed and externally validated a self-supervised machine learning step detection model, trained on an open-source and step-annotated free-living dataset. 39 individuals will free-living ground-truth annotated step counts were used for model development. An open-source dataset with 30 individuals was used for external validation. Epidemiological analysis was performed using 75,263 UK Biobank participants without prevalent cardiovascular disease (CVD) or cancer. Cox regression was used to test the association of daily step count with fatal CVD and all-cause mortality after adjustment for potential confounders. RESULTS: The algorithm substantially outperformed reference models (free-living mean absolute percent error of 12.5%, versus 65-231%). Our data indicate an inverse dose-response association, where taking 6,430-8,277 daily steps was associated with 37% [25-48%] and 28% [20-35%] lower risk of fatal CVD and all-cause mortality up to seven years later, compared to those taking fewer steps each day. CONCLUSIONS: We have developed an open and transparent method that markedly improves the measurement of steps in large-scale wrist-worn accelerometer datasets. The application of this method demonstrated expected associations with CVD and all-cause mortality, indicating excellent face validity. This reinforces public health messaging for increasing physical activity and can help lay the groundwork for the inclusion of target step counts in future public health guidelines.

Self-supervised learning of accelerometer data provides new insights for sleep and its association with mortality.

Sleep is essential to life. Accurate measurement and classification of sleep/wake and sleep stages is important in clinical studies for sleep disorder diagnoses and in the interpretation of data from consumer devices for monitoring physical and mental well-being. Existing non-polysomnography sleep classification techniques mainly rely on heuristic methods developed in relatively small cohorts. Thus, we aimed to establish the accuracy of wrist-worn accelerometers for sleep stage classification and subsequently describe the association between sleep duration and efficiency (proportion of total time asleep when in bed) with mortality outcomes. We developed a self-supervised deep neural network for sleep stage classification using concurrent laboratory-based polysomnography and accelerometry. After exclusion, 1448 participant nights of data were used for training. The difference between polysomnography and the model classifications on the external validation was 34.7 min (95% limits of agreement (LoA): -37.8-107.2 min) for total sleep duration, 2.6 min for REM duration (95% LoA: -68.4-73.4 min) and 32.1 min (95% LoA: -54.4-118.5 min) for NREM duration. The sleep classifier was deployed in the UK Biobank with 100,000 participants to study the association of sleep duration and sleep efficiency with all-cause mortality. Among 66,214 UK Biobank participants, 1642 mortality events were observed. Short sleepers (<6 h) had a higher risk of mortality compared to participants with normal sleep duration of 6-7.9 h, regardless of whether they had low sleep efficiency (Hazard ratios (HRs): 1.58; 95% confidence intervals (CIs): 1.19-2.11) or high sleep efficiency (HRs: 1.45; 95% CIs: 1.16-1.81). Deep-learning-based sleep classification using accelerometers has a fair to moderate agreement with polysomnography. Our findings suggest that having short overnight sleep confers mortality risk irrespective of sleep continuity.

Feasibility and usability of remote monitoring in Alzheimer's disease.

Introduction: Remote monitoring technologies (RMTs) can measure cognitive and functional decline objectively at-home, and offer opportunities to measure passively and continuously, possibly improving sensitivity and reducing participant burden in clinical trials. However, there is skepticism that age and cognitive or functional impairment may render participants unable or unwilling to comply with complex RMT protocols. We therefore assessed the feasibility and usability of a complex RMT protocol in all syndromic stages of Alzheimer's disease and in healthy control participants. Methods: For 8 weeks, participants (N = 229) used two activity trackers, two interactive apps with either daily or weekly cognitive tasks, and optionally a wearable camera. A subset of participants participated in a 4-week sub-study (N = 45) using fixed at-home sensors, a wearable EEG sleep headband and a driving performance device. Feasibility was assessed by evaluating compliance and drop-out rates. Usability was assessed by problem rates (e.g., understanding instructions, discomfort, forgetting to use the RMT or technical problems) as discussed during bi-weekly semi-structured interviews. Results: Most problems were found for the active apps and EEG sleep headband. Problem rates increased and compliance rates decreased with disease severity, but the study remained feasible. Conclusions: This study shows that a highly complex RMT protocol is feasible, even in a mild-to-moderate AD population, encouraging other researchers to use RMTs in their study designs. We recommend evaluating the design of individual devices carefully before finalizing study protocols, considering RMTs which allow for real-time compliance monitoring, and engaging the partners of study participants in the research.

Self-supervised learning for human activity recognition using 700,000 person-days of wearable data.

