Behaviour-based movement cut-off points in 3-year old children comparing wrist- with hip-worn actigraphs MW8 and GT3X

Daniel Jansson; Rikard Westlander; Jonas Sandlund; Christina E West; Magnus Domellöf; Katharina Wulff

doi:10.1371/journal.pone.0316747

. 2025 Mar 26;20(3):e0316747. doi: 10.1371/journal.pone.0316747

Behaviour-based movement cut-off points in 3-year old children comparing wrist- with hip-worn actigraphs MW8 and GT3X

Daniel Jansson ^1,², Rikard Westlander ³, Jonas Sandlund ⁴, Christina E West ³, Magnus Domellöf ^3,^*, Katharina Wulff ^5,^6,^*

Editor: Duncan S Buchan⁷

PMCID: PMC11940821 PMID: 40138295

Abstract

Background

Behaviour-based physical intensity evaluation requires rigorous calibration before application in long-term recordings of children’s sleep/activity patterns. This study aimed at (i) calibrating activity counts of motor behaviour measured simultaneously with MotionWatch 8 (MW8) and ActiGraph (GT3X) in 3-year-old children, (ii) documenting movement intensities in 30s-epochs at wrist/hip positions, and (iii) evaluating the accuracy of cut-off agreements between different behavioural activities.

Methods

Thirty 3-year-old children of the NorthPop cohort performed six directed behavioural activities individually, each for 8–10 minutes while wearing two pairs of devices at hip and wrist position. These naturally-occurring behaviours were aligned to movement intensities from ‘motionless’ (watching cartoons) and ‘sedentary’ (recumbent story listening, sit and handcraft) to ‘light activity’ (floor play with toys), ‘moderate activity’ (engaging in a brisk walk) and ‘vigorous activity (a sprinting game). Time-keeping was ensured using direct observation by an observer. Receiver-Operating-Curve classification was applied to determine activity thresholds and to assign two composite movement classes.

Results

Activity counts of MW8 and GT3X pairs of wrist-worn (rho = 0.94) and hip-worn (rho = 0.90) devices correlated significantly (p < 0.001). Activity counts at hip position were significantly lower compared to those at the wrist position (p < 0.001), irrespective of device type. Sprinting, floorball/walk and floorplay assigned as ‘physically mobile’ classes achieved outstanding accuracy (AUC > 0.9) and two sedentary and a motionless activities assigned into ‘physically stationary’ classes achieved excellent accuracy (AUC > 0.8).

Conclusion

This calibration provides useful cut-offs for physical activity levels of preschool children. Contextual information of behaviour is advantageous over intensity classifications only, because interventions will focus on behaviour-allocated time to reduce a sedentary lifestyle. Our comparative calibration is one step forward to behaviour-based movement guidelines for 3-year-old children.

Introduction

Movement-related assessments of habitual activities, such as physical activity (PA) levels or sleep/circadian timing, have typically taken a segregated approach in fields such as epidemiology, sports medicine, rehabilitation or chronobiology [1–3]. Historically, terminology also developed independently, with ‘actimeter/actigraph’ used in sleep/chronobiology and ‘accelerometer’ in sports medicine (S1 Table). Similarly, evidence regarding the combination of movement behaviours over a 24-h period using compositional analyses [4] is uncommon but growing [5] with studies emerging [6–8]. The Commission on Ending Childhood Obesity recognised the importance of interaction among PA, sedentary behaviour and adequate sleep on the child’s well-being [9]. Canadian and Australian 24-hour movement guidelines were developed [4,10], followed by World Health Organisation (WHO) guidelines on the amount of time in a 24-hour day particularly for children under five years of age [1]. Global recommendations state that children 3–4 years old should spend at least 180 minutes a day in a variety of physical activities at any intensity, of which at least 60 minutes is moderate- to vigorous-intense PA [1]. This PA recommendation is considered strong, albeit of very low-quality evidence [1].

Challenges for compositional analysis

There are a number of calibration studies for wearable movement sensors using short epochs (5–15s) in preschool children [11–21] (S1 Table). Objective measurements of physical movements with wearables have proven feasible and valid for estimating sedentary time during waking hours in young children in the field [22–24]. However large-scale cohort recordings capture behaviour over many days and require the distinction of motionless alert from daytime sleep, which necessitates devices to be compared and calibrated for their accuracy in measuring PA and derived sleep [25,26]. Every device type needs to be calibrated for typical behaviours of a given population [27], and epoch as well as commonly used body sites, mainly the wrist and hip [28,29]. The challenges with accelerometer calibration, dividing-up absolute and relative intensities, have been competently outlined by Arvidsson et al. 2019 [30] and consensus recently discussed by Migueles et al. 2022 [31].

In brief, the standard placement of devices for PA alone has been at the hip by fitting it on an elastic belt around the waist [32], while devices developed for sleep/circadian rhythm research have typically been attached to the non-dominant wrist [33,34] (S1 Table). Accordingly, the devices’ sensors and firmwares operate differently according to their purpose: Measurements for assessing sleep/circadian patterns last over complete 24h cycles and longer epochs (30 to 60s) for several weeks with sensors and firmware adjusted for wrist movements [35]; while sensors used for PA measurements are usually implemented for high-resolution (seconds), short-term (min-hours) measurements on the hip-position to determine energy expenditure [36–38]. Ramification are intricate, including systematically different activity levels from different brands, because their raw accelerometer outputs are unequal [39] or when using same brand devices, they classify sedentary behaviour and PA accurately under laboratory settings but not under free-living conditions [40]. Scaling factors implemented into algorithms may overcome inherent discrepancies, provided that the research communities share data collections and meta-data from as many models as possible [2].

On the nature of compliance, wrist position was found to achieve higher compliance than hip position [41]. Therefore, preschool children are recommended to attach the device to the non-dominant wrist [42,43].

Rational of the study

The aforementioned recommendations were based on systematic reviews with an overall ‘moderate’ certainty of the evidence, stating little research was available to inform about specific aspects such as dose-response studies on the type, duration, intensity and context of individual behaviours [4]. In turn, uncertainty was reported in how light PA and moderate-to-vigorous intensity PA is best defined for young children. Taken together, the guidelines recommend durations and PA quantities for several age ranges without accounting for different detection methods and different body positions to track behavioural situations. These gaps can be closed by: (i) comparative assessments of different devices; (ii) worn simultaneously at different body positions; (iii) during age-appropriate behaviours accounting for different PA intensities; (iv) in children at narrower age ranges.

However, many devices are on the market and researchers are confronted with analytical challenges in face of the technical diversity. The appropriateness in filter settings, algorithms, recording modes, size and position for the combination of PA and sleep are under scrutiny [44]. Furthermore, various research-grade data loggers capture light exposure levels due to its effect on circadian/sleep behaviour [45] in addition to body movements [46], at the expense of recording period and size. Two of those have been validated for sleep/circadian quantities and PA in adults and children: MotionWatch 8 (MW8, Camntech LTd, UK) and ActiGraph GT3X (theactigraph, FL, US). Both use different algorithms and output modes [47–49]. The MW8 is primarily used in longitudinal sleep/circadian studies, which prioritises a 30s epoch as a result of the validation with polysomnography to derive sleep parameters from movement patterns [50] and to allow longer recording periods. The GT3X uses the Cole-Kripke [51] and Sadeh et al. [52] algorithms as their standard sleep algorithms.

The present study

The present study was specifically designed to address some of the aforementioned gaps by measuring PA intensities of different behaviours simultaneously at two body positions (non-dominant wrist and hip) with two models (MW8 and GT3X) using the same epoch length of 30s in children within a narrow age range (3.5 years). Both device types were expected to produce systematically different intensities for the same behaviour. It was important to investigate, how much of a difference it made for time spent ‘physically mobile’ (body in motion) from ‘physically stationary’ (motionless/sedentary activities, excluding naps/sleep). The two devices and positions were also necessary, because each type has generally favoured only one of the positions. Our design made a comparison of the distribution of intensity levels within and between device types across positions possible. We first piloted various activities and then chosen six different behaviours to calculate behaviour-based activities for which we determined cut-off boundaries and their accuracies. This design goes beyond general calibration and insights were anticipated in the accuracy of discriminating intensity classes with a 30s eoch length and how to interpret scores between positions and devices. Finally, we envisage to assign behavioural classes according to cut-off intensity boundaries in our prospective, population-based NorthPop birth cohort (https://www.katlab.org/ [under ‘people’], www.northpop.se), compare their allocated time with movement targets for preschool children and relate to health indicators.

Materials and methods

Participants

In total, 30 children were recruited from the ongoing prospective, population-based NorthPop cohort. Three-year-old, typically developing children were included. Children with any chronic disease or weight outside the normative range ( ±2 standard deviations) using a Swedish growth reference [53] were not included to minimise confounding variables, which could potentially skew the calibration of activity counts. This was decided to ensure that the cut-off points we develop are generalisable to the broader population of typically developing 3-year-old children.

Children in this study were part of a larger actigraphic project within the original NorthPop cohort (Dnr 2014/224-31), so all children had experience wearing actigraphs during free-living conditions. Therefore, all participants were already familiar with the procedure, and adding more devices was well-tolerated. Further, all legal guardians received written and oral information about the study and signed a written consent prior to the data collection. In addition to legal guardians’ consent, we explained the procedures to the children in simple terms and demonstrated where the monitors would be placed using a teddy bear to ensure their understanding and comfort. This calibration study was conducted in agreement with the declaration of Helsinki and approved by the regional ethical review board (Dnr 2020/01254).

Procedures

All measurements were conducted between October 2020 and January 2021. Each parent-child pair was studied separately on one test occasion in the E-health laboratory at Umeå University. The E-health laboratory was designed to mimic a small apartment to simulate daily living. We collected the data during daytime between 9.00 to 12.00 or 13.00 to 16.00. Basic anthropometric measurements were collected for each child. Height was measured to the nearest 0.1 cm using a folding rule and body mass to the nearest 0.1 kg using an electronic scale. A pilot evaluation of the feasibility of various activities was carried out before commencing the data collection. A final study protocol was developed to monitor three stationary behaviors and three physically demanding behaviors typical for children to engage in at this age under free-living conditions. The six behaviours represent different movement patterns and intensity levels: Physically mobile levels endorsed physically ‘vigorous’, ‘moderate’ and ‘light’ activities, and physically stationary levels endorsed ‘sedentary crafts’, ‘recumbent listening’ and ‘sedentary screen time’ (Fig 1 with explanatory table).

Fig 1 — The activity intensity was transformed into count values recorded every 30 seconds (epochs). Each type of behaviour was observed by stop-watch to last between 8 to 10 minutes. Bottom: Explanatory table describing the six behavioural activities at the top.

Sedentary screen time involved watching a cartoon on a large screen TV with their parent and recumbent listening involved a parent reading their child’s favourite story. The ‘sedentary crafts’ activity involved the child sitting at a table, drawing or placing stickers in a sticker book. The three active behaviours were planned to be progressively more intense: Playing on the floor with toys represented as light activity (defined as little increase in heart rate, can talk while playing);, moderate activity (defined as increased heart rate and breathing rate) was represented by walking or playing floorball, and vigorous activity (defined as significantly high heart rate and breathing hard and fast) represented by competitive sprinting games. Here the children were asked to carry balls from a box on one side of a long corridor to another box at the other end as quickly as possible. They had short breaks to take their breath at each side and they did not run continuously fast for 10 min. Each activity was performed for approximately 10 minutes, with a minimum of 3 minutes rest between activities. If the activity was less than 8 minutes, we repeated that specific activity until it was within the time limits (8–10 min). The activities could be executed in any order depending on the child’s preferences, whilst a relaxation break with a snack was offered after about half way through the activities. After the vigorous activity, the child had a little longer break (≈10 min) to minimize the influence of fatigue on the following activity. Two research team members were skilled to explain the study to the child and their parent, to handle the technical devices and software, and trained in momentary time sampling, which is here termed ‘directly observed’. The ‘direct observation protocol’ entails one person acting as a ‘bystander’ watching (observing) each child’s activities from a ‘bird’s perspective’ and documenting each activity’s start and stop with a stopwatch. This enabled the second person to engage with the child without being distracted by checking the time. The documented start and stop times were essential for extracting the sequences of activities from the times series. Children were fitted with four movement sensors (MotionWatch 8 ([MW8], CamNtech Ltd., Cambridge UK), ActiGraph [GT3X], ActiGraph, LLC, Fort Walton Beach, FL) and a heart-rate monitor (Actiheart 5, CamNtech), whose data were not part of the current analyses. A pair of MW8 and GT3X were attached to the non-dominant wrist and another pair to the hip. The hip-worn sensors were mounted to an elastic belt so that the black circle of the GT3X and the label of the MW8 were both facing downwards. The wrist-worn sensors were mounted to the same strap pointing towards the hand, following our standard operational procedures. The children wore the belt around the waist, placed 1–2 cm beneath the umbilicus. The same four devices were used in all children during all activities to minimise variability between sensors. This set-up enabled comparisons between different types of sensors in the same position (wrist-wrist versus hip-hip) as well as within and between sensors across different positions (wrist versus hip). Missing data were rare and when discovered, participant would be replaced until we had complete datasets. However, for one participant we missed data for the MW8 at hip position.

Data from MW8 and GT3X were captured in 30-second epochs. The GT3X is a lightweight (27 g), relatively obtrusive (3.8 × 3.7 × 1.8 cm) device with a rechargeable battery. Batteries were charged before each trial to avoid missing data. The GT3X collects motion data on three axes (x, y, z) and is designed to record accelerations ranging from 0.05 to 2.5g. The sampling frequency of 30 Hz and “triaxial mode” was used for both GT3X devices and the “vector magnitude (VM)” for the analyses. The MW8 is battery-powered and weighs 9.1g. It is of relatively unobtrusive dimensions (3.6 × 2.82 × 0.94 cm), which is an advantage for long-term wear in younger children (e.g., 3-year-olds). Its continuous operation has the option of single-axis or tri-axial recording mode (producing a vector magnitude count per epoch). The MW8 has a built-in light sensor and a time-stamp marker button. The MW8 sensing ranges between 0.01 and 8 g, and the minimum ‘not moving’ threshold is 0.1 g. Data are sampled at 50 Hz and bandwidth-filtered between 3 Hz and 11 Hz. An ‘activity count’ (unitless) is derived from the highest of the 50 samples/second, and these values are accumulated over the length of an epoch (30 seconds) [54]. Both MW8 devices were set up to record in single-axis ‘MotionWatch Mode 1’ and not tri-axial mode, because this mode has been validated against polysomnography. It also allows longer recording periods, as implemented in the ongoing collection of sleep-wake and light exposure data in the Northpop cohort. After data collection, the raw time series data were downloaded using Motionware software and exported into an Excel spreadsheet. The start and stop times of the behavioural observation protocol were matched to the 8 to 10 minutes-long periods of raw time series data in Excel. Periods of the same behaviour across the children were successively concatenated, separately for each position and device. The total duration per behaviour was calculated and compared between the six behaviours to ensure similar data length distribution (S3 Fig).

Statistical analysis

All statistical analyses were performed using the SPSS statistical package (SPSS, v. 27, Chicago, IL). Anthropometrical data at baseline were visually inspected using histograms, kurtosis, and skewness. Data were considered normally distributed and compared between males and females using a standard Student T-test. Activity count data of different intensities were not normally distributed and therefore the Spearman’s-Rank-Order correlation was used to determine the monotonic relationship between the activity measured with the wrist-worn and hip-worn devices GT3X and MW8. To approximate the similarity in the shape of the data distribution between algorithms, the data of wrist-worn devices were analyzed with linear regression. The data of hip-worn devices showed a non-linear relationship, therefore polynomial nonlinear regression models were applied. Due to the skewed distribution of activity counts were plotted as boxplots in a log scale to visualise their proportional overlap between adjacent behavioural activities.

Receiver Operator Characteristics (ROC) curves were used to determine cut-off points (intensity thresholds) from activity counts of pre-defined, observer-based behaviours [55,56]. ROC curves have previously been used to determine activity count thresholds in children between 4–8 years for Actigraph and MW8 models separately [12,13,49]. Here we used both brands simultaneously and applied ROC curves to examine the classification accuracy across brands and positions. We used a binarised approach in ROC curve classification acknowledging the activity counts to be of descending order from vigorous PA> moderate PA> light PA> sedentary> motionless alert. Classification can be performed in two ways: pair-wise ‘One-vs-One’ (Fig 2a) or one-group/composite against all other ‘One-vs-Rest’ (Fig 2b). Although we report One-vs-One details in Supplementary Information (S2 Table), here we focus on the One-vs-Rest (OvR) scheme [57]. We collapsed counts of certain adjacent behavioural activities (and assumed as one) and compared those against all other collapsed behavioural activities (Fig 2b, see ROC curves in Supporting information S3–S7 Figs). The behavioral activities were assigned with intensity class labels: the most vigorous activity ‘Sprinting’ was assigned to ‘vigorous PA’ (VPA); ‘Sprinting’ and ‘Floor ball’ combined were assigned to ‘moderate-vigorous PA’ (MVPA); ‘Sprinting’, ‘Floor ball’ and ‘Play on floor’ combined were assigned to ‘light-moderate-vigorous PA’ (LMVPA); and all these together were labelled ‘mobile PA’ class. They were set against all physically ‘stationary’ behaviours, thereby dividing ‘mobile PA’ from ‘stationary PA’. Stationary PA included ‘sedentary crafts’, ‘recumbent listening’ and ‘sedentary screentime’. We merged ‘sedentary crafts’ and ‘recumbent listening’ behaviours into ‘sedentary PA’ (SED) on the basis of their considerable overlap in movement patterns. The most immobile behaviour ‘sedentary screentime’ was assigned to ‘motionless alert’ (MOA). We delibertly split off MOA from SED to better understand their closeness to quiet awake and sleep. We did not run a separate ROC-AUC for the SED, but report the range between lower boundary of LMVPA intensity and the upper boundary of MOA intensity.

Fig 2 — Double-sided arrows indicate the cut between categories for each ROC analysis. In OvsR, two sedentary categories (Sedentary craft + Recumbent listening) were merged into SED (for which ROC analysis were done only in OvsO, see S2 Table). Moreover, SED and MOA were collapsed into an composite ‘*stationary* PA’ class and tested against the composite ‘*mobile* PA’.

To determine the best possible compromise between sensitivity and specificity for each cut-off, we applied the Youden Index (J), which is determined by calculating the sum of sensitivity and specificity minus one [58]. The cut-offs (synonym for intensity thresholds) were derived from the raw data distribution’s highest total sensitivity and specificity ratio. All values are reported as mean ± standard deviation (SD). The Area Under the Curve (AUC) quantified how well the ROC curve performed at classifying data. Values for AUC ranges from 0 to 1. It is used as a measure of accuracy for the ROC curve in discriminating classes, here in classifying count cut-offs from pre-defined, observed behavioural activities. ROC-AUC values above 0.90 are considered outstanding, 0.80–0.89 excellent, 0.70–0.79 acceptable (fair), less than 0.70 are considered poor, and an AUC as low as 0.5 indicates no greater predictive ability than by random guessing [59]. A significance level of p < 0.05 was considered significant.

Results

Anthropometrical data for all participants are presented in Table 1. The study included 30 participants, 20 males and 10 females. There was no significant difference in age (p = 0.69), height (p = 0.898) or waist circumferences (p = 0.052) between males and females. The males were heavier than females (p = 0.023), which was also reflected in their BMI values (p < 0.05). Mean activity counts were lowest when children were watching cartoons and highest when running in a sprinting game (Fig 1 and Table 2).

Table 1. Descriptive characteristics of the participants (n = 30). Shown are means ± SD.

	Males (n = 20)	Females (n = 10)	All (n = 30)
Age (years)	3.5 ± 0.1	3.6 ± 0.1	3.5 ± 0.1
Weight (kg)	17.1 ± 1.6	15.3 ± 1.0^*	16.5 ± 1.7
Height (cm)	102.8 ± 3.1	98.0 ± 3.6	101.3 ± 4.0
BMI (kg/m²)	17.2 ± 1.5	15.3 ± 1.1^*	16.5 ± 1.7
Waist circumference (cm)	53.2 ± 2.8	48.5 ± 12.9	51.7 ± 7.9

Open in a new tab

Significantly different between females and males; p < 0.05.

Table 2. Mean ± SD of activity counts (counts/30 seconds) for each activity ranked from highest to lowest intensity measured at wrist-level and hip-level with ActiGraph GT3X and MotionWatch 8. Mean ± SD of the duration spent in activity is consistently similar.

			Wrist-worn		Hip-worn
	N	Duration (min)	ActiGraph GT3X (counts/30 s)	MotionWatch 8 (counts/30 s)	ActiGraph GT3X (counts/30 s)	MotionWatch 8 (counts/30 s)
Vigorous activity	30	9.7	9017.8 ± 2745.2	1419.2 ± 457.5	2240.0 ± 586.6	1172.3 ± 357.9
Moderate activity	30	9.9	4677.3 ± 1875.3	672.2 ± 309.8	1964.7 ± 753.4	533.6 ± 250.5
Light activity	30	10.0	2359.8 ± 818.6	291.1 ± 143.9	1044.1 ± 550.4	80.3 ± 85.6
Sedentary crafts	30	10.0	1329.0 ± 736.1	128.9 ± 95	291.8 ± 317.7	16.0 ± 29.3
Recumbent listening	30	9.9	998.0 ± 1198.2	110.7 ± 142.1	253.5 ± 479.1	19.9 ± 47.4
Sedentary screen time	30	9.9	540.6 ± 724.8	62.9 ± 97.4	137.2 ± 260.3	10.2 ± 30.9

Open in a new tab

Comparing MW8 with GT3X revealed significant correlations in activity counts at their respective positions, despite different operating scales

The GT3X’s “vector magnitude” producing proportionally greater values than those of the uniaxis MW8 (Table 2). When data from the same brand and position were pooled across all activities, significant correlations were detected between counts measured with the MW8 and GT3X for wrist-worn (rho = 0.94) and hip-worn devices (rho = 0.90) (all p < 0.001, Fig 3). Averaging across all six behaviours, the mean counts/30s were significantly greater for the wrist position compared to the hip-position within the respective brands (MW8: wrist: 438.6 ± 537 counts versus hip: 300.8 ± 464.3 counts; GT3X: wrist: 3165.3 ± 3357.6 counts versus hip: 959.3 ± 937.8 counts, all p < 0.001). The mean activity counts/30s at wrist- and hip positions simultaneously measured with GT3X and MW8 for each behavioural activity are reported (Table 2) and visualized as boxplots (Fig 4). The boxplots reveal that the three physically mobile behaviours (light, moderate, vigorous activities) had a narrow within-class variation, enabling to separate each from one another. In contrast, the three physically stationary behaviours (recumbent listening, sedentary crafts, sedentary screen time) showed large interquartile ranges and overlap between classes, regardless of position or brand.

Fig 3 — = 30) during six activities ranging from ‘Motionless alert’ to ‘Vigorous PA’ for MotionWatch 8 (MW8) and ActiGraph (GT3X) at A) wrist position and B) hip position.The six activities are colour-coded in the graphs.

Fig 4 — A value of ‘1’ was added to every count in order to increase the visibility of the lower end of the scale, in particular for the stationary, calmer activities through a logarithmic scale. Zero movement over 30s epoch is not uncommon, for example during a break. Box contains 50% of values (Interquartile range [IQR] with values between Q1 and Q3 (25th–75th percentile), mid-line represents median, Whiskers represent min-max ([Q1 − 1.5 * IQR] − Q3 + 1.5 * IQR]).

Collapsing behavioural classes indicated outstanding accuracy thresholds for ‘mobile PA’ and excellent accuracy thresholds for ‘stationary PA’ intensities

OvR ROC-AUC analysis were carried out four times per device (VPA, MVPA, LMVPA and MOA). Sensitivity ranged from 89–95% and specificity from 90–94% for MVPA and VPA of wrist-worn devices. Sensitivity for mobile PA composite against stationary PA was 88% and 91%, and specificity 86% and 83% for wrist-worn MW8 and GT3X, respectively. AUC ranged from 0.95–0.98. Sensitivity for MOA (watching cartoon) against the rest was 75% and 77%, and specificity 82% and 85% for wrist-worn MW8 and GT3X, respectively. AUC was 0.86 and 0.87, respectively (Table 3).

Table 3. ROC-AUC analysis for wrist-worn MotionWatch 8 and ActiGraph GT3X devices after collapsing categories. See details for collapsed categories in Fig 2.

		Wrist- worn MotionWatch 8 (MW8)				Wrist-worn ActiGraph (GT3X)
	Intensity class	Cut-off value (counts)	Sensitivity (%)	Specificity (%)	AUC (95% CI)	Cut-off value (counts)	Sensitivity (%)	Specificity (%)	AUC (95% CI)
Mobile PA	VPA¹	>787	93	94	0.98	>4607	95	90	0.97
					(0.97 to 0.98)^*				(0.97 to 0.98)^*
	MVPA²	>408	90	93	0.97	>3038	89	93	0.97
					(0.97 to 0.98)^*				(0.96 to 0.98)^*
	LMVPA³^,^#	>215	88	86	0.95	>1782	91	83	0.95
					(0.94 to 0.96)^*				(0.94 to 0.95)^*
Stationary PA	SED⁴	118–215	NA	NA	NA	1148–1782	NA	NA	NA
	MOA⁵	<118	75	82	0.86	<1148	77	85	0.87
					(0.84–0.87)^*				(0.86–0.89)^*

Open in a new tab

p < 0.001.

LMVPA: Separates mobile physical activities (Mobile PA) from stationary activities (Stationary PA).

¹VPA: Vigorous PA versus all other activity levels.

²MVPA: Moderate-vigorous PA versus all other activity levels (light, sedentary [handcraft + recumbent], motionless).

³LMVPA: Light moderate vigorous PA versus the rest (sedentary [handcraft + recumbent], motionless).

⁴SED: Sedentary range: above MOA threshold and below LMVPA threshold.

⁵MOA: Motionless-alert [Screentime] versus all other activity level.

Sensitivity ranged from 91–96%, and specificity from 81–98% for MVPA and VPA of hip-worn devices. Sensitivity for mobile PA composite against stationary PA was 85% and 91%, and specificity 91% and 89% for hip-worn MW8 and GT3X, respectively. AUC ranged from 0.90–0.98. Sensitivity for MOA (watching cartoon) against the rest was 64 and 75%, and specificity 87 and 78% for hip-worn MW8 and GT3X, respectively. AUC was 0.81 and 0.86, respectively (Table 4).

Table 4. ROC-AUC analysis for hip-worn MotionWatch 8 and ActiGraph GT3X devices after collapsing categories. See details for collapsed categories in Fig. 2.

		Hip-worn MotionWatch 8 (MW8)				Hip-worn ActiGraph (GT3X)
	Intensity class	Cut-off value (counts)	Sensitivity (%)	Specificity (%)	AUC (95% CI)	Cut-off value (counts)	Sensitivity (%)	Specificity (%)	AUC (95% CI)
Mobile PA	VPA¹	>637	92	94	0.98	>1509	91	81	0.90
					(0.97 to 0.99)^*				(0.89 to 0.91)^*
	MVPA²	>214	96	98	0.99	>1006	94	85	0.96
					(0.99 to 0.99)^*				(0.95 to 0.96)^*
	LMVPA³^,^#	>46	85	91	0.95	>631	91	89	0.96
					(0.94 to 0.95)^*				(0.95 to 0.96)^*
Stationary PA	SED⁴	22–46	NA	NA	NA	183–631	NA	NA	NA
	MOA⁵	<22	64	87	0.81	<183	75	78	0.86
					(0.79–0.83)^*				(0.84–0.87)^*

Open in a new tab

p < 0.001.

LMVPA: Separates mobile physical activities (Mobile PA) from stationary activities (Stationary PA).

¹VPA: Vigorous PA versus all other activity levels.

²MVPA: Moderate-vigorous PA versus all other activity levels (light, sedentary [handcraft + recumbent], motionless).

³LMVPA: Light moderate vigorous PA versus the rest (sedentary [handcraft + recumbent], motionless).

⁴SED: Sedentary range: above MOA threshold and below LMVPA threshold.

⁵MOA: Motionless-alert [Screentime] versus all other activity levels.

Discussion

The motivation of conducting this calibration study came from intending to use the daily actigraphic data not only for sleep patterns but also for PA monitoring in children of an ongoing prospective birth cohort. This study aimed at providing calibration data for 3-year-old children performing various play activities and passive behaviours by means of count cut-off thresholds from wrist-worn MW8 and GT3X activity counts/30s. Additional purposes included comparison of classification accuracy between concomitantly worn wrist- and hip-mounted devices using sensitivity, specificity and ROC-AUC. Our results show that our cut-off points corroborate the separation of physically mobile behaviours from physically stationary behaviours (specificities of 83%-86% for wrist-worn cut-offs), which is in line with results of earlier studies, despite using different measures and criteria of classifying PA intensities [10,12,32,49]. The movement intensities differed not just by behaviour but also by type of device and body position. The count cut-offs from both devices (MW8, GT3X) at both positions (wrist, hip) showed outstanding accuracy (range: 0.95–0.98) for all mobile PA classes. Nevertheless, the classification accuracy for ‘motionless alert’ against all other classes showed a fair accuracy (range: 0.81–0.87).

Observational method

Direct observation was the choice of criterion to calibrate activity counts because it avoids interpretation errors associated with MET conversions or errors associated with extrapolation from atypical activities (walk/dance to a certain pace) to free-living behaviours [60]. The pilot phase has been valuable to adjust activities to the children’s age, i.e. children refused to dance to the beat of music. Instead this moderate activity was changed into a moderately paced walk with their parent or playing with miniature floorball. The use of pre-determined behaviours allowed us to classify sets of activities, i.e. as stationary. No previous studies have developed activity intensity thresholds for wrist- and hip-worn GT3X models at 30s-epochs in vector mode (VM). But thresholds have been published for GT3X models at 5s-epochs using a sampling rate of 30Hz and wrist/hip positions [12] and for the Actiwatch (Minimitter) at 1-min epochs and a 32 Hz sampling rate [58]. While the present study employed single child observations over six directed activities, Johansson et al. (2016) [12] and Finn and Specker (2000) [58] employed video recording during free activities of groups of children and subsequently assigned behaviours according to the ‘Children’s Activity Rating Scale’ [61,62]. The activities in the present study included similar types as in the other studies, i.e. watching cartoons and drawing; refrained from other types, i.e. dancing, outdoor: and included different types, i.e. sprinting game, floor-ball. [12]. Compared to Johansson’s et al. (2016), the intensity thresholds established for GT3X in VM in the present study were higher. This was expected from an epoch five times as long, confirming thresholds to be epoch-specific in addition to being age-specific [12].

Types and terminology of movement behaviours

We found substantial overlap in activity counts among the three physically stationary’ activities as evident from the sizable variation in the interquartile range of the box plots. Already two decades ago, correlations between activity counts and oxygen consumption were reported for rest/structured activities and free play (r = 0.82 and r = 0.66, respectively) in 3–5-year old children [13], whereupon stationary behaviours, also termed ‘physical inactivity’ became recognised as a distinct construct [60]. The Children’s PA Research Group at the University of South Carolina has made significant contributions to better understand PA behaviours in children [21,63]. Among those, most notably the development of observational instruments to assess PA context-specific, and the implementation of accelerometry cut-off points in settings of different age groups. The overlap of activity counts for stationary behaviours implies that non-stationary PA can relatively clearly defined in future large-scale data-driven approaches. Cut-off points can be applied in longitudinal actigraphic data collected in sleep research and PA analysed while sleep is treated as a different vigilance state than quiet awake. In the future, PA and sleep analysis of actigraphic/accelerometer population data should be combined as movement-based 24-h behaviour. The correlation between PA activity and oxygen consumption [13] justifies to address circumstances and conditions surrounding low PA activities, for example the transition from home to school or living in city, urban or rural areas. Activity monitoring can be helpful for identifying children in need of improving their fitness and tailoring interventions for both, sleep and PA with staff and family support.

The current study adopted the term ‘motionless alert’ instead of using ‘rest/resting’ because the term ‘rest’ is inconsistently used [64]. In chronobiology, ‘rest’ covers ‘sleep’, including sleep-defining stages ‘time-in-bed’ ‘latency to sleep’, and ‘sleep period’ (motionless body with an active brain). The Sedentary Behaviour Research Network has published consensus definitions to standardise terminology for sedentary behaviours using ‘posture’ information and ‘energy expenditure’ [65]. They also adopted the term ‘stationary’ and suggested to apply ‘behaviour’ when the context is known, and ‘time’ in the absence of context, which is a sensible linguistic differentiation. We advocate to use terminology that is applicable not only for PA but for different disciplines such as chronobiology, sleep and epidemiology.

An important objective of observing the behaviour while measuring movement intensities was to test the discrimination power for ‘motionless alert’ behaviour, such as watching a cartoon, from all other classes. Although acceptable, ‘motionless alert’ from the wrist-worn MW8 and GT3X cut-offs showed only fair sensitivity (75% and 77%), while sensitivity was even lower for hip-worn cut-off points. The better outcome of the mobile PA classes over stationary PA classes can be explained by greater density of counts with much narrower ranges due to more steady, continuous activity during ‘sprinting, ‘floor ball’ and ‘playing on the floor’. In comparison, the stationary PA classes, and motionless alert in particular, included children engaged in mental tasks (listening, concentrating, observing) with the occasional, discontinued burst of movements while sitting still, for example when pointing or changing of position. Whether distinguishing the different stationary behaviours for time spent in low PA per day is useful can be questioned, but distinguishing motionless alert from sleep is necessary. Visual inspection, time-stamp markers or diary entries (night sleep, nap) have been proven to work in chronobiology and sleep science.

Sensory modes and body positions

In direct comparison, the two device models systematically scaled count quantities differently irrespective of placement (wrist/hip). Since all manufacturers calibrate their models at factory for optimal recording performance suited for certain body position, there is an expected imbalance in counts related to body positions [40,66]. For example, the MW8 offers different recording options such as uni- or triaxial mode, specific for wrist movements or body movements, respectively. Its uni-axial mode has been calibrated for wrist position, making use of omni-rotary arm movements [67], and sensitivity thresholds were determined and validated for sleep analyses algorithms (S1 Table). Its tri-axial mode produces VM/epochs that is best suited for the trunk, i.e., the hip, but not for measuring sleep nor light [54]. We mounted a second MW8 in uni-axial mode at the hip position to see how the same mode scales wrist against hip position, accepting its compromised status of reduced resolution. Lower activity counts were found for all behavioural activities at hip position compared to the wrist not only for MW8 but also GT3X. Differences in intensity output using two raw (with bandpass filters) acceleration metrics (Mean amplitude deviation, Movement acceleration intensity) also showed higher values for the wrist position compared to hip position when measured simultaneously in children 8–13 years old [68]. The GT3X model, which was set to tri-axial mode as recommended, generated approximately 6 to 10 times higher values in VM mode than the uni-axial MW8 mode, regardless of position. The challenge with recording and processing raw 3-dimensonal longitudinal data is their volume - in the TeraByte range - and the need to develop approaches on how to visualise 3-dimensional patterns, most summarise them into a single vector/epoch [69]. Taken together, users need to be aware of these facts and decide accordingly on sensor modes, pre-filter/raw settings and algorithms when implementing activity monitoring into longitudinal recordings.

Within model comparison of cut-off points

The MW8 has previously been validated to measure PA in older adults (S1 Table) [47] and in 9–13 year old children [49]. Lin et al.’s (45) protocol differed from our study in that they captured data from the dominant wrist position of older children. We preferred the non-dominant wrist position because it has been shown to produce less misclassification for sedentary behaviours [42]. Despite the differences, our results are in line for mobile PA activities with Lin et al. [49], who reported a good ability of the device to differentiate light from moderate-to-vigorous activity (MVPA >371.5 counts/30 s), and moderate from vigorous activity (VPA >859.5 counts/30 s). Our boundaries for VPA is a little lower (MVPA > 408 counts/30s; VPA >787 counts/30s), likely as a result of age-related differences in motor development and muscle strength. In Johansson et al. [12], authors argued for age-specific thresholds since they found that 4-year-old preschool children, who wore a GT3X triaxial on their non-dominant wrist, produced generally higher intensity thresholds per 5s-epoch in comparison to toddlers of 2 years [19]. Differentiation of sedentary from light activity was much greater in the study by Lin et al. (45), given the much lower sedentary threshold score (SED < 32 counts/30s) compared to the present study (SED < 215 counts/30s, MOA < 118 counts/30s). This is likely related to Lin et al. (45) prioritizing higher specificity at the cost of lower sensitivity in order to minimise false positives due to arm movements. Johansson et al. [12] used the opposite adjustment - towards higher sensitivity at the expense of a lower specificity for the SED threshold in order to avoid underestimating the time spent in SED. In the present study we used a step-approach with the upper boundary for SED being the lower boundary of mobile PA classes against all combined stationary PA classes, which includes arm movements, i.e. when sitting and doing crafts. All threshholds were reported as the greatest sum of the sensitivity and specificity, without prioritising specificity nor sensitivity.

Strength and limitations

One of the strengths is the comparative study design that included two research-grade devices (MW8, GT3X) worn in parallel at two body positions (wrist and hip) in a young age group (3 years). This allowed us to point out distributional regularities, similarities and differences for the devices and positions, knowledge that is important to consider when planning a study or when comparing data collected with these devices. In addition, the use of a 30 second epoch matches the calibration of MW8 devices for sleep assessments and meets the requirements for longitudinal, long-term actigraphic data collection for circadian rhythm/sleep assessments. The present calibration allows PA analysis from measurements of physical activity in combination with circadian/sleep assessments. The purpose of it is a multi-modal outcome with relevant information for trajectories of 24-h sleep and PA sufficiency to be comparable and interpretable with other studies and contribute to charts of 24h sleep and PA percentile curves. Half a minute may seem very long for sport activities but human behaviours under real situations last for several minutes, in particular conversations, meal and play times, sedentary activities like handcrafts, drawing or screen time. On the other hand, vigorous activities over the sequence of 10 minutes were not uninterrupted. Instead, there were periods of short light level movements.

Another strengths is the application of a commonly used classification method on directly observed behaviours to quantify intensity levels from which count cut-off points were determined. The intensity thresholds were chosen based on the best compromise between sensitivity and specificity, since prioritising sensitivity for ‘physically stationary’ may inflate sedentary time, when it is actually time spent in light mobile PA. Prioritsing specificity for stationary PA may overestimate ‘physically mobile’ and underestimate time spent sedentary. Our approach is a compromise as are those of others [5]. Our calibration may be applicable to existing population datasets since it allows to derive scaling coefficients. When other studies used 5 second epochs these can be re-scaled to match the 30s epoch for comparison, but this conversion needs verification as deliniated by Orme et al. [70]. The dataset generated during the current study is freely available under figshare.

The absence of indirect calorimetry [71] could be seen as a limitation, but oxygen consumption has been measured indirectly in 4-year old preschool children and activity intensity levels at free play correlated well with oxygen consumption, yet it has been discussed to have limitations due to the delayed rate of change with a change in activity levels [13]. We are aware that our thresholds have not undergone cross-validation and all of the observed activities were performed indoors, although resembling everyday habits of a three-year-old child. Outdoor activities were deliberately left out because of the high seasonal variability in outdoor activities in Northern Sweden. Further work is required to examine season-based behavioural outdoor activity patterns.

Conclusions

In conclusion, MotionWatch 8 and the ActiGraph GT3X underwent calibration processing simultaneously on the non-dominant wrist and hip based on direct observation of six naturally-occuring behavioural activities in healthy, 3-year-old children. ROC-AUC curves revealed cut-off points with outstanding (Sprinting, Floorball/Walk and Playing on the floor) and excellent (Sedentary screen time) discrimination power. The accuracy indicates that the calibrated cut-off points can be adequately used to determine time spent in different activity levels. When applying the cut-off points, five calibration factors need to be matched (i) age(ii) the device-specific sensor mode(iii) epoch (scaling coefficient, if different)(iv) body position, and (v) behaviours equivalent to occurring in natural settings. The dataset is freely available.. It is hoped to be useful for the community interested in a 24-hour movement-related approach, integrating PA intensities into projects on sleep, circadian rhythms and light exposure.

Supporting information

S1 Table. Highlighting some of the different approaches taken from sports medicine and chronobiology/sleep, with references of published examples (References Supplementary to S1 Table: 42 articles).

(DOCX)

pone.0316747.s001.docx^{(124.5KB, docx)}

S2 Table. Receiver operating characteristics curve (ROC) analysis in One-vs-One scheme, comparing pairwise combination of adjacent behavioural classes as described in Fig 2a.

Cut-off values, sensitivity, specificity and accuracy (AUC) are reported from ‘vigorous’ to ‘motionless alert’. ‘Sedentary crafts’ and ‘Recumbent listening’ were merged into ‘Sedentary (active)’. ‘Sedentary screen time is categorised as ‘Motionless alert’.

(DOCX)

pone.0316747.s002.docx^{(44.8KB, docx)}

S3 Fig. Duration in minutes of cumulated wear-time of devices (MW8, GT3X) simultaneously worn at wrist and hip positions while performing each of the six behaviours.

The epoch for storing the activity counts was 0.5 min.

(DOCX)

pone.0316747.s003.docx^{(69.6KB, docx)}

S4 Fig. ROC curves for the wrist-worn Motionwatch 8.

OvR ROC curves illustrating the results in Tables 3 and 4 in the main text for: A) vigorous physical activity (VPA), B) moderate-vigorous physical activity (MVPA), C) light moderate vigorous physical activity (LMVPA) and D) motionless-alert (MOA).

(DOCX)

pone.0316747.s004.docx^{(146KB, docx)}

S5 Fig. ROC curves for wrist-worn GT3X.

(DOCX)

pone.0316747.s005.docx^{(144.3KB, docx)}

S6 Fig. ROC curves for hip-worn Motionwatch 8.

(DOCX)

pone.0316747.s006.docx^{(148.3KB, docx)}

S7 Fig. ROC curves for hip-worn GT3X.

(DOCX)

pone.0316747.s007.docx^{(198.8KB, docx)}

Acknowledgments

The authors thank all children and their parents for their participation. Also, we thank Anna Tellström and Rebecca Rönnholm for their great assistance in the data collection; Tobias Stenlund for all organisational help with the e-health laboratory; Patrik Wennberg for lending us the two Actigraph GT3X devices and the accompanying temporary license; and Richard Lundberg for his administrative support with the NorthPop database.

Data Availability

Sources for raw data and manuals: RAW DATA EXCEL SPREADSHEET 10.6084/m9.figshare.27896760 ACTIGRAPHY OPERATIONAL MANUAL https://www.katlab.org/wp-content/uploads/2023/10/ACTIGRAPHY-OPERATIONAL-MANUAL-Nordic-Daylight-Research-Programme-2023.pdf

Funding Statement

We thank for the financial support from the Swedish Research Council (Vetenskapsrådet grant number 2019-01005 to MD), Region Västerbotten (ALF research infrastructure grant) and Umeå University research infrastructure grants to MD and CW, as well as a Särskild satsning grant from the Wallenberg centrum för molekylär medicin (WCMM, proj no: FS 2.1.6-849-209) to KW. The work of KW was partially supported by the Knut and Alice Wallenberg Foundation. The work of DJ was funded by institutional funds from the department of community medicine and rehabilitation, Umeå University. None of the funders played any role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.

References

1.Bull FC, Al-Ansari SS, Biddle S, Borodulin K, Buman MP, Cardon G, et al. World Health Organization 2020 guidelines on physical activity and sedentary behaviour. Br J Sports Med. 2020;54(24):1451–62. doi: 10.1136/bjsports-2020-102955 [DOI] [PMC free article] [PubMed] [Google Scholar]
2.Hammad G, Reyt M, Beliy N, Baillet M, Deantoni M, Lesoinne A, et al. pyActigraphy: open-source python package for actigraphy data visualization and analysis. PLoS Comput Biol. 2021;17(10):e1009514. doi: 10.1371/journal.pcbi.1009514 [DOI] [PMC free article] [PubMed] [Google Scholar]
3.Weitz M, Syed S, Hopstock LA, Morseth B, Prasad DK, Horsch A. Discrimination of sleep and wake periods from a hip-worn raw acceleration sensor using recurrent neural networks. medRxiv. 2022. p. 2022.03.07.22270992. doi: 10.1101/2022.03.07.22270992 [DOI] [Google Scholar]
4.Tremblay MS, Carson V, Chaput J-P, Connor Gorber S, Dinh T, Duggan M, et al. Canadian 24-hour movement guidelines for children and youth: an integration of physical activity, sedentary behaviour, and sleep. Appl Physiol Nutr Metab. 2016;41(6 Suppl 3):S311–27. doi: 10.1139/apnm-2016-0151 [DOI] [PubMed] [Google Scholar]
5.Rosenberger ME, Fulton JE, Buman MP, Troiano RP, Grandner MA, Buchner DM, et al. The 24-hour activity cycle: a new paradigm for physical activity. Med Sci Sports Exerc. 2019;51(3):454–64. doi: 10.1249/MSS.0000000000001811 [DOI] [PMC free article] [PubMed] [Google Scholar]
6.Chang Z, Lei W. A study on the relationship between physical activity, sedentary behavior, and sleep duration in preschool children. Front Public Health. 2021;9:618962. doi: 10.3389/fpubh.2021.618962 [DOI] [PMC free article] [PubMed] [Google Scholar]
7.Butte NF, Puyau MR, Wilson TA, Liu Y, Wong WW, Adolph AL, et al. Role of physical activity and sleep duration in growth and body composition of preschool-aged children. Obesity (Silver Spring). 2016;24(6):1328–35. doi: 10.1002/oby.21489 [DOI] [PMC free article] [PubMed] [Google Scholar]
8.Møller NC, Christensen LB, Mølgaard C, Ejlerskov KT, Pfeiffer KA, Michaelsen KF. Descriptive analysis of preschool physical activity and sedentary behaviors – a cross sectional study of 3-year-olds nested in the SKOT cohort. BMC Public Health. 2017;17(1):613. doi: 10.1186/s12889-017-4521-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
9.World Health Organization. Report of the commission on ending childhood obesity. Geneva, Switzerland: World Health Organization [Internet]; 2016. Available from: https://www.who.int/publications/i/item/9789241510066 [Google Scholar]
10.Okely AD, Ghersi D, Hesketh KD, Santos R, Loughran SP, Cliff DP, et al. A collaborative approach to adopting/adapting guidelines - the Australian 24-hour movement guidelines for the early years (birth to 5 years): an integration of physical activity, sedentary behavior, and sleep. BMC Public Health. 2017;17(Suppl 5):869. doi: 10.1186/s12889-017-4867-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
11.Trost SG, Cliff DP, Ahmadi MN, Tuc NV, Hagenbuchner M. Sensor-enabled activity class recognition in preschoolers: hip versus wrist data. Med Sci Sports Exerc. 2018;50(3):634–41. doi: 10.1249/MSS.0000000000001460 [DOI] [PubMed] [Google Scholar]
12.Johansson E, Larisch L-M, Marcus C, Hagströmer M. Calibration and validation of a wrist- and hip-worn actigraph accelerometer in 4-year-old children. PLoS One. 2016;11(9):e0162436. doi: 10.1371/journal.pone.0162436 [DOI] [PMC free article] [PubMed] [Google Scholar]
13.Pate RR, Almeida MJ, McIver KL, Pfeiffer KA, Dowda M. Validation and calibration of an accelerometer in preschool children. Obesity (Silver Spring). 2006;14(11):2000–6. doi: 10.1038/oby.2006.234 [DOI] [PubMed] [Google Scholar]
14.van Cauwenberghe E, Labarque V, Trost SG, de Bourdeaudhuij I, Cardon G. Calibration and comparison of accelerometer cut points in preschool children. Int J Pediatr Obes. 2011;6(2–2):e582-9. doi: 10.3109/17477166.2010.526223 [DOI] [PubMed] [Google Scholar]
15.Roscoe CMP, James RS, Duncan MJ. Calibration of GENEActiv accelerometer wrist cut-points for the assessment of physical activity intensity of preschool aged children. Eur J Pediatr. 2017;176(8):1093–8. doi: 10.1007/s00431-017-2948-2 [DOI] [PubMed] [Google Scholar]
16.Pfeiffer KA, McIver KL, Dowda M, Almeida MJCA, Pate RR. Validation and calibration of the Actical accelerometer in preschool children. Med Sci Sports Exerc. 2006;38(1):152–7. doi: 10.1249/01.mss.0000183219.44127.e7 [DOI] [PubMed] [Google Scholar]
17.Janssen X, Cliff DP, Reilly JJ, Hinkley T, Jones RA, Batterham M, et al. Validation and calibration of the activPAL™ for estimating METs and physical activity in 4-6 year olds. J Sci Med Sport. 2014;17(6):602–6. doi: 10.1016/j.jsams.2013.10.252 [DOI] [PubMed] [Google Scholar]
18.Ahmadi MN, Chowdhury A, Pavey T, Trost SG. Laboratory-based and free-living algorithms for energy expenditure estimation in preschool children: a free-living evaluation. PLoS ONE. 2020;15(5):e0233229. doi: 10.1371/journal.pone.0233229 [DOI] [PMC free article] [PubMed] [Google Scholar]
19.Johansson E, Ekelund U, Nero H, Marcus C, Hagströmer M. Calibration and cross-validation of a wrist-worn Actigraph in young preschoolers. Pediatr Obes. 2015;10(1):1–6. doi: 10.1111/j.2047-6310.2013.00213.x [DOI] [PubMed] [Google Scholar]
20.Costa S, Barber SE, Cameron N, Clemes SA. Calibration and validation of the ActiGraph GT3X+ in 2-3 year olds. J Sci Med Sport. 2014;17(6):617–22. doi: 10.1016/j.jsams.2013.11.005 [DOI] [PubMed] [Google Scholar]
21.Pate RR, Dishman RK, Dowda M, McIver KL, Pfeiffer KA, Porter DE, et al. A summary of one research team’s contributions to understanding physical activity behavior in children and youth. Int J Environ Res Public Health. 2022;19(21):14136. doi: 10.3390/ijerph192114136 [DOI] [PMC free article] [PubMed] [Google Scholar]
22.Van Cauwenberghe E, Gubbels J, De Bourdeaudhuij I, Cardon G. Feasibility and validity of accelerometer measurements to assess physical activity in toddlers. Int J Behav Nutr Phys Act. 2011;8:67. doi: 10.1186/1479-5868-8-67 [DOI] [PMC free article] [PubMed] [Google Scholar]
23.Dolinsky DH, Brouwer RJN, Evenson KR, Siega-Riz AM, Østbye T. Correlates of sedentary time and physical activity among preschool-aged children. Prev Chronic Dis. 2011;8(6):A131. [PMC free article] [PubMed] [Google Scholar]
24.Timmons BW, Leblanc AG, Carson V, Connor Gorber S, Dillman C, Janssen I, et al. Systematic review of physical activity and health in the early years (aged 0-4 years). Appl Physiol Nutr Metab. 2012;37(4):773–92. doi: 10.1139/h2012-070 [DOI] [PubMed] [Google Scholar]
25.Liu F, Schrack J, Wanigatunga SK, Rabinowitz JA, He L, Wanigatunga AA, et al. Comparison of sleep parameters from wrist-worn ActiGraph and Actiwatch devices. Sleep. 2024;47(2):zsad155. doi: 10.1093/sleep/zsad155 [DOI] [PMC free article] [PubMed] [Google Scholar]
26.Schoch SF, Jenni OG, Kohler M, Kurth S. Actimetry in infant sleep research: an approach to facilitate comparability. Sleep. 2019;42(7):zsz083. doi: 10.1093/sleep/zsz083 [DOI] [PMC free article] [PubMed] [Google Scholar]
27.Chandler JL, Brazendale K, Beets MW, Mealing BA. Classification of physical activity intensities using a wrist-worn accelerometer in 8-12-year-old children. Pediatr Obes. 2016;11(2):120–7. doi: 10.1111/ijpo.12033 [DOI] [PubMed] [Google Scholar]
28.Hislop J, Palmer N, Anand P, Aldin T. Validity of wrist worn accelerometers and comparability between hip and wrist placement sites in estimating physical activity behaviour in preschool children. Physiol Meas. 2016;37(10):1701–14. doi: 10.1088/0967-3334/37/10/1701 [DOI] [PubMed] [Google Scholar]
29.Fairclough SJ, Noonan R, Rowlands AV, Van Hees V, Knowles Z, Boddy LM. Wear compliance and activity in children wearing wrist- and hip-mounted accelerometers. Med Sci Sports Exerc. 2016;48(2):245–53. doi: 10.1249/MSS.0000000000000771 [DOI] [PubMed] [Google Scholar]
30.Arvidsson D, Fridolfsson J, Börjesson M, Andersen LB, Ekblom Ö, Dencker M, et al. Re-examination of accelerometer data processing and calibration for the assessment of physical activity intensity. Scand J Med Sci Sports. 2019;29(10):1442–52. doi: 10.1111/sms.13470 [DOI] [PubMed] [Google Scholar]
31.Migueles JH, Aadland E, Andersen LB, Brønd JC, Chastin SF, Hansen BH, et al. GRANADA consensus on analytical approaches to assess associations with accelerometer-determined physical behaviours (physical activity, sedentary behaviour and sleep) in epidemiological studies. Br J Sports Med. 2022;56(7):376–84. doi: 10.1136/bjsports-2020-103604 [DOI] [PMC free article] [PubMed] [Google Scholar]
32.Montoye AHK, Westgate BS, Fonley MR, Pfeiffer KA. Cross-validation and out-of-sample testing of physical activity intensity predictions with a wrist-worn accelerometer. J Appl Physiol (1985). 2018;124(5):1284–93. doi: 10.1152/japplphysiol.00760.2017 [DOI] [PubMed] [Google Scholar]
33.Troiano RP, McClain JJ, Brychta RJ, Chen KY. Evolution of accelerometer methods for physical activity research. Br J Sports Med. 2014;48(13):1019–23. doi: 10.1136/bjsports-2014-093546 [DOI] [PMC free article] [PubMed] [Google Scholar]
34.Wulff K, Dedek A, Siegmund R. Circadian and Ultradian time patterns in human behavior: Part 2: social synchronisation during the development of the infant’s diurnal activity-rest pattern. Biol Rhythm Res. 2001;32(5):529–46. doi: 10.1076/brhm.32.5.529.1295 [DOI] [Google Scholar]
35.Wulff K, Siegmund R. Time pattern analysis of activity-rest rhythms in families with infants using actigraphy. In: Salzarulo P, Ficca G, editors. Awakening and Sleep–Wake Cycle Across Development. John Benjamins Publishing Company; 2002. p. 149–69. [Google Scholar]
36.Freedson PS, John D. Comment on “estimating activity and sedentary behavior from an accelerometer on the hip and wrist”. Med Sci Sports Exerc. 2013;45(5):962–3. doi: 10.1249/MSS.0b013e31827f024d [DOI] [PubMed] [Google Scholar]
37.Jeong D-H, Kim S-E, Choi W-H, Ahn S-H. A comparative study on the influence of undersampling and oversampling techniques for the classification of physical activities using an imbalanced accelerometer dataset. Healthcare (Basel). 2022;10(7):1255. doi: 10.3390/healthcare10071255 [DOI] [PMC free article] [PubMed] [Google Scholar]
38.Cellini N, McDevitt EA, Mednick SC, Buman MP. Free-living cross-comparison of two wearable monitors for sleep and physical activity in healthy young adults. Physiol Behav. 2016;157:79–86. doi: 10.1016/j.physbeh.2016.01.034 [DOI] [PMC free article] [PubMed] [Google Scholar]
39.Van Remoortel H, Giavedoni S, Raste Y, Burtin C, Louvaris Z, Gimeno-Santos E, et al. Validity of activity monitors in health and chronic disease: a systematic review. Int J Behav Nutr Phys Act. 2012;9:84. doi: 10.1186/1479-5868-9-84 [DOI] [PMC free article] [PubMed] [Google Scholar]
40.Migueles JH, Cadenas-Sanchez C, Ekelund U, Delisle Nyström C, Mora-Gonzalez J, Löf M, et al. Accelerometer data collection and processing criteria to assess physical activity and other outcomes: a systematic review and practical considerations. Sports Med. 2017;47(9):1821–45. doi: 10.1007/s40279-017-0716-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
41.Montoye AHK, Pivarnik JM, Mudd LM, Biswas S, Pfeiffer KA. Comparison of activity type classification accuracy from accelerometers worn on the hip, wrists, and thigh in young, apparently healthy adults. Meas Phys Educ Exerc Sci. 2016;20(3):173–83. doi: 10.1080/1091367x.2016.1192038 [DOI] [Google Scholar]
42.Gao Z, Liu W, McDonough DJ, Zeng N, Lee JE. The dilemma of analyzing physical activity and sedentary behavior with wrist accelerometer data: challenges and opportunities. J Clin Med. 2021;10(24):5951. doi: 10.3390/jcm10245951 [DOI] [PMC free article] [PubMed] [Google Scholar]
43.Ancoli-Israel S, Cole R, Alessi C, Chambers M, Moorcroft W, Pollak CP. The role of actigraphy in the study of sleep and circadian rhythms. Sleep. 2003;26(3):342–92. doi: 10.1093/sleep/26.3.342 [DOI] [PubMed] [Google Scholar]
44.Maczák B, Vadai G, Dér A, Szendi I, Gingl Z. Detailed analysis and comparison of different activity metrics. PLoS One. 2021;16(12):e0261718. doi: 10.1371/journal.pone.0261718 [DOI] [PMC free article] [PubMed] [Google Scholar]
45.Skeldon AC, Dijk D-J, Meyer N, Wulff K. Extracting circadian and sleep parameters from longitudinal data in schizophrenia for the design of pragmatic light interventions. Schizophr Bull. 2022;48(2):447–56. doi: 10.1093/schbul/sbab124 [DOI] [PMC free article] [PubMed] [Google Scholar]
46.Danilenko KV, Stefani O, Voronin KA, Mezhakova MS, Petrov IM, Borisenkov MF, et al. Wearable light-and-motion dataloggers for sleep/wake research: a review. Appl Sci. 2022;12(22):11794. doi: 10.3390/app122211794 [DOI] [Google Scholar]
47.Falck RS, Best JR, Li MCR, Eng JJ, Liu-Ambrose T. Revisiting the MotionWatch8©: calibrating cut-points for measuring physical activity and sedentary behavior among adults with stroke. Front Aging Neurosci. 2019;11:203. doi: 10.3389/fnagi.2019.00203 [DOI] [PMC free article] [PubMed] [Google Scholar]
48.Landry GJ, Falck RS, Beets MW, Liu-Ambrose T. Measuring physical activity in older adults: calibrating cut-points for the MotionWatch 8©. Front Aging Neurosci. 2015;7. [DOI] [PMC free article] [PubMed] [Google Scholar]
49.Lin H-P, Lynk N, Moore LL, Cabral HJ, Heffernan KS, Dumas AK, et al. A pragmatic approach to the comparison of wrist-based cutpoints of physical activity intensity for the MotionWatch8 accelerometer in children. PLoS One. 2020;15(6):e0234725. doi: 10.1371/journal.pone.0234725 [DOI] [PMC free article] [PubMed] [Google Scholar]
50.Kushida CA, Chang A, Gadkary C, Guilleminault C, Carrillo O, Dement WC. Comparison of actigraphic, polysomnographic, and subjective assessment of sleep parameters in sleep-disordered patients. Sleep Med. 2001;2(5):389–96. doi: 10.1016/s1389-9457(00)00098-8 [DOI] [PubMed] [Google Scholar]
51.Cole RJ, Kripke DF, Gruen W, Mullaney DJ, Gillin JC. Automatic sleep/wake identification from wrist activity. Sleep. 1992;15(5):461–9. doi: 10.1093/sleep/15.5.461 [DOI] [PubMed] [Google Scholar]
52.Sadeh A, Sharkey KM, Carskadon MA. Activity-based sleep-wake identification: an empirical test of methodological issues. Sleep. 1994;17(3):201–7. doi: 10.1093/sleep/17.3.201 [DOI] [PubMed] [Google Scholar]
53.Wikland KA, Luo ZC, Niklasson A, Karlberg J. Swedish population-based longitudinal reference values from birth to 18 years of age for height, weight and head circumference. Acta Paediatr. 2002;91(7):739–54. doi: 10.1080/08035250213216 [DOI] [PubMed] [Google Scholar]
54.Camntech. The MotionWatch and MotionWare user guide. Version 1226a. 2019. [Google Scholar]
55.Powers D. Evaluation: from precision, recall and F-measure to ROC, Informedness, Markedness and correlation. J Machine Learn Technol. 2011;2:37–63. [Google Scholar]
56.Metz CE. Basic principles of ROC analysis. Semin Nucl Med. 1978;8(4):283–98. doi: 10.1016/s0001-2998(78)80014-2 [DOI] [PubMed] [Google Scholar]
57.Pedregosa F, Varoquaux G, Gramfort A, Michel V, Thirion B, Grisel O. Scikit-learn: machine learning in python. J Machine Learn Res. 2011. [Google Scholar]
58.Akobeng AK. Understanding diagnostic tests 3: receiver operating characteristic curves. Acta Paediatr. 2007;96(5):644–7. doi: 10.1111/j.1651-2227.2006.00178.x [DOI] [PubMed] [Google Scholar]
59.Hosmer DW Jr, Lemenshow S, Sturdivant RX. Assessing the Fit of the Model. Applied Logistic Regression. John Wiley & Sons, Ltd; 2000. p. 143–202. [Google Scholar]
60.Reilly JJ, Coyle J, Kelly L, Burke G, Grant S, Paton JY. An objective method for measurement of sedentary behavior in 3- to 4-year olds. Obes Res. 2003;11(10):1155–8. doi: 10.1038/oby.2003.158 [DOI] [PubMed] [Google Scholar]
61.DuRant RH, Baranowski T, Puhl J, Rhodes T, Davis H, Greaves KA, et al. Evaluation of the Children’s Activity Rating Scale (CARS) in young children. Med Sci Sports Exerc. 1993;25(12):1415–21. doi: 10.1249/00005768-199312000-00016 [DOI] [PubMed] [Google Scholar]
62.Finn KJ, Specker B. Comparison of Actiwatch activity monitor and children’s activity rating scale in children. Med Sci Sports Exerc. 2000;32(10):1794–7. doi: 10.1097/00005768-200010000-00021 [DOI] [PubMed] [Google Scholar]
63.Brown WH, Pfeiffer KA, Mclver KL, Dowda M, Almeida MJCA, Pate RR. Assessing preschool children’s physical activity: the Observational System for Recording Physical Activity in children-preschool version. Res Q Exerc Sport. 2006;77(2):167–76. doi: 10.1080/02701367.2006.10599351 [DOI] [PubMed] [Google Scholar]
64.Bernhofer EI. Investigating the concept of rest for research and practice. J Adv Nurs. 2016;72(5):1012–22. doi: 10.1111/jan.12910 [DOI] [PubMed] [Google Scholar]
65.Tremblay MS, Aubert S, Barnes JD, Saunders TJ, Carson V, Latimer-Cheung AE, et al. Sedentary Behavior Research Network (SBRN) - Terminology Consensus Project process and outcome. Int J Behav Nutr Phys Act. 2017;14(1):75. doi: 10.1186/s12966-017-0525-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
66.Brønd JC, Arvidsson D. Sampling frequency affects the processing of Actigraph raw acceleration data to activity counts. J Appl Physiol (1985). 2016;120(3):362–9. doi: 10.1152/japplphysiol.00628.2015 [DOI] [PubMed] [Google Scholar]
67.Cyriax EF. On the rotary movements of the wrist. J Anat. 1926;60(Pt 2):199–201. [PMC free article] [PubMed] [Google Scholar]
68.Beck F, Marzi I, Eisenreich A, Seemüller S, Tristram C, Reimers AK. Determination of cut-off points for the Move4 accelerometer in children aged 8-13 years. BMC Sports Sci Med Rehabil. 2023;15(1):163. doi: 10.1186/s13102-023-00775-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
69.Tsanas A. Investigating wrist-based acceleration summary measures across different sample rates towards 24-hour physical activity and sleep profile assessment. Sensors (Basel). 2022;22(16):6152. doi: 10.3390/s22166152 [DOI] [PMC free article] [PubMed] [Google Scholar]
70.Orme M, Wijndaele K, Sharp SJ, Westgate K, Ekelund U, Brage S. Combined influence of epoch length, cut-point and bout duration on accelerometry-derived physical activity. Int J Behav Nutr Phys Act. 2014;11(1):34. doi: 10.1186/1479-5868-11-34 [DOI] [PMC free article] [PubMed] [Google Scholar]
71.Kim Y, Beets MW, Welk GJ. Everything you wanted to know about selecting the “right” Actigraph accelerometer cut-points for youth, but…: a systematic review. J Sci Med Sport. 2012;15(4):311–21. doi: 10.1016/j.jsams.2011.12.001 [DOI] [PubMed] [Google Scholar]

PLoS One. doi: 10.1371/journal.pone.0316747.r001

Decision Letter 0

Duncan S Buchan

9 Jun 2024

PONE-D-24-07355Behaviour-based movement cut-off points in 3-year old children comparing wrist- with hip-worn actigraphs MW8 and GT3XPLOS ONE

Dear Dr. Wulff,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript by Jul 24 2024 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org . When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.
A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.
An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols . Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols .

We look forward to receiving your revised manuscript.

Kind regards,

Duncan S Buchan

Academic Editor

PLOS ONE

Journal requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. Thank you for uploading your study's underlying data set. Unfortunately, the repository you have noted in your Data Availability statement does not qualify as an acceptable data repository according to PLOS's standards.

At this time, please upload the minimal data set necessary to replicate your study's findings to a stable, public repository (such as figshare or Dryad) and provide us with the relevant URLs, DOIs, or accession numbers that may be used to access these data. For a list of recommended repositories and additional information on PLOS standards for data deposition, please see https://journals.plos.org/plosone/s/recommended-repositories.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Partly

Reviewer #2: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: Dear Editor, Dear Author,

The article "Behaviour-based movement cut-off points in 3-year old children comparing 4 wrist- with hip-worn actigraphs MW8 and GT3X" is a study focusing on calibration of activity counts of motor behavior measures simultaneously with two devices.

Thank you for the opportunity to review this manuscript. The article deals with a relevant and contemporary issue and is based on a good scientific methodological quality. The research question at hand and the methodological approach are discussed comprehensively. Nevertheless, I have some concerns that should be addressed by the authors before publication.

General

- Please use PA instead of physical activity throughout the whole manuscript, since you introduce the abbreviation at the beginning.

- Please improve quality of all figures.

Abstract

1. Line 46, page 2: What is meant with rigorous calibration? Is this something different than just a calibration?

2. Line 52, page 2: Are the six activities allocated to different intensities? Please specify.

3. Line53, page 2: I wonder if the the time of each activity is too long, especially for vigorous intensities when participants are three year old? In Line 53 (page 2) you also mentioned sprinting about 10 minutes. Are 3-year old children able to do this for 10 minutes?

4. Line 54, page 2: What do you mean with directly observed? You mentioned this the term but we wonder what exactly is meant with this and couldn’t find any more information in the manuscript neither in the results section, nor in the discussion.

5. Line 61, page 2: Why classifying into mobile and stationary? Maybe you can add one sentence to clarify this already in the abstract.

6. Line 65, page 2: I wonder what you mean with context information. Please specify.

Introduction

General:

- Throughout the whole introduction, sentences are very long. I would recommend, separating long sentences (especially Line 78-84, page 3; Line 129-136, page 5).

- The introduction is very long. I would recommend shorten it. The history of accelerometers is not that important in my opinion (Line 73, page 3). Further, some information could be more summarized. Please critically check the introduction for shortening.

1. Line 83, page 3: World Health Organization (WHO)

2. Line 140, page 5: What is the intention to assess context-related movements and distinct between mobile and stationary? From a public health perspective, it is necessary and more common to focus on intensities, as especially MVPA leads to health benefits and LPA does not impact health status in that way.

Method

1. Page 6, Line 150: Healthy 3-year-old children. What does that mean? Without disabilities? Without illness? I would recommend rephrasing the sentence in a Subject-first language (3-year old children without…).

2. Page 7, Line 170: Did you use expected MET-values? How did the classification of the activities to the different intensities took place?

3. Page 7, Line 181: On which basis did you chose the activities?

4. Page 8, Line 197: I wonder why you didn’t validate your data with the heart rate when you have the data anyway? Did you do something to validate the data? Or just the correlation between two similar devices?

5. Page 8, Line 211. Please remove the space between 250 g.

6. Page 9, Line 219: The raw data of the accelerometer are accelerations (g as unit). How do you get counts from the acceleration?

7. Page 9, Line 225: proprietary software – what is this? I am not sure if everybody know what this means. Please contextualize.

8. Page 9, Line 226: what is the behavioral observation protocol. Please describe this in more detail. This term is frequently in your manuscript but for me it is not sure what is meant with that.

9. Page 9, Line 231: Why didn’t you use test-retest to assess the reliability? How did you assess the reliability? Known studies often focus on validity and reliability in the context of calibration.

10. Page 10, Line 241: observer-based behaviors – what is this? Please contextualize.

11. Page 10, Line 246: I wonder why you make a distinction between mobility and stationary behavior when the common classification are the intensity levels.

12. Page 10, Line 252: on which basis did you do that? MET-values?

Results

1. Do you have the individual raw data/activity counts of each conducted activity?

2. Page 16, Table 3: Cut-off values per which second?

Discussion

1. General: Some paragraphs are difficult to understand. Maybe it would be helpful to add some subheadings to help the reader to go through the discussion (e.g., cut-off values, sensor position, …) see also (Beck et al., 2023).

2. Page 16, Line 382: Again, what is the observation technique? Are there no results concerning this?

3. Page 16, Line 383: Children’s activity rating scale --> in your study? The sentence is not clear to me.

4. Page 19, 390 ff: Is this also a phenomenon found in actual literature as you mention “Already two decades ago”?

5. Page 20, Line 397: You did not mention the qualitative-quantitative combination before. Please state this also in the methods section/abstract.

6. Page 20, Line 411- 422: In this section, you just describe your results. Please shorten this and state the main finding and discuss this.

7. Page 21, Line 430: Why didn’t you mention your motivation for this study at the beginning of the discussion?

8. Page 21, Line 430 ff: I suggest restructure this section. What was your main result? Further there are many statements not supported by literature (e.g. Since all manufacturers calibrate their models at factory for optimal recording performance suited for certain body position, there is an expected imbalance in counts related to body positions). Which implications does this have on your results?

9. Page 22, Line 443: Compare this with existing studies (e.g. Beck et al., 2023)

10. Page 22, Line 452 ff: I suggest restructure this section and firstly mention your results, then discussing. Further, in this section wearing positions (method) are discussed. It

11. Page 23, Line 475: Please add a subheading --> Strengths and Limitations

12. Page 23, Line 475: I would prefer restructuring the strength-section. Actually, you are not contextualizing your first strength - “One of the strengths is the study design that included two research-grade devices (MW8, GT3X) worn in parallel at two body positions (wrist and hip) in a young age group (3 years)” – why is this a strength of your study? Maybe you could pronounce this in one more sentence.

Reference:

Beck, F., Marzi, I., Eisenreich, A., Seemüller, S., Tristram, C., & Reimers, A. K. (2023). Determination of cut-off points for the Move4 accelerometer in children aged 8–13 years. BMC Sports Science, Medicine and Rehabilitation, 15(1). https://doi.org/10.1186/s13102-023-00775-4

Reviewer #2: A nicely conducted study. A few recommendations to improve its readability:

INTRODUCTION

• Line 72-73: Repeated use of "segregated" and "integrated" could be streamlined for clarity. SUGGESTED REVISION: "Movement-related assessments of habitual activities, such as physical activity (PA) levels or timing of sleep/circadian rhythms, have typically been segregated across disciplines like epidemiology, sports medicine, rehabilitation, and chronobiology [1–3]."

• Line 72-75: SUGGESTED REVISION: "Assessments of habitual activities like physical activity (PA) levels or sleep/circadian rhythms have often taken a segregated approach in fields such as epidemiology, sports medicine, rehabilitation, and chronobiology [1–3]. Historically, terminology also developed independently, with 'actimeter' or 'actigraph' used in sleep/chronobiology and 'accelerometer' in sport/physical activity (Tab S1)."

• Line 75: SUGGESTED REVISION: "Similarly, evidence regarding the combination of movement behaviours over a 24-hour period using compositional analyses [4] is uncommon but growing [5], with emerging studies [6–8]."

• Line 131: "https://www.katlab.org/ [under people]". SUGGESTED REVISION: "https://www.katlab.org/ [under 'people'], www.northpop.se"

• Context and Rationale: The introduction could benefit from a more explicit rationale for why comparing wrist- and hip-worn actigraphs is essential. While the text touches on different devices and algorithms, it should clarify the practical implications of these comparisons for assessing PA and sleep in children. i.e. explain why both wrist and hip placements are necessary and what specific insights are gained from comparing these placements.

• Research Gap: The introduction mentions the need for more studies but does not clearly state what specific gap this study aims to fill beyond general calibration. Clarify what unique aspect of children's activity measurement this study addresses.

• Detail on Current Standards and Practices: While the text mentions various guidelines and recommendations, it could provide a clearer connection between these guidelines and the specific challenges or limitations this study aims to overcome. For example, how current guidelines fail to account for the differences between wrist- and hip-worn devices in practical terms.

METHODS

• Participant Recruitment and Criteria: Lines 148-152: The recruitment criteria are clear, but additional context on the rationale for these criteria would be beneficial. Why specifically exclude children with chronic diseases or those outside the normative weight range?

• Lines 158-159: A brief mention of how the children's consent was obtained, beyond the legal guardians' consent, would be useful.

• Lines 166-167: A pilot is a crucial step for validating the methods, but the results of this pilot study are not discussed. A brief mention of any adjustments or findings from the pilot could enhance the credibility of the methods.

• Line 168-172: The description of the six behaviors is clear, but the introduction of "vigorous," "moderate," and "light" activity could be linked more explicitly to how these terms are operationalized in the study. Consider briefly defining these terms in this section for clarity.

• Lines 220-221: It would be helpful to reference specific studies or data supporting the Epoch Length choice.

STATISTICS

• Line 233: "… using a standard Students Unpaired T-test.". The choice of the Student's Unpaired T-test is standard for comparing means between two independent groups. However, if the sample size is small, a mention of checking assumptions of normality and equal variances (e.g., via Levene's test) would strengthen the methodological rigor.

• Line 235: "The two-tailed Pearson product-moment correlation was used ...". Pearson's correlation assumes linear relationships and normally distributed variables. Clarifying if these assumptions were tested and met would be beneficial. If not, a non-parametric correlation test (e.g., Spearman's rank correlation) might be more appropriate.

• Regression Models Lines 236-237: This statement lacks clarity on why different regression models are used for wrist vs. hip data. Justifying this choice with underlying data characteristics or preliminary analysis results would provide better context. Moreover, specifying the type of nonlinear regression model used (e.g., polynomial, exponential) would enhance transparency.

• Line 238: Using boxplots on a log scale is a good approach for skewed data. However, explaining why a log transformation is necessary (e.g., due to the skewed distribution of activity counts) would add clarity.

• Lines 240-250: The approach of using ROC curves to determine cut-off points is well-established. However, the explanation could be improved by: (1) Clarifying why a binarized approach is used and its advantages over multi-class ROC analysis. (2)Providing more details on how the specific behavioural categories were chosen and merged for the ROC analysis.

• Line 273: A brief rationale for choosing the Youden Index over other potential indices (e.g., F1 score, balanced accuracy) would be beneficial.

• Lines 277-282: The interpretation of ROC-AUC values is clearly stated, but ensuring that these values are consistently used throughout the analysis section will strengthen the overall analysis. Any variations or deviations from these interpretations should be noted and justified.

• There is no mention of a power analysis or justification of the sample size. Given that only 30 children were included, a brief explanation of how this sample size was determined to be adequate for the statistical tests used would be important.

• There is no discussion on how missing data were handled. Addressing whether there were any missing data points, and if so, how they were managed (e.g., imputation methods) would enhance the transparency and reliability of the results.

RESULTS

• The results section is thorough and well-structured.

DISCUSSION

• Aim and Context: (1) The aim of the study is clearly stated in the opening sentences (Lines 375-379). However, the text could be more concise, directly stating the key objectives without repetition. (2) Including a brief summary of the main findings at the beginning would help set the stage for the detailed discussion that follows.

• Comparison with Previous Studies: (1) The comparison with previous studies (Lines 379-387) is thorough but somewhat scattered. It would be beneficial to organize this section by specific themes or metrics (e.g., epoch length, device differences) to improve readability. (2) The explanation of why the thresholds in this study are higher due to the longer epoch length is informative but could be simplified for better clarity.

• Activity Counts and Behavior Distinction: (1) The discussion of overlap in activity counts among stationary activities (Lines 388-389) is clear, but it would be helpful to provide more context on the implications of this overlap for practical applications or further research. (2) The reference to historical data on correlations between activity counts and oxygen consumption (Lines 390-393) is useful, but the relevance to the current study should be made more explicit.

• Terminology and Definitions: (1) The section on terminology (Lines 398-407) is detailed but could be streamlined. The distinction between 'motionless alert' and 'rest' is important, but the explanation could be more concise. (2) The discussion on standardized definitions by the Sedentary Behaviour Research Network is relevant, but it would benefit from a clearer link to how these definitions were applied or interpreted in this study.

• Statistical Analysis and Accuracy- ROC-AUC and Sensitivity/Specificity: The explanation of why mobile PA classes showed better accuracy than stationary PA classes (Lines 423-428) is clear, but further elaboration on potential strategies to improve stationary PA classification could be beneficial.

• Device Comparison: (1) The comparison between MW8 and GT3X devices (Lines 430-451) is detailed but could be more structured. Breaking this section into sub-sections (e.g., device-specific performance, position-specific performance) would improve readability. (2) The discussion on the different calibration modes and their implications (Lines 436-450) is important but could be condensed for clarity.

CONCLUSION:

• The conclusion could be clearer by breaking down complex sentences. For example, the sentence starting with "The accuracy indicates..." (Lines 511-515) is quite dense and could be split for better readability.

• The mention of "ROC-AUC curves revealed cut-off points with outstanding, excellent and good discrimination power" (Lines 510-511) could be more specific. It would be beneficial to state which activities or behaviors correspond to each level of discrimination power.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean? ). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy .

Reviewer #1: No

Reviewer #2: No

**********

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/ . PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org . Please note that Supporting Information files do not need this step.

PLoS One. 2025 Mar 26;20(3):e0316747. doi: 10.1371/journal.pone.0316747.r002

Author response to Decision Letter 1

25 Nov 2024

RESPONSE TO REVIEWERS

PONE-D-24-07355

Behaviour-based movement cut-off points in 3-year old children comparing wrist- with hip-worn actigraphs MW8 and GT3X

We thank the editor and reviewers for their critical reading of the manuscript and their questions and comments to clarify phrases and procedures we described in the manuscript. Below, we have written a reply point by point, starting with +++ and changes we made in the manuscript according to the comments and suggestions are in track changes in the revised manuscript.

Reviewer #1: Dear Editor, Dear Author,

General

- Please use PA instead of physical activity throughout the whole manuscript, since you introduce the abbreviation at the beginning.

+++ We have replaced ‘physical activity with PA throughout since abbreviation was introduced. We replaced ‘physical’ with ‘motor’ in a sentence related to sleep.

- Please improve quality of all figures.

+++ We noticed that the figures in the PDF are of poorer quality than the original figures. We anticipate that the higher quality will remain in the type-setting process.

Abstract

1. Line 46, page 2: What is meant with rigorous calibration? Is this something different than just a calibration?

+++ Rigorous is meant to underscore that we considered every part of the process to make certain that it is correct. This includes testing the sensors by putting the devices on a rotor with a constant round per second to check at the recoding epoch that the devices themselves do produce very similar counts. We made sure they arrow on the devices were always placed in the same direction on the wrist and the waist. The wrist-worn devices need to be placed either on the dominant or non-dominant wrist for comparison. None-rigorous calibration ignores these facts, which the increases systematic errors. We wanted to highlight that we are aware and have tried our best to minimise systematic errors.

2. Line 52, page 2: Are the six activities allocated to different intensities? Please specify.

+++ Yes, we have amended the abstract to include the categories.

+++ The length was calculated by the need to have enough data per child with an epoch of 30s (minimum 16 data points for 8 min). The vigorous activity was a sprinting game along the corridor, where they had to carry balls from a box on one side into a box at the other side. So, they had not a continuous sprint for 10 min, which would not reflect their usual behaviour. Instead they had short breaks at each side to take their breath. This is something we tested in the pilot and found that 3-year olds can do that very well without exhausting themselves too much. In terms of intensity levels, since the epoch was 30s, these short breaks would not create zero values but give a representative activity count for vigorous activity for this age. We have explained this under methods:

Here the children were asked to carry balls from a box on one side of a long corridor to another box at the other end. They had short breaks to take their breath at each side. They did not run continuously fast for 10 min.

+++ ‘Directly observed’ means we had one person, who was not involved in the execution of the activities but watching (observing) from a ‘bird’s perspective’ over each activity with a stopwatch. This person had the list of activities on a sheet of paper and kept accurate note of the start time and end time to make sure the activity lasted for 8 to10 min and can be accurately identified later in the time series of the recordings. This enabled the second person to focus entirely on the child without being distracted by checking the time. We made it more explicit under methods, see below and dedicated it a subheading in the discussion:

Under methods: …which is here termed ‘directly observed’. The ‘direct observation protocol’ entails one person acting as a ‘bystander’ watching (observing) each child’s activities from a ‘bird’s perspective’ and documenting each activity’s start and stop with a stopwatch. This enabled the second person to engage with the child without being distracted by checking the time. The documented start and stop times were essential for extracting the sequences of activities from the times series.

+++ We used this terminology in the abstract:

Time-keeping was ensured using direct observation by an observer.

The term ‘direct observation’ was used by Freedson et al (2005) Calibration of Accelerometer Output for Children. Med Sci Sports Exerc. 2005 Nov;37(11 Suppl):S523-30. doi:10.1249/01.mss.0000185658.28284.ba.: page S529 “Behavioral approaches (e.g., direct observation) for calibrating accelerometry output may be particularly useful when studying young children where measurement and interpretation of energy expenditure data are difficult tasks.”

5. Line 61, page 2: Why classifying into mobile and stationary? Maybe you can add one sentence to clarify this already in the abstract.

+++ This classification goes back to the terminology consensus project process and outcome (Ref. 65,Tremblay et al, 2017). ‘Stationary’ is defined as the superordinate concept uniting different types of behaviours: sedentary (sitting), standing, screen time, reclining, lying. ‘Mobile’ is the superordinate concept uniting light PA, moderate PA and vigorous PA.

6. Line 65, page 2: I wonder what you mean with context information. Please specify.

+++ Contextual information of behaviour refers to the knowledge of what the children were doing over time while measurements took place. Movement intensities of the children’s real habitual repertoire - and not arbitrary exercise, like dancing to a specific beat – is less prone to under- and overestimation of time spent in a given intensity range in observational studies.

+++ We amended the abstract, saying: … Receiver-Operating-Curve classification was applied to determine activity thresholds and to assign two composite movement classes.

For point 5 and 6, we would have added the reference of Tremblay et al (2017) in the abstract, but references are not allowed in the abstract.

Introduction

General:

- Throughout the whole introduction, sentences are very long. I would recommend, separating long sentences (especially Line 78-84, page 3; Line 129-136, page 5).

+++ Revised and shorter.

1. Line 83, page 3: World Health Organization (WHO)

+++ Included.

+++ Context-related movements refer to movements that are required for different behaviours. It is outlined in the Terminology Consensus Project by Tremblay et al (2017). Stationary means you are awake and you do something, where you can move hands, arms, legs and body, but without changing position, which relates to body postures described as standing, squatting, lying, reclining, but not sleeping.

Mobile includes moving your body to a different position, which can include LPA as well as MVPA. Our study primarily reports useful cut-off points to identify MVPA and one of our aims was to differentiate between different intensities of PA.

When you have small children, the level of movements is mechanically and anatomically different from adult PA. While the line is drawn between MVPA and LPA in adults, this does not necessarily work for small children. The intention is to draw the line between stationary activities and LMVPA. Playing on the floor, in the sandpit, in the snow or climbing trees, which adolescents and adults hardly do, belong to LPA and should not be underestimated for its health benefits for small children for sensory-muscular-eye coordination, balance and cognitive development. Reilly et al (2003) has pointed the way to not only measure physical activity levels but also (stationary) inactivity duration for its usefulness as an outcome measure for premature adiposity and/or risk factors, see OBESITY RESEARCH Vol. 11 No. 10 October 2003.

Method

+++ Done

It says now: 3-year-old children without any chronic disease or weight outside the normative range (± 2 standard deviations) using a Swedish growth reference [51] were included.”

2. Page 7, Line 170: Did you use expected MET-values? How did the classification of the activities to the different intensities took place?

+++ We used direct observations during the pilot in characterising movement patterns into intensities from our expertise within the group of 27 years of analysing actigraphic patterns, from newborn babies to the elderly. Specifically, we came up with a list of behaviours that demand different intensities and tested them out in the pilot. Once we had behaviours, which the children liked doing, we ranked them to become progressively more intense.

3. Page 7, Line 181: On which basis did you chose the activities?

+++ By expertise in child behaviour and actigraphic time series analysis.

+++ The criterion here is direct observation in calibrating two different models at two different positions for six different behaviours in a narrow age range. The heart rate data will get their own space in a separate manuscript and refer to this study.

The devices measure what they are intended to measure, but given their different sampling frequencies and sensors, the different acceleration from wrist and waist movements will likely differ in counts and therefore need to be compared within and between devices under observed behaviours The correlations between the actigraphic outputs for the two pairs of devices provide a level of consistency.

5. Page 8, Line 211. Please remove the space between 250 g.

+++ Done.

6. Page 9, Line 219: The raw data of the accelerometer are accelerations (g as unit). How do you get counts from the acceleration?

+++ via the piezoelectric effect, where the acceleration in the piezoelectric sensing material is turned into an electrical charge, and the level of charge is then converted into a count, which is proportional to the level of acceleration, see piezoelectric sensor (https://en.wikipedia.org/wiki/Piezoelectric_sensor).

7. Page 9, Line 225: proprietary software – what is this? I am not sure if everybody know what this means. Please contextualize.

+++ replaced by Motionware software

8. Page 9, Line 226: what is the behavioral observation protocol. Please describe this in more detail. This term is frequently in your manuscript but for me it is not sure what is meant with that.

+++ please find above under point 4 .

+++ We calibrated behavioural classes using movement intensities. Reliability refers to the consistency of a measure. A high test-retest correlation makes sense when the construct being measured is assumed to be consistent over time, which is the case for example for intelligence, but constructs of behaviour under the influence of emotions (how they feel right now) are not assumed to be stable over time. The very nature of behaviour that produces a low test-retest correlation over a period of a week or so would not be a cause for concern. Validity is the extent to which the intensity from a measurement represents the behaviour it intends to cover. We use content validity as the extend to which the measurement method appears “to cover” the construct of behavioural class of interest. For this we use the device models for quantitative assessment and checking the measurement against the conceptualised definition by the ‘Direct observation protocol’. We used ‘direct observation’ as criterion.

10. Page 10, Line 241: observer-based behaviors – what is this? Please contextualize.

+++ please find above under point 4 and figure 1, Bottom explanatory table.

11. Page 10, Line 246: I wonder why you make a distinction between mobility and stationary behavior when the common classification are the intensity levels.

+++ We use classification guided by behaviour not intensity levels. Behaviour-based movement cut-offs means movement cut-offs derived from intensity levels. We do not use intensity classifications for physical activity only – we look holistically into behaviours.

12. Page 10, Line 252: on which basis did you do that? MET-values?

+++ We did not use exertion of energy expenditure during physical activity in the children. We used behaviour, broadly defined as anything a living being does, which includes actions. Here, we used the actions of different behaviours as described.

Results

1. Do you have the individual raw data/activity counts of each conducted activity?

+++ We do have the activity counts/30s as raw data over each person and conducted behaviour and have provided open access to non-commercial use.

2. Page 16, Table 3: Cut-off values per which second?

+++ We used counts per 30s.

Discussion

+++ We have included sub-headings into the discussion.

2. Page 16, Line 382: Again, what is the observation technique? Are there no results concerning this?

+++ I refer to #4 above on the technique. Regarding results of this technique, we used the start and end times to identify and extract each sequence of behaviour correctly in the time series. Observation was the choice over indirect calorimetry because of the age of the children.

The observation results are reflected in the cumulated wear-time of devices by behaviour, reported in the supplementary figure S1, Same duration for every behaviour avoids introducing a bias from unequal variation from different durations of the behaviours performed.

3. Page 16, Line 383: Children’s activity rating scale --> in your study? The sentence is not clear to me.

+++ We have amended the sentences to clarify.

4. Page 19, 390 ff: Is this also a phenomenon found in actual literature as you mention “Already two decades ago”?

+++ . ‘Already’ refers to the fact that this method has been used in a calibration study in healthy 3-5 year old children. To our knowledge, indirect calorimetry has been used but only in older children (Freedson et al, 2005) or

Attachment

Submitted filename: PONE - RESPONSE TO REVIEWERS FINAL.docx

pone.0316747.s008.docx^{(67.8KB, docx)}

PLoS One. doi: 10.1371/journal.pone.0316747.r003

Decision Letter 1

Duncan S Buchan

17 Dec 2024

Behaviour-based movement cut-off points in 3-year old children comparing wrist- with hip-worn actigraphs MW8 and GT3X

PONE-D-24-07355R1

Dear Dr. Wulff,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice will be generated when your article is formally accepted. Please note, if your institution has a publishing partnership with PLOS and your article meets the relevant criteria, all or part of your publication costs will be covered. Please make sure your user information is up-to-date by logging into Editorial Manager at Editorial Manager® and clicking the ‘Update My Information' link at the top of the page. If you have any questions relating to publication charges, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Duncan S Buchan

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

Reviewer #2: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

Reviewer #1: Yes

Reviewer #2: Yes

**********

6. Review Comments to the Author

Reviewer #1: Dear Authors,

Thank you for the thorough revision of your manuscript. The revisions were carried out with great care, and all comments and suggestions have been addressed in an excellent manner.

Based on the significant improvements and the high quality of the work, I am pleased to recommend your manuscript for acceptance.

I sincerely appreciate your efforts and wish you continued success in your research.

Reviewer #2: A thorough revision of an already excellent paper. Thank you for the nice read. No further recommendations!

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean? ). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy .

Reviewer #1: No

Reviewer #2: No

**********

PLoS One. doi: 10.1371/journal.pone.0316747.r004

Acceptance letter

Duncan S Buchan

PONE-D-24-07355R1

PLOS ONE

Dear Dr. Wulff,

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now being handed over to our production team.

At this stage, our production department will prepare your paper for publication. This includes ensuring the following:

* All references, tables, and figures are properly cited

* All relevant supporting information is included in the manuscript submission,

* There are no issues that prevent the paper from being properly typeset

If revisions are needed, the production department will contact you directly to resolve them. If no revisions are needed, you will receive an email when the publication date has been set. At this time, we do not offer pre-publication proofs to authors during production of the accepted work. Please keep in mind that we are working through a large volume of accepted articles, so please give us a few weeks to review your paper and let you know the next and final steps.

Lastly, if your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

If we can help with anything else, please email us at customercare@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Duncan S Buchan

Academic Editor

PLOS ONE

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

S1 Table. Highlighting some of the different approaches taken from sports medicine and chronobiology/sleep, with references of published examples (References Supplementary to S1 Table: 42 articles).

(DOCX)

pone.0316747.s001.docx^{(124.5KB, docx)}

S2 Table. Receiver operating characteristics curve (ROC) analysis in One-vs-One scheme, comparing pairwise combination of adjacent behavioural classes as described in Fig 2a.

(DOCX)

pone.0316747.s002.docx^{(44.8KB, docx)}

S3 Fig. Duration in minutes of cumulated wear-time of devices (MW8, GT3X) simultaneously worn at wrist and hip positions while performing each of the six behaviours.

The epoch for storing the activity counts was 0.5 min.

(DOCX)

pone.0316747.s003.docx^{(69.6KB, docx)}

S4 Fig. ROC curves for the wrist-worn Motionwatch 8.

(DOCX)

pone.0316747.s004.docx^{(146KB, docx)}

S5 Fig. ROC curves for wrist-worn GT3X.

(DOCX)

pone.0316747.s005.docx^{(144.3KB, docx)}

S6 Fig. ROC curves for hip-worn Motionwatch 8.

(DOCX)

pone.0316747.s006.docx^{(148.3KB, docx)}

S7 Fig. ROC curves for hip-worn GT3X.

(DOCX)

pone.0316747.s007.docx^{(198.8KB, docx)}

Attachment

Submitted filename: PONE - RESPONSE TO REVIEWERS FINAL.docx

pone.0316747.s008.docx^{(67.8KB, docx)}

Data Availability Statement

[pone.0316747.ref001] 1.Bull FC, Al-Ansari SS, Biddle S, Borodulin K, Buman MP, Cardon G, et al. World Health Organization 2020 guidelines on physical activity and sedentary behaviour. Br J Sports Med. 2020;54(24):1451–62. doi: 10.1136/bjsports-2020-102955 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0316747.ref002] 2.Hammad G, Reyt M, Beliy N, Baillet M, Deantoni M, Lesoinne A, et al. pyActigraphy: open-source python package for actigraphy data visualization and analysis. PLoS Comput Biol. 2021;17(10):e1009514. doi: 10.1371/journal.pcbi.1009514 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0316747.ref003] 3.Weitz M, Syed S, Hopstock LA, Morseth B, Prasad DK, Horsch A. Discrimination of sleep and wake periods from a hip-worn raw acceleration sensor using recurrent neural networks. medRxiv. 2022. p. 2022.03.07.22270992. doi: 10.1101/2022.03.07.22270992 [DOI] [Google Scholar]

[pone.0316747.ref004] 4.Tremblay MS, Carson V, Chaput J-P, Connor Gorber S, Dinh T, Duggan M, et al. Canadian 24-hour movement guidelines for children and youth: an integration of physical activity, sedentary behaviour, and sleep. Appl Physiol Nutr Metab. 2016;41(6 Suppl 3):S311–27. doi: 10.1139/apnm-2016-0151 [DOI] [PubMed] [Google Scholar]

[pone.0316747.ref005] 5.Rosenberger ME, Fulton JE, Buman MP, Troiano RP, Grandner MA, Buchner DM, et al. The 24-hour activity cycle: a new paradigm for physical activity. Med Sci Sports Exerc. 2019;51(3):454–64. doi: 10.1249/MSS.0000000000001811 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0316747.ref006] 6.Chang Z, Lei W. A study on the relationship between physical activity, sedentary behavior, and sleep duration in preschool children. Front Public Health. 2021;9:618962. doi: 10.3389/fpubh.2021.618962 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0316747.ref007] 7.Butte NF, Puyau MR, Wilson TA, Liu Y, Wong WW, Adolph AL, et al. Role of physical activity and sleep duration in growth and body composition of preschool-aged children. Obesity (Silver Spring). 2016;24(6):1328–35. doi: 10.1002/oby.21489 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0316747.ref008] 8.Møller NC, Christensen LB, Mølgaard C, Ejlerskov KT, Pfeiffer KA, Michaelsen KF. Descriptive analysis of preschool physical activity and sedentary behaviors – a cross sectional study of 3-year-olds nested in the SKOT cohort. BMC Public Health. 2017;17(1):613. doi: 10.1186/s12889-017-4521-3 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0316747.ref009] 9.World Health Organization. Report of the commission on ending childhood obesity. Geneva, Switzerland: World Health Organization [Internet]; 2016. Available from: https://www.who.int/publications/i/item/9789241510066 [Google Scholar]

[pone.0316747.ref010] 10.Okely AD, Ghersi D, Hesketh KD, Santos R, Loughran SP, Cliff DP, et al. A collaborative approach to adopting/adapting guidelines - the Australian 24-hour movement guidelines for the early years (birth to 5 years): an integration of physical activity, sedentary behavior, and sleep. BMC Public Health. 2017;17(Suppl 5):869. doi: 10.1186/s12889-017-4867-6 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0316747.ref011] 11.Trost SG, Cliff DP, Ahmadi MN, Tuc NV, Hagenbuchner M. Sensor-enabled activity class recognition in preschoolers: hip versus wrist data. Med Sci Sports Exerc. 2018;50(3):634–41. doi: 10.1249/MSS.0000000000001460 [DOI] [PubMed] [Google Scholar]

[pone.0316747.ref012] 12.Johansson E, Larisch L-M, Marcus C, Hagströmer M. Calibration and validation of a wrist- and hip-worn actigraph accelerometer in 4-year-old children. PLoS One. 2016;11(9):e0162436. doi: 10.1371/journal.pone.0162436 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0316747.ref013] 13.Pate RR, Almeida MJ, McIver KL, Pfeiffer KA, Dowda M. Validation and calibration of an accelerometer in preschool children. Obesity (Silver Spring). 2006;14(11):2000–6. doi: 10.1038/oby.2006.234 [DOI] [PubMed] [Google Scholar]

[pone.0316747.ref014] 14.van Cauwenberghe E, Labarque V, Trost SG, de Bourdeaudhuij I, Cardon G. Calibration and comparison of accelerometer cut points in preschool children. Int J Pediatr Obes. 2011;6(2–2):e582-9. doi: 10.3109/17477166.2010.526223 [DOI] [PubMed] [Google Scholar]

[pone.0316747.ref015] 15.Roscoe CMP, James RS, Duncan MJ. Calibration of GENEActiv accelerometer wrist cut-points for the assessment of physical activity intensity of preschool aged children. Eur J Pediatr. 2017;176(8):1093–8. doi: 10.1007/s00431-017-2948-2 [DOI] [PubMed] [Google Scholar]

[pone.0316747.ref016] 16.Pfeiffer KA, McIver KL, Dowda M, Almeida MJCA, Pate RR. Validation and calibration of the Actical accelerometer in preschool children. Med Sci Sports Exerc. 2006;38(1):152–7. doi: 10.1249/01.mss.0000183219.44127.e7 [DOI] [PubMed] [Google Scholar]

[pone.0316747.ref017] 17.Janssen X, Cliff DP, Reilly JJ, Hinkley T, Jones RA, Batterham M, et al. Validation and calibration of the activPAL™ for estimating METs and physical activity in 4-6 year olds. J Sci Med Sport. 2014;17(6):602–6. doi: 10.1016/j.jsams.2013.10.252 [DOI] [PubMed] [Google Scholar]

[pone.0316747.ref018] 18.Ahmadi MN, Chowdhury A, Pavey T, Trost SG. Laboratory-based and free-living algorithms for energy expenditure estimation in preschool children: a free-living evaluation. PLoS ONE. 2020;15(5):e0233229. doi: 10.1371/journal.pone.0233229 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0316747.ref019] 19.Johansson E, Ekelund U, Nero H, Marcus C, Hagströmer M. Calibration and cross-validation of a wrist-worn Actigraph in young preschoolers. Pediatr Obes. 2015;10(1):1–6. doi: 10.1111/j.2047-6310.2013.00213.x [DOI] [PubMed] [Google Scholar]

[pone.0316747.ref020] 20.Costa S, Barber SE, Cameron N, Clemes SA. Calibration and validation of the ActiGraph GT3X+ in 2-3 year olds. J Sci Med Sport. 2014;17(6):617–22. doi: 10.1016/j.jsams.2013.11.005 [DOI] [PubMed] [Google Scholar]

[pone.0316747.ref021] 21.Pate RR, Dishman RK, Dowda M, McIver KL, Pfeiffer KA, Porter DE, et al. A summary of one research team’s contributions to understanding physical activity behavior in children and youth. Int J Environ Res Public Health. 2022;19(21):14136. doi: 10.3390/ijerph192114136 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0316747.ref022] 22.Van Cauwenberghe E, Gubbels J, De Bourdeaudhuij I, Cardon G. Feasibility and validity of accelerometer measurements to assess physical activity in toddlers. Int J Behav Nutr Phys Act. 2011;8:67. doi: 10.1186/1479-5868-8-67 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0316747.ref023] 23.Dolinsky DH, Brouwer RJN, Evenson KR, Siega-Riz AM, Østbye T. Correlates of sedentary time and physical activity among preschool-aged children. Prev Chronic Dis. 2011;8(6):A131. [PMC free article] [PubMed] [Google Scholar]

[pone.0316747.ref024] 24.Timmons BW, Leblanc AG, Carson V, Connor Gorber S, Dillman C, Janssen I, et al. Systematic review of physical activity and health in the early years (aged 0-4 years). Appl Physiol Nutr Metab. 2012;37(4):773–92. doi: 10.1139/h2012-070 [DOI] [PubMed] [Google Scholar]

[pone.0316747.ref025] 25.Liu F, Schrack J, Wanigatunga SK, Rabinowitz JA, He L, Wanigatunga AA, et al. Comparison of sleep parameters from wrist-worn ActiGraph and Actiwatch devices. Sleep. 2024;47(2):zsad155. doi: 10.1093/sleep/zsad155 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0316747.ref026] 26.Schoch SF, Jenni OG, Kohler M, Kurth S. Actimetry in infant sleep research: an approach to facilitate comparability. Sleep. 2019;42(7):zsz083. doi: 10.1093/sleep/zsz083 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0316747.ref027] 27.Chandler JL, Brazendale K, Beets MW, Mealing BA. Classification of physical activity intensities using a wrist-worn accelerometer in 8-12-year-old children. Pediatr Obes. 2016;11(2):120–7. doi: 10.1111/ijpo.12033 [DOI] [PubMed] [Google Scholar]

[pone.0316747.ref028] 28.Hislop J, Palmer N, Anand P, Aldin T. Validity of wrist worn accelerometers and comparability between hip and wrist placement sites in estimating physical activity behaviour in preschool children. Physiol Meas. 2016;37(10):1701–14. doi: 10.1088/0967-3334/37/10/1701 [DOI] [PubMed] [Google Scholar]

[pone.0316747.ref029] 29.Fairclough SJ, Noonan R, Rowlands AV, Van Hees V, Knowles Z, Boddy LM. Wear compliance and activity in children wearing wrist- and hip-mounted accelerometers. Med Sci Sports Exerc. 2016;48(2):245–53. doi: 10.1249/MSS.0000000000000771 [DOI] [PubMed] [Google Scholar]

[pone.0316747.ref030] 30.Arvidsson D, Fridolfsson J, Börjesson M, Andersen LB, Ekblom Ö, Dencker M, et al. Re-examination of accelerometer data processing and calibration for the assessment of physical activity intensity. Scand J Med Sci Sports. 2019;29(10):1442–52. doi: 10.1111/sms.13470 [DOI] [PubMed] [Google Scholar]

[pone.0316747.ref031] 31.Migueles JH, Aadland E, Andersen LB, Brønd JC, Chastin SF, Hansen BH, et al. GRANADA consensus on analytical approaches to assess associations with accelerometer-determined physical behaviours (physical activity, sedentary behaviour and sleep) in epidemiological studies. Br J Sports Med. 2022;56(7):376–84. doi: 10.1136/bjsports-2020-103604 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0316747.ref032] 32.Montoye AHK, Westgate BS, Fonley MR, Pfeiffer KA. Cross-validation and out-of-sample testing of physical activity intensity predictions with a wrist-worn accelerometer. J Appl Physiol (1985). 2018;124(5):1284–93. doi: 10.1152/japplphysiol.00760.2017 [DOI] [PubMed] [Google Scholar]

[pone.0316747.ref033] 33.Troiano RP, McClain JJ, Brychta RJ, Chen KY. Evolution of accelerometer methods for physical activity research. Br J Sports Med. 2014;48(13):1019–23. doi: 10.1136/bjsports-2014-093546 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0316747.ref034] 34.Wulff K, Dedek A, Siegmund R. Circadian and Ultradian time patterns in human behavior: Part 2: social synchronisation during the development of the infant’s diurnal activity-rest pattern. Biol Rhythm Res. 2001;32(5):529–46. doi: 10.1076/brhm.32.5.529.1295 [DOI] [Google Scholar]

[pone.0316747.ref035] 35.Wulff K, Siegmund R. Time pattern analysis of activity-rest rhythms in families with infants using actigraphy. In: Salzarulo P, Ficca G, editors. Awakening and Sleep–Wake Cycle Across Development. John Benjamins Publishing Company; 2002. p. 149–69. [Google Scholar]

[pone.0316747.ref036] 36.Freedson PS, John D. Comment on “estimating activity and sedentary behavior from an accelerometer on the hip and wrist”. Med Sci Sports Exerc. 2013;45(5):962–3. doi: 10.1249/MSS.0b013e31827f024d [DOI] [PubMed] [Google Scholar]

[pone.0316747.ref037] 37.Jeong D-H, Kim S-E, Choi W-H, Ahn S-H. A comparative study on the influence of undersampling and oversampling techniques for the classification of physical activities using an imbalanced accelerometer dataset. Healthcare (Basel). 2022;10(7):1255. doi: 10.3390/healthcare10071255 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0316747.ref038] 38.Cellini N, McDevitt EA, Mednick SC, Buman MP. Free-living cross-comparison of two wearable monitors for sleep and physical activity in healthy young adults. Physiol Behav. 2016;157:79–86. doi: 10.1016/j.physbeh.2016.01.034 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0316747.ref039] 39.Van Remoortel H, Giavedoni S, Raste Y, Burtin C, Louvaris Z, Gimeno-Santos E, et al. Validity of activity monitors in health and chronic disease: a systematic review. Int J Behav Nutr Phys Act. 2012;9:84. doi: 10.1186/1479-5868-9-84 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0316747.ref040] 40.Migueles JH, Cadenas-Sanchez C, Ekelund U, Delisle Nyström C, Mora-Gonzalez J, Löf M, et al. Accelerometer data collection and processing criteria to assess physical activity and other outcomes: a systematic review and practical considerations. Sports Med. 2017;47(9):1821–45. doi: 10.1007/s40279-017-0716-0 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0316747.ref041] 41.Montoye AHK, Pivarnik JM, Mudd LM, Biswas S, Pfeiffer KA. Comparison of activity type classification accuracy from accelerometers worn on the hip, wrists, and thigh in young, apparently healthy adults. Meas Phys Educ Exerc Sci. 2016;20(3):173–83. doi