IMPORTANCE: Recommendations for the number of steps per day may be easier to enact for some people than the current time- and intensity-based physical activity guidelines, but the evidence to support ...steps-based goals is limited. OBJECTIVE: To describe the associations of step count and intensity with all-cause mortality and cancer and cardiovascular disease (CVD) incidence and mortality. DESIGN, SETTING, AND PARTICIPANTS: This population-based prospective cohort study used data from the UK Biobank for 2013 to 2015 (median follow-up, 7 years) and included adults 40 to 79 years old in England, Scotland, and Wales. Participants were invited by email to partake in an accelerometer study. Registry-based morbidity and mortality were ascertained through October 2021. Data analyses were performed during March 2022. EXPOSURES: Baseline wrist accelerometer-measured daily step count and established cadence-based step intensity measures (steps/min): incidental steps, (<40 steps/min), purposeful steps (≥40 steps/min); and peak-30 cadence (average steps/min for the 30 highest, but not necessarily consecutive, min/d). MAIN OUTCOMES AND MEASURES: All-cause mortality and primary and secondary CVD or cancer mortality and incidence diagnosis. For cancer, analyses were restricted to a composite cancer outcome of 13 sites that have a known association with reduced physical activity. Cox restricted cubic spline regression models were used to assess the dose-response associations. The linear mean rate of change (MRC) in the log-relative hazard ratio for each outcome per 2000 daily step increments were also estimated. RESULTS: The study population of 78 500 individuals (mean SD age, 61 8 years; 43 418 55% females; 75 874 97% White individuals) was followed for a median of 7 years during which 1325 participants died of cancer and 664 of CVD (total deaths 2179). There were 10 245 incident CVD events and 2813 cancer incident events during the observation period. More daily steps were associated with a lower risk of all-cause (MRC, −0.08; 95% CI, −0.11 to −0.06), CVD (MRC, −0.10; 95% CI, −0.15 to −0.06), and cancer mortality (MRC, 95% CI, −0.11; −0.15 to −0.06) for up to approximately 10 000 steps. Similarly, accruing more daily steps was associated with lower incident disease. Peak-30 cadence was consistently associated with lower risks across all outcomes, beyond the benefit of total daily steps. CONCLUSIONS AND RELEVANCE: The findings of this population-based prospective cohort study of 78 500 individuals suggest that up to 10 000 steps per day may be associated with a lower risk of mortality and cancer and CVD incidence. Steps performed at a higher cadence may be associated with additional risk reduction, particularly for incident disease.
Machine learning (ML) accelerometer data processing methods have potential to improve the accuracy of device-based assessments of physical activity (PA) in young children. Yet the uptake of ML ...methods by health researchers has been minimal and the use of cut-points (CP) continues to be the norm, despite evidence of significant misclassification error. The lack of studies demonstrating a relative advantage for ML approaches over CP methods maybe a key contributing factor.
The current study compared the accuracy of PA intensity predictions provided by ML classification models and previously published CPs for preschool-aged children.
In a free-living study, 31 preschool-aged children (mean age = 4.0 ± 0.9 y) wore wrist and hip ActiGraph GT3X+ accelerometers while completing a video recorded 20-minute free play session. Ground truth PA intensity was coded continuously using the Children's Activity Rating Scale (CARS). Accelerometer data was classified as sedentary (SED), light intensity (LPA), or moderate-to-vigorous intensity (MVPA) using ML random forest PA classifiers and published CPs for preschool-aged children. Performance differences were evaluated in a hold-out sample by comparing weighted kappa statistics, classification accuracy for each intensity band, and equivalence testing.
ML classification models (hip: κ = 0.76; wrist: κ = 0.72) exhibited significantly higher agreement with ground truth PA intensity than CP methods (hip: κ = 0.38-0.49; wrist: κ = 0.31-0.44). For the ML models, classification accuracy for SED and LPA ranged from 83% - 88%, while classification accuracy for MVPA ranged from 68% - 78%. For the CP's, classification accuracy ranged from 50% - 94% for SED, 19% - 75% for LPA, and 44% - 76.1% for MVPA. ML classification models showed equivalence (within ± 0.5 SD) with directly observed time in SED, LPA, and MVPA. None of the CP's exhibited evidence of equivalence.
Under free living conditions, ML classification models for hip or wrist accelerometer data provide more accurate assessments of PA intensity in young children than CP methods. The results demonstrate the relative advantage of ML methods over threshold-based approaches and adds to a growing evidence base supporting the feasibility and accuracy of ML accelerometer data processing methods.
Celotno besedilo
Dostopno za:
DOBA, IZUM, KILJ, NUK, PILJ, PNG, SAZU, SIK, UILJ, UKNU, UL, UM, UPUK
Machine learning (ML) activity classification models trained on laboratory-based activity trials exhibit low accuracy under free-living conditions. Training new models on free-living accelerometer ...data, reducing the number of prediction windows comprised of multiple activity types by using shorter windows, including temporal features such as standard deviation in lag and lead windows, and using multiple sensors may improve the classification accuracy under free-living conditions. The objective of this study was to evaluate the accuracy of Random Forest (RF) activity classification models for preschool-aged children trained on free-living accelerometer data. Thirty-one children (mean age = 4.0 ± 0.9 years) completed a 20 min free-play session while wearing an accelerometer on their right hip and non-dominant wrist. Video-based direct observation was used to categorize the children's movement behaviors into five activity classes. The models were trained using prediction windows of 1, 5, 10, and 15 s, with and without temporal features. The models were evaluated using leave-one-subject-out-cross-validation. The F-scores improved as the window size increased from 1 to 15 s (62.6%-86.4%), with only minimal improvements beyond the 10 s windows. The inclusion of temporal features increased the accuracy, mainly for the wrist classification models, by an average of 6.2 percentage points. The hip and combined hip and wrist classification models provided comparable accuracy; however, both the models outperformed the models trained on wrist data by 7.9 to 8.2 percentage points. RF activity classification models trained with free-living accelerometer data provide accurate recognition of young children's movement behaviors under real-world conditions.
Wearable devices can capture unexplored movement patterns such as brief bursts of vigorous intermittent lifestyle physical activity (VILPA) that is embedded into everyday life, rather than being done ...as leisure time exercise. Here, we examined the association of VILPA with all-cause, cardiovascular disease (CVD) and cancer mortality in 25,241 nonexercisers (mean age 61.8 years, 14,178 women/11,063 men) in the UK Biobank. Over an average follow-up of 6.9 years, during which 852 deaths occurred, VILPA was inversely associated with all three of these outcomes in a near-linear fashion. Compared with participants who engaged in no VILPA, participants who engaged in VILPA at the sample median VILPA frequency of 3 length-standardized bouts per day (lasting 1 or 2 min each) showed a 38%-40% reduction in all-cause and cancer mortality risk and a 48%-49% reduction in CVD mortality risk. Moreover, the sample median VILPA duration of 4.4 min per day was associated with a 26%-30% reduction in all-cause and cancer mortality risk and a 32%-34% reduction in CVD mortality risk. We obtained similar results when repeating the above analyses for vigorous physical activity (VPA) in 62,344 UK Biobank participants who exercised (1,552 deaths, 35,290 women/27,054 men). These results indicate that small amounts of vigorous nonexercise physical activity are associated with substantially lower mortality. VILPA in nonexercisers appears to elicit similar effects to VPA in exercisers, suggesting that VILPA may be a suitable physical activity target, especially in people not able or willing to exercise.
ABSTRACT Introduction Tailoring physical activity interventions to individual chronotypes and preferences by time of day could promote more effective and sustainable behavior change; however, our ...understanding of circadian physical behavior patterns is very limited. Objective To characterize and compare 24‐h physical behavior patterns expressed relative to clock time (the standard measurement of time‐based on a 24‐h day) versus wake‐up time in a large British cohort age 46. Methods Data were analyzed from 4979 participants in the age 46 sweep of the 1970 British Cohort Study who had valid activPAL accelerometer data across ≥4 days. Average steps and upright time (time standing plus time stepping) per 30‐min interval were determined for weekdays and weekends, both in clock time and synchronized to individual wake‐up times. Results The mean weekday steps were 9588, and the mean weekend steps were 9354. The mean weekday upright time was 6.6 h, and the mean weekend upright time was 6.4 h. When synchronized to wake‐up time, steps peaked 1 h after waking on weekdays and 2.5 h after waking on weekends. Upright time peaked immediately, in the first 30‐min window, after waking on both weekdays and weekends. Conclusions Aligning accelerometer data to wake‐up times revealed distinct peaks in stepping and upright times shortly after waking. Activity built up more gradually across clock time in the mornings, especially on weekends. Synchronizing against wake‐up times highlighted the importance of circadian rhythms and personal schedules in understanding population 24‐h physical behavior patterns, and this may have important implications for promoting more effective and sustainable behavior change.
•There was a dose–response of healthy lifestyle behaviour combinations against infection mortality.•Associations were independent of multiple markers of overall health status.•Mortality risk was ...reduced among both the low and high-risk segments of the population.
In this community-based cohort study, we investigated the relationship between combinations of modifiable lifestyle risk factors and infectious disease mortality. Participants were 468,569 men and women (56.5 ± 8.1, 54.6% women) residing in the United Kingdom. Lifestyle indexes included traditional and emerging lifestyle risk factors based on health guidelines and best practice recommendations for: physical activity, sedentary behaviour, sleep quality, diet quality, alcohol consumption, and smoking status. The main outcome was mortality from infectious diseases, including pneumonia, and coronavirus disease 2019 (COVID-19). Meeting public health guidelines or best practice recommendations among combinations of lifestyle risk factors was inversely associated with mortality. Hazard ratios ranged between 0.26 (0.23–0.30) to 0.69 (0.60–0.79) for infectious disease and pneumonia. Among participants with pre-existing cardiovascular disease or cancer, hazard ratios ranged between 0.30 (0.25–0.34) to 0.73 (0.60–0.89). COVID-19 mortality risk ranged between 0.42 (0.28–0.63) to 0.75 (0.49–1.13). We found a beneficial dose–response association with a higher lifestyle index against mortality that was consistent across sex, age, BMI, and socioeconomic status. There was limited evidence of synergistic interactions between most lifestyle behaviour pairs, suggesting that the dose–response relationship among different lifestyle behaviours is not greater than the sum of the risk induced by each behaviour. Improvements in lifestyle risk factors and meeting public health guidelines or best practice recommendations could be used as an ancillary measure to ameliorate infectious disease mortality.
Vigorous physical activity (VPA) is a time-efficient way to achieve recommended physical activity levels. There is a very limited understanding of the minimal and optimal amounts of vigorous physical ...activity in relation to mortality and disease incidence.
A prospective study in 71 893 adults median age (IQR): 62.5 years (55.3, 67.7); 55.9% female from the UK Biobank cohort with wrist-worn accelerometry. VPA volume (min/week) and frequency of short VPA bouts (≤2 min) were measured. The dose-response associations of VPA volume and frequency with mortality all-cause, cardiovascular disease (CVD) and cancer, and CVD and cancer incidence were examined after excluding events occurring in the first year. During a mean post-landmark point follow-up of 5.9 years (SD ± 0.8), the adjusted 5-year absolute mortality risk was 4.17% (95% confidence interval: 3.19%, 5.13%) for no VPA, 2.12% (1.81%, 2.44%) for >0 to <10 min, 1.78% (1.53%, 2.03%) for 10 to <30 min, 1.47% (1.21%, 1.73%) for 30 to <60 min, and 1.10% (0.84%, 1.36%) for ≥60 min. The 'optimal dose' (nadir of the curve) was 53.6 (50.5, 56.7) min/week hazard ratio (HR): 0.64 (0.54, 0.77) relative to the 5th percentile reference (2.2 min/week). There was an inverse linear dose-response association of VPA with CVD mortality. The 'minimal' volume dose (50% of the optimal dose) was ∼15 (14.3, 16.3) min/week for all-cause HR: 0.82 (0.75, 0.89) and cancer HR: 0.84 (0.74, 0.95) mortality, and 19.2 (16.5, 21.9) min/week HR: 0.60 (0.50, 0.72) for CVD mortality. These associations were consistent for CVD and cancer incidence. There was an inverse linear association between VPA frequency and CVD mortality. 27 (24, 30) bouts/week was associated with the lowest all-cause mortality HR: 0.73 (0.62, 0.87).
VPA of 15-20 min/week were associated with a 16-40% lower mortality HR, with further decreases up to 50-57 min/week. These findings suggest reduced health risks may be attainable through relatively modest amounts of VPA accrued in short bouts across the week.
To evaluate the accuracy of LAB EE prediction models in preschool children completing a free-living active play session. Performance was benchmarked against EE prediction models trained on free ...living (FL) data.
25 children (mean age = 4.1±1.0 y) completed a 20-minute active play session while wearing a portable indirect calorimeter and ActiGraph GT3X+ accelerometers on their right hip and non-dominant wrist. EE was predicted using LAB models which included Random Forest (RF) and Support Vector Machine (SVM) models for the wrist, and RF and Artificial Neural Network (ANN) models for the hip. Two variations of the LAB models were evaluated; 1) an "off the shelf" model without additional training; 2) models retrained on free-living data, replicating the methodology used in the original calibration study (retrained LAB). Prediction errors were evaluated in a hold-out sample of 10 children.
Root mean square error (RMSE) for the FL and retrained LAB models ranged from 0.63-0.67 kcals/min. In the hold out sample, RMSE's for the hip LAB (0.62-0.71), retrained LAB (0.58-0.62) and FL models (0.61-0.65) were similar. For the wrist placement, FL SVM had a significantly higher RMSE (0.73 ± 0.29 kcals/min) than the retrained LAB SVM (0.63 ± 0.30 kcals/min) and LAB SVM (0.64 ± 0.18 kcals/min). The LAB (0.64 ± 0.28), retrained LAB (0.64 ± 0.25), and FL (0.62 ± 0.26) RF exhibited comparable accuracy.
Machine learning EE prediction models trained on LAB and FL data had similar accuracy under free-living conditions.
Celotno besedilo
Dostopno za:
DOBA, IZUM, KILJ, NUK, PILJ, PNG, SAZU, SIK, UILJ, UKNU, UL, UM, UPUK
PURPOSEPattern recognition approaches to accelerometer data processing have emerged as viable alternatives to cut-point methods. However, few studies have explored the validity of pattern recognition ...approaches in pre-schoolers; and none have compared supervised learning algorithms trained on hip and wrist data. To develop, test, and compare activity class recognition algorithms trained on hip, wrist, and combined hip and wrist accelerometer data in pre-schoolers.
METHODS11 children aged 3 - 6 y (mean age 4.8 ± 0.9 y) completed 12 developmentally appropriate PA trials while wearing an ActiGraph GT3X+ accelerometer on the right hip and non-dominant wrist. PA trials were categorised as sedentary (SED), light activity games (LG), moderate-to-vigorous games (MVG), walking (WA), and running (RU). Random forest (RF) and support vector machine (SVM) classifiers were trained using time and frequency domain features from the vector magnitude of the raw signal. Features were extracted from 15 s non-overlapping windows. Classifier performance was evaluated using leave-one-out-cross-validation.
RESULTSCross-validation accuracy for the hip, wrist, and combine hip and wrist RF models was 0.80 (95% CI:0.79 - 0.82), 0.78 (95% CI:0.77-0.80), 0.82 (95% CI:0.80 - 0.83), respectively. Accuracy for Hact, Wact, and HWact SVM models was 0.81 (95% CI:0.80 - 0.83), 0.80 (95% CI:0.79-0.80), 0.85 (95% CI:0.84 - 0.86), respectively. Recognition accuracy was consistently excellent for SED (> 90%), moderate for LG, MVG, and RU (69-79%), and modest for WA (61-71%).
CONCLUSIONSMachine learning algorithms such as RF and SVM are useful for predicting PA class from accelerometer data collected in preschool children. While classifiers trained on hip or wrist data provided acceptable recognition accuracy, the combination of hip and wrist accelerometer delivered better performance.
Detection of non-wear periods is an important step in accelerometer data processing. This study evaluated five non-wear detection algorithms for wrist accelerometer data and two rules for non-wear ...detection when non-wear and sleep algorithms are implemented in parallel. Non-wear algorithms were based on the standard deviation (SD), the high-pass filtered acceleration, or tilt angle. Rules for differentiating sleep from non-wear consisted of an override rule in which any overlap between non-wear and sleep was deemed non-wear; and a 75% rule in which non-wear periods were deemed sleep if the duration was < 75% of the sleep period. Non-wear algorithms were evaluated in 47 children who wore an ActiGraph GT3X+ accelerometer during school hours for 5 days. Rules for differentiating sleep from non-wear were evaluated in 15 adults who wore a GeneActiv Original accelerometer continuously for 24 hours. Classification accuracy for the non-wear algorithms ranged between 0.86-0.95, with the SD of the vector magnitude providing the best performance. The override rule misclassified 37.1 minutes of sleep as non-wear, while the 75% rule resulted in no misclassification. Non-wear algorithms based on the SD of the acceleration signal can effectively detect non-wear periods, while application of the 75% rule can effectively differentiate sleep from non-wear when examined concurrently.