Accurate physical activity monitoring is essential to understand the impact of physical activity on one's physical health and overall well-being. However, advances in human activity recognition algorithms have been constrained by the limited availability of large labelled datasets. This study aims to leverage recent advances in self-supervised learning to exploit the large-scale UK Biobank accelerometer dataset-a 700,000 person-days unlabelled dataset-in order to build models with vastly improved generalisability and accuracy. Our resulting models consistently outperform strong baselines across eight benchmark datasets, with an F1 relative improvement of 2.5-130.9% (median 24.4%). More importantly, in contrast to previous reports, our results generalise across external datasets, cohorts, living environments, and sensor devices. Our open-sourced pre-trained models will be valuable in domains with limited labelled data or where good sampling coverage (across devices, populations, and activities) is hard to achieve.

A systematic review of the performance of actigraphy in measuring sleep stages.

The accuracy of actigraphy for sleep staging is assumed to be poor, but examination is limited. This systematic review aimed to assess the performance of actigraphy in sleep stage classification of adults. A systematic search was performed using MEDLINE, Web of Science, Google Scholar, and Embase databases. We identified eight studies that compared sleep architecture estimates between wrist-worn actigraphy and polysomnography. Large heterogeneity was found with respect to how sleep stages were grouped, and the choice of metrics used to evaluate performance. Quantitative synthesis was not possible, so we performed a narrative synthesis of the literature. From the limited number of studies, we found that actigraphy-based sleep staging had some ability to classify different sleep stages compared with polysomnography.

Digital health technologies and machine learning augment patient reported outcomes to remotely characterise rheumatoid arthritis.

Digital measures of health status captured during daily life could greatly augment current in-clinic assessments for rheumatoid arthritis (RA), to enable better assessment of disease progression and impact. This work presents results from weaRAble-PRO, a 14-day observational study, which aimed to investigate how digital health technologies (DHT), such as smartphones and wearables, could augment patient reported outcomes (PRO) to determine RA status and severity in a study of 30 moderate-to-severe RA patients, compared to 30 matched healthy controls (HC). Sensor-based measures of health status, mobility, dexterity, fatigue, and other RA specific symptoms were extracted from daily iPhone guided tests (GT), as well as actigraphy and heart rate sensor data, which was passively recorded from patients' Apple smartwatch continuously over the study duration. We subsequently developed a machine learning (ML) framework to distinguish RA status and to estimate RA severity. It was found that daily wearable sensor-outcomes robustly distinguished RA from HC participants (F1, 0.807). Furthermore, by day 7 of the study (half-way), a sufficient volume of data had been collected to reliably capture the characteristics of RA participants. In addition, we observed that the detection of RA severity levels could be improved by augmenting standard patient reported outcomes with sensor-based features (F1, 0.833) in comparison to using PRO assessments alone (F1, 0.759), and that the combination of modalities could reliability measure continuous RA severity, as determined by the clinician-assessed RAPID-3 score at baseline (r2, 0.692; RMSE, 1.33). The ability to measure the impact of the disease during daily life-through objective and remote digital outcomes-paves the way forward to enable the development of more patient-centric and personalised measurements for use in RA clinical trials.

Device-Measured Physical Activity in 3506 Individuals with Knee or Hip Arthroplasty.

PURPOSE: Hip and knee arthroplasty aims to reduce joint pain and increase functional mobility in patients with osteoarthritis; however, the degree to which arthroplasty is associated with higher physical activity is unclear. The current study sought to assess the association of hip and knee arthroplasty with objectively measured physical activity. METHODS: This cross-sectional study analyzed wrist-worn accelerometer data collected in 2013-2016 from UK Biobank participants (aged 43-78 yr). Multivariable linear regression was performed to assess step count, cadence, overall acceleration, and activity behaviors between nonarthritic controls, end-stage arthritic, and postoperative cohorts, controlling for demographic and behavioral confounders. From a cohort of 94,707 participants with valid accelerometer wear time and complete self-reported data, electronic health records were used to identify 3506 participants having undergone primary or revision hip or knee arthroplasty and 68,389 nonarthritic controls. RESULTS: End-stage hip or knee arthritis was associated with taking 1129 fewer steps per day (95% confidence interval (CI), 811-1447; P < 0.001) and having 5.8 fewer minutes per day (95% CI, 3.0-8.7; P < 0.001) of moderate-to-vigorous activity compared with nonarthritic controls. Unilateral primary hip and knee arthroplasties were associated with 877 (95% CI, 284-1471; P = 0.004) and 893 (95% CI, 232-1554; P = 0.008) more steps than end-stage osteoarthritic participants, respectively. Postoperative unilateral hip arthroplasty participants demonstrated levels of moderate-to-vigorous physical activity and daily step count equivalent to nonarthritic controls. No difference in physical activity was observed between any cohorts in terms of overall acceleration, or time spent in daily light activity, sedentary behavior, or sleep. CONCLUSIONS: Hip and knee arthroplasties are associated with higher levels of physical activity compared with participants with end-stage arthritis. Unilateral hip arthroplasty patients, in particular, demonstrate equivalence to nonarthritic peers at more than 1 yr after surgery.

Device-measured movement behaviours in over 20,000 China Kadoorie Biobank participants.

BACKGROUND: Movement behaviours, including physical activity, sedentary behaviour, and sleep have been shown to be associated with several chronic diseases. However, they have not been objectively measured in large-scale prospective cohort studies in low-and middle-income countries. We aim to describe the patterns of device-measured movement behaviours collected in the China Kadoorie Biobank (CKB) study. METHODS: During 2020 and 2021, a random subset of 25,087 surviving CKB individuals participated in the 3rd resurvey of the CKB. Among them, 22,511 (89.7%) agreed to wear an Axivity AX3 wrist-worn triaxial accelerometer for seven consecutive days to assess their habitual movement behaviours. We developed a machine-learning model to infer time spent in four movement behaviours [i.e. sleep, sedentary behaviour, light intensity physical activity (LIPA), and moderate-to-vigorous physical activity (MVPA)]. Descriptive analyses were performed for wear-time compliance and patterns of movement behaviours by different participant characteristics. RESULTS: Data from 21,897 participants (aged 65.4 ± 9.1 years; 35.4% men) were received for demographic and wear-time analysis, with a median wear-time of 6.9 days (IQR: 6.1-7.0). Among them, 20,370 eligible participants were included in movement behavior analyses. On average, they had 31.1 mg/day (total acceleration) overall activity level, accumulated 7.7 h/day (32.3%) of sleep time, 8.8 h/day (36.6%) sedentary, 5.7 h/day (23.9%) in light physical activity, and 104.4 min/day (7.2%) in moderate-to-vigorous physical activity. There was an inverse relationship between age and overall acceleration with an observed decline of 5.4 mg/day (17.4%) per additional decade. Women showed a higher activity level than men (32.3 vs 28.8 mg/day) and there was a marked geographical disparity in the overall activity level and time allocation. CONCLUSIONS: This is the first large-scale accelerometer data collected among Chinese adults, which provides rich and comprehensive information about device-measured movement behaviour patterns. This resource will enhance our knowledge about the potential relevance of different movement behaviours for chronic disease in Chinese adults.

Dose-response of accelerometer-measured physical activity, step count, and cancer risk in the UK Biobank: a prospective cohort analysis.

BACKGROUND: Cancer is an age-related condition, but changes to modifiable lifestyle-related behaviours, including physical activity, could impact risk. While step count is an accessible metric of activity for older adults, its association with cancer risk remains poorly understood. We investigated the association between accelerometer-measured total activity, step count, and cancer risk. METHODS: We analysed data from a prospective UK Biobank cohort of consenting participants who wore wrist-based Axivity AX3 accelerometer devices for 7 days between June 1, 2013 and Dec 23, 2015, had valid accelerometer data, and no previous cancer diagnosis at baseline. Machine learning models estimated total physical activity (vector magnitude) and step count. The primary outcome, a composite of 13 cancers previously associated with physical activity, was obtained from national registries. Hazard ratios (HR) and were calculated using Cox proportional hazard models, with attained age as the underlying timescale and adjustment for sex, ethnicity, smoking status, alcohol consumption, education, and Townsend Deprivation Index. The impact of reallocating time between behaviours was evaluated using compositional data analyses. Dose-response associations were assessed with restricted cubic splines. FINDINGS: We analysed data from 86 556 participants, who were followed up during an average of 6·1 years (age range 43-78; 48 478 [56%] female and 38 078 [44%] male; 83 830 [97%] white). 5577 incident malignant cancers occurred among these 86 556 participants. Greater total physical activity was associated with a lower risk of physical-activity-related cancer (HR per 1 SD [+8·33 milligravity per day] 0·85, 95% CI 0·81-0·89). Reallocating 30 min/day from other activities to moderate-to-vigorous physical activity behaviour was associated with lower cancer risk (HR 0·96, 0·94-0·98), as was reallocating 1 h/day to light intensity activity (HR 0·94, 0·92-0·96), compared with the mean behaviour composition among included participants. Compared with taking 5000 steps per day, taking 10 000 daily steps was associated with a significantly lower risk of physical-activity-related cancer (HR 0·81, 0·73-0·90). INTERPRETATION: In this sample from the UK Biobank, higher total physical activity and daily step count were associated with lower risk of physical-activity-related cancers. Findings suggest additional physical activity time, irrespective of intensity, may be beneficial. Increasing low intensity activity time and increasing daily step counts could be practical public health interventions to lower cancer risk, especially for aging adults. FUNDING: National Institute of Health Oxford Cambridge Scholars Program, Wellcome Trust, Swiss Re, Health Data Research UK, and Cancer Research UK.

Prognostic Significance of Different Ventricular Ectopic Burdens During Submaximal Exercise in Asymptomatic UK Biobank Subjects.

BACKGROUND: The consequences of exercise-induced premature ventricular contractions (PVCs) in asymptomatic individuals remain unclear. This study aimed to assess the association between PVC burdens during submaximal exercise and major adverse cardiovascular events (MI/HF/LTVA: myocardial infarction [MI], heart failure [HF], and life-threatening ventricular arrhythmia [LTVA]), and all-cause mortality. Additional end points were MI, LTVA, HF, and cardiovascular mortality. METHODS: A neural network was developed to count PVCs from ECGs recorded during exercise (6 minutes) and recovery (1 minute) in 48 315 asymptomatic participants from UK Biobank. Associations were estimated using multivariable Cox proportional hazard models. Explorative studies were conducted in subgroups with cardiovascular magnetic resonance imaging data (n=6290) and NT-proBNP (N-terminal Pro-B-type natriuretic peptide) levels (n=4607) to examine whether PVC burden was associated with subclinical cardiomyopathy. RESULTS: Mean age was 56.8±8.2 years; 51.1% of the participants were female; and median follow-up was 12.6 years. Low PVC counts during exercise and recovery were both associated with MI/HF/LTVA risk, independently of clinical factors: adjusted hazard ratio (HR), 1.2 (1-5 exercise PVCs, P<0.001) and HR, 1.3 (1-5 recovery PVCs, P<0.001). Risks were higher with increasing PVC count: HR, 1.8 (>20 exercise PVCs, P<0.001) and HR, 1.6 (>5 recovery PVCs, P<0.001). A similar trend was observed for all-cause mortality, although associations were only significant for high PVC burdens: HRs, 1.6 (>20 exercise PVCs, P<0.001) and 1.5 (>5 recovery PVCs, P<0.001). Complex PVC rhythms were associated with higher risk compared with PVC count alone. PVCs were also associated with incident HF, LTVA, and cardiovascular mortality, but not MI. In the explorative studies, high PVC burden was associated with larger left ventricular volumes, lower ejection fraction, and higher levels of NT-proBNP compared with participants without PVCs. CONCLUSIONS: In this cohort of middle-aged and older adults, PVC count during submaximal exercise and recovery were both associated with MI/HF/LTVA, all-cause mortality, HF, LTVAs, and cardiovascular mortality, independent of clinical and exercise test factors, indicating an incremental increase in risk as PVC count rises. Complex PVC rhythms were associated with higher risk compared with PVC count alone. Underlying mechanisms may include the presence of subclinical cardiomyopathy.

Brief bouts of device-measured intermittent lifestyle physical activity and its association with major adverse cardiovascular events and mortality in people who do not exercise: a prospective cohort study.

BACKGROUND: Guidelines emphasise the health benefits of bouts of physical activity of any duration. However, the associations of intermittent lifestyle physical activity accumulated through non-exercise with mortality and major adverse cardiovascular events (MACE) remain unclear. We aimed to examine the associations of bouts of moderate-to-vigorous intermittent lifestyle physical activity (MV-ILPA) and the proportion of vigorous activity contributing within these bouts with mortality and MACE. METHODS: In this prospective cohort study, we used data from the UK Biobank on adults who do not exercise (ie, those who did not report leisure-time exercise) who had wrist-worn accelerometry data available. Participants were followed up until Nov 30, 2022, with the outcome of interest of all-cause mortality obtained through linkage with NHS Digital of England and Wales, and the NHS Central Register and National Records of Scotland, and MACE obtained from inpatient hospitalisation data provided by the Hospital Episode Statistics for England, the Patient Episode Database for Wales, and the Scottish Morbidity Record for Scotland. MV-ILPA bouts were derived using a two-level Random Forest classifier and grouped as short (<1 min), medium (1 to <3 min; 3 to <5 min), and long (5 to <10 min). We further examined the dose-response relationship of the proportion of vigorous physical activity contributing to the MV-ILPA bout. FINDINGS: Between June 1, 2013, and Dec 23, 2015, 103 684 Biobank participants wore an accelerometer on their wrist. 25 241 adults (mean age 61·8 years [SD 7·6]), of whom 14 178 (56·2%) were women, were included in our analysis of all-cause mortality. During a mean follow-up duration of 7·9 years (SD 0·9), 824 MACE and 1111 deaths occurred. Compared with bouts of less than 1 min, mortality risk was lower for bouts of 1 min to less than 3 min (hazard ratio [HR] 0·66 [0·53-0·81]), 3 min to less than 5 min (HR 0·56 [0·46-0·69]), and 5 to less than 10 min (HR 0·48 [0·39-0·59]). Similarly, compared with bouts of less than 1 min, risk of MACE was lower for bouts of 1 min to less than 3 min (HR 0·71 [0·54-0·93]), 3 min to less than 5 min (0·62 [0·48-0·81]), and 5 min to less than 10 min (0·59 [0·46-0·76]). Short bouts (<1 min) were associated with lower MACE risk only when bouts were comprised of at least 15% vigorous activity. INTERPRETATION: Intermittent non-exercise physical activity was associated with lower mortality and MACE. Our results support the promotion of short intermittent bouts of non-exercise physical activity of moderate-to-vigorous intensity to improve longevity and cardiovascular health among adults who do not habitually exercise in their leisure time. FUNDING: Australian National Health, Medical Research Council, and Wellcome Trust.

Characterising personal, household, and community PM2.5 exposure in one urban and two rural communities in China.

BACKGROUND: Cooking and heating in households contribute importantly to air pollution exposure worldwide. However, there is insufficient investigation of measured fine particulate matter (PM2.5) exposure levels, variability, seasonality, and inter-spatial dynamics associated with these behaviours. METHODS: We undertook parallel measurements of personal, household (kitchen and living room), and community PM2.5 in summer (May-September 2017) and winter (November 2017-Janauary 2018) in 477 participants from one urban and two rural communities in China. After stringent data cleaning, there were 67,326-80,980 person-hours (ntotal = 441; nsummer = 384; nwinter = 364; 307 had repeated PM2.5 data in both seasons) of processed data per microenvironment. Age- and sex-adjusted geometric means of PM2.5 were calculated by key participant characteristics, overall and by season. Spearman correlation coefficients between PM2.5 levels across different microenvironments were computed. FINDINGS: Overall, 26.4 % reported use of solid fuel for both cooking and heating. Solid fuel users had 92 % higher personal and kitchen 24-h average PM2.5 exposure than clean fuel users. Similarly, they also had a greater increase (83 % vs 26 %) in personal and household PM2.5 from summer to winter, whereas community levels of PM2.5 were 2-4 times higher in winter across different fuel categories. Compared with clean fuel users, solid fuel users had markedly higher weighted annual average PM2.5 exposure at personal (78.2 [95 % CI 71.6-85.3] µg/m3 vs 41.6 [37.3-46.5] µg/m3), kitchen (102.4 [90.4-116.0] µg/m3 vs 52.3 [44.8-61.2] µg/m3) and living room (62.1 [57.3-67.3] µg/m3 vs 41.0 [37.1-45.3] µg/m3) microenvironments. There was a remarkable diurnal variability in PM2.5 exposure among the participants, with 5-min moving average from 10 µg/m3 to 700-1200 µg/m3 across different microenvironments. Personal PM2.5 was moderately correlated with living room (Spearman r: 0.64-0.66) and kitchen (0.52-0.59) levels, but only weakly correlated with community levels, especially in summer (0.15-0.34) and among solid fuel users (0.11-0.31). CONCLUSION: Solid fuel use for cooking and heating was associated with substantially higher personal and household PM2.5 exposure than clean fuel users. Household PM2.5 appeared a better proxy of personal exposure than community PM2.5.

Self-supervised learning of accelerometer data provides new insights for sleep and its association with mortality.

Background: Sleep is essential to life. Accurate measurement and classification of sleep/wake and sleep stages is important in clinical studies for sleep disorder diagnoses and in the interpretation of data from consumer devices for monitoring physical and mental well-being. Existing non-polysomnography sleep classification techniques mainly rely on heuristic methods developed in relatively small cohorts. Thus, we aimed to establish the accuracy of wrist-worn accelerometers for sleep stage classification and subsequently describe the association between sleep duration and efficiency (proportion of total time asleep when in bed) with mortality outcomes. Methods: We developed and validated a self-supervised deep neural network for sleep stage classification using concurrent laboratory-based polysomnography and accelerometry data from three countries (Australia, the UK, and the USA). The model was validated within-cohort using subject-wise five-fold cross-validation for sleep-wake classification and in a three-class setting for sleep stage classification wake, rapid-eye-movement sleep (REM), non-rapid-eye-movement sleep (NREM) and by external validation. We assessed the face validity of our model for population inference by applying the model to the UK Biobank with 100,000 participants, each of whom wore a wristband for up to seven days. The derived sleep parameters were used in a Cox regression model to study the association of sleep duration and sleep efficiency with all-cause mortality. Findings: After exclusion, 1,448 participant nights of data were used to train the sleep classifier. The difference between polysomnography and the model classifications on the external validation was 34.7 minutes (95% limits of agreement (LoA): -37.8 to 107.2 minutes) for total sleep duration, 2.6 minutes for REM duration (95% LoA: -68.4 to 73.4 minutes) and 32.1 minutes (95% LoA: -54.4 to 118.5 minutes) for NREM duration. The derived sleep architecture estimate in the UK Biobank sample showed good face validity. Among 66,214 UK Biobank participants, 1,642 mortality events were observed. Short sleepers (<6 hours) had a higher risk of mortality compared to participants with normal sleep duration (6 to 7.9 hours), regardless of whether they had low sleep efficiency (Hazard ratios (HRs): 1.69; 95% confidence intervals (CIs): 1.28 to 2.24 ) or high sleep efficiency (HRs: 1.42; 95% CIs: 1.14 to 1.77). Interpretation: Deep-learning-based sleep classification using accelerometers has a fair to moderate agreement with polysomnography. Our findings suggest that having short overnight sleep confers mortality risk irrespective of sleep continuity.

Development and Validation of a Machine Learning Wrist-worn Step Detection Algorithm with Deployment in the UK Biobank.

Background: Step count is an intuitive measure of physical activity frequently quantified in a range of health-related studies; however, accurate quantification of step count can be difficult in the free-living environment, with step counting error routinely above 20% in both consumer and research-grade wrist-worn devices. This study aims to describe the development and validation of step count derived from a wrist-worn accelerometer and to assess its association with cardiovascular and all-cause mortality in a large prospective cohort study. Methods: We developed and externally validated a hybrid step detection model that involves self-supervised machine learning, trained on a new ground truth annotated, free-living step count dataset (OxWalk, n=39, aged 19-81) and tested against other open-source step counting algorithms. This model was applied to ascertain daily step counts from raw wrist-worn accelerometer data of 75,493 UK Biobank participants without a prior history of cardiovascular disease (CVD) or cancer. Cox regression was used to obtain hazard ratios and 95% confidence intervals for the association of daily step count with fatal CVD and all-cause mortality after adjustment for potential confounders. Findings: The novel step algorithm demonstrated a mean absolute percent error of 12.5% in free-living validation, detecting 98.7% of true steps and substantially outperforming other recent wrist-worn, open-source algorithms. Our data are indicative of an inverse dose-response association, where, for example, taking 6,596 to 8,474 steps per day was associated with a 39% [24-52%] and 27% [16-36%] lower risk of fatal CVD and all-cause mortality, respectively, compared to those taking fewer steps each day. Interpretation: An accurate measure of step count was ascertained using a machine learning pipeline that demonstrates state-of-the-art accuracy in internal and external validation. The expected associations with CVD and all-cause mortality indicate excellent face validity. This algorithm can be used widely for other studies that have utilised wrist-worn accelerometers and an open-source pipeline is provided to facilitate implementation.

The impact of selected methodological factors on data collection outcomes in observational studies of device-measured physical behaviour in adults: A systematic review.

BACKGROUND: Accelerometer measures of physical behaviours (physical activity, sedentary behaviour and sleep) in observational studies offer detailed insight into associations with health and disease. Maximising recruitment and accelerometer wear, and minimising data loss remain key challenges. How varying methods used to collect accelerometer data influence data collection outcomes is poorly understood. We examined the influence of accelerometer placement and other methodological factors on participant recruitment, adherence and data loss in observational studies of adult physical behaviours. METHODS: The review was in accordance with the Preferred Reporting Items for Systematic Reviews and Meta Analyses (PRISMA). Observational studies of adults including accelerometer measurement of physical behaviours were identified using database (MEDLINE (Ovid), Embase, PsychINFO, Health Management Information Consortium, Web of Science, SPORTDiscus and Cumulative Index to Nursing & Allied Health Literature) and supplementary searches to May 2022. Information regarding study design, accelerometer data collection methods and outcomes were extracted for each accelerometer measurement (study wave). Random effects meta-analyses and narrative syntheses were used to examine associations of methodological factors with participant recruitment, adherence and data loss. RESULTS: 123 accelerometer data collection waves were identified from 95 studies (92.5% from high-income countries). In-person distribution of accelerometers was associated with a greater proportion of invited participants consenting to wear an accelerometer (+ 30% [95% CI 18%, 42%] compared to postal distribution), and adhering to minimum wear criteria (+ 15% [4%, 25%]). The proportion of participants meeting minimum wear criteria was higher when accelerometers were worn at the wrist (+ 14% [ 5%, 23%]) compared to waist. Daily wear-time tended to be higher in studies using wrist-worn accelerometers compared to other wear locations. Reporting of information regarding data collection was inconsistent. CONCLUSION: Methodological decisions including accelerometer wear-location and method of distribution may influence important data collection outcomes including recruitment and accelerometer wear-time. Consistent and comprehensive reporting of accelerometer data collection methods and outcomes is needed to support development of future studies and international consortia. Review supported by the British Heart Foundation (SP/F/20/150002) and registered (Prospero CRD42020213465).

Association of Accelerometer-Measured Physical Activity Level With Risks of Hospitalization for 25 Common Health Conditions in UK Adults.

Importance: Higher physical activity levels are associated with lower risks of cancer, cardiovascular disease, and diabetes, but associations with many common and less severe health conditions are not known. These conditions impose large health care burdens and reduce quality of life. Objectives: To investigate the association between accelerometer-measured physical activity and the subsequent risk of hospitalization for 25 common reasons for hospitalization and to estimate the proportion of these hospitalizations that might have been prevented if participants had higher levels of physical activity. Design, Setting, and Participants: This prospective cohort study used data from a subset of 81 717 UK Biobank participants aged 42 to 78 years. Participants wore an accelerometer for 1 week (between June 1, 2013, and December 23, 2015) and were followed up over a median (IQR) of 6.8 (6.2-7.3) years; follow-up for the current study ended in 2021 (exact date varied by location). Exposures: Mean total and intensity-specific accelerometer-measured physical activity. Main Outcomes and Measures: Hospitalization for the most common health conditions. Cox proportional hazards regression analysis was used to estimate hazard ratios (HRs) and 95% CIs for mean accelerometer-measured physical activity (per 1-SD increment) and risks of hospitalization for 25 conditions. Population-attributable risks were used to estimate the proportion of hospitalizations for each condition that might be prevented if participants increased their moderate to vigorous physical activity (MVPA) by 20 minutes per day. Results: Among 81 717 participants, the mean (SD) age at accelerometer assessment was 61.5 (7.9) years; 56.4% were female, and 97.0% self-identified as White. Higher levels of accelerometer-measured physical activity were associated with lower risks of hospitalization for 9 conditions: gallbladder disease (HR per 1 SD, 0.74; 95% CI, 0.69-0.79), urinary tract infections (HR per 1 SD, 0.76; 95% CI, 0.69-0.84), diabetes (HR per 1 SD, 0.79; 95% CI, 0.74-0.84), venous thromboembolism (HR per 1 SD, 0.82; 95% CI, 0.75-0.90), pneumonia (HR per 1 SD, 0.83; 95% CI, 0.77-0.89), ischemic stroke (HR per 1 SD, 0.85; 95% CI, 0.76-0.95), iron deficiency anemia (HR per 1 SD, 0.91; 95% CI, 0.84-0.98), diverticular disease (HR per 1 SD, 0.94; 95% CI, 0.90-0.99), and colon polyps (HR per 1 SD, 0.96; 95% CI, 0.94-0.99). Positive associations were observed between overall physical activity and carpal tunnel syndrome (HR per 1 SD, 1.28; 95% CI, 1.18-1.40), osteoarthritis (HR per 1 SD, 1.15; 95% CI, 1.10-1.19), and inguinal hernia (HR per 1 SD, 1.13; 95% CI, 1.07-1.19), which were primarily induced by light physical activity. Increasing MVPA by 20 minutes per day was associated with reductions in hospitalization ranging from 3.8% (95% CI, 1.8%-5.7%) for colon polyps to 23.0% (95% CI, 17.1%-28.9%) for diabetes. Conclusions and Relevance: In this cohort study of UK Biobank participants, those with higher physical activity levels had lower risks of hospitalization across a broad range of health conditions. These findings suggest that aiming to increase MVPA by 20 minutes per day may be a useful nonpharmaceutical intervention to reduce health care burdens and improve quality of life.

Charting a Course for Smartphones and Wearables to Transform Population Health Research.

The use of data from smartphones and wearable devices has huge potential for population health research, given the high level of device ownership; the range of novel health-relevant data types available from consumer devices; and the frequency and duration with which data are, or could be, collected. Yet, the uptake and success of large-scale mobile health research in the last decade have not met this intensely promoted opportunity. We make the argument that digital person-generated health data are required and necessary to answer many top priority research questions, using illustrative examples taken from the James Lind Alliance Priority Setting Partnerships. We then summarize the findings from 2 UK initiatives that considered the challenges and possible solutions for what needs to be done and how such solutions can be implemented to realize the future opportunities of digital person-generated health data for clinically important population health research. Examples of important areas that must be addressed to advance the field include digital inequality and possible selection bias; easy access for researchers to the appropriate data collection tools, including how best to harmonize data items; analysis methodologies for time series data; patient and public involvement and engagement methods for optimizing recruitment, retention, and public trust; and methods for providing research participants with greater control over their data. There is also a major opportunity, provided through the linkage of digital person-generated health data to routinely collected data, to support novel population health research, bringing together clinician-reported and patient-reported measures. We recognize that well-conducted studies need a wide range of diverse challenges to be skillfully addressed in unison (eg, challenges regarding epidemiology, data science and biostatistics, psychometrics, behavioral and social science, software engineering, user interface design, information governance, data management, and patient and public involvement and engagement). Consequently, progress would be accelerated by the establishment of a new interdisciplinary community where all relevant and necessary skills are brought together to allow for excellence throughout the life cycle of a research study. This will require a partnership of diverse people, methods, and technologies. If done right, the synergy of such a partnership has the potential to transform many millions of people's lives for the better.

Amount and intensity of physical activity and risk of incident cancer in the UK Biobank.

Importance: The influence of total daily and light intensity activity on cancer risk remains unclear, as most existing knowledge is drawn from studies relying on self-reported leisure-time activities of moderate-vigorous intensity. Objective: To investigate associations between total daily activity, including step counts, and activity intensity on incident cancer risk. Design Setting and Participants: Prospective analysis of cancer-free UK Biobank participants who wore accelerometers for 7-days (between 2013-2015), followed for cancer incidence through national registries (mean follow-up 5.8 years (SD=1.3)). Exposures: Time-series machine learning models derived daily total activity (average acceleration), behaviour time, step counts, and peak 30-minute cadence from wrist-based accelerometer data. Main Outcomes and Measures: A composite cancer outcome of 13 cancers previously associated with low physical activity (bladder, breast, colon, endometrial, oesophageal adenocarcinoma, gastric cardia, head and neck, kidney, liver, lung, myeloid leukaemia, myeloma, and rectum) based on previous studies of self-reported activity. Cox proportional hazards regression models estimated hazard ratios (HR) and 95% confidence intervals (CI), adjusted for age, sex, ethnicity, smoking, alcohol, education, Townsend Deprivation Index, and reproductive factors. Associations of reducing sedentary time in favour of increased light and moderate-vigorous activity were examined using compositional data analyses. Results: Among 86 556 participants (mean age 62.0 years (SD=7.9) at accelerometer assessment), 2 669 cancers occurred. Higher total physical activity was associated with a lower overall cancer risk (HR1SD=0.85, [95%CI 0.81-0.89]). On average, reallocating one hour/day from sedentary behaviour to moderate-vigorous physical activity was associated with a lower risk (HR=0.92, [0.89-0.95]), as was reallocating one hour/day to light-intensity physical activity (HR=0.94, [0.92-0.96]). Compared to individuals taking 5 000 daily steps, those who took 9 000 steps had an 18% lower risk of physical-activity-related cancer (HR=0.82, [0.74-0.90]). We found no significant association with peak 30-minute cadence after adjusting for total steps. Conclusion and Relevance: Higher total daily physical activity and less sedentary time, in favour of both light and moderate-vigorous intensity activity, were associated with a lower risk of certain cancers. For less active adults, increasing step counts by 4 000 daily steps may be a practical public health intervention for lowering the risk of some cancers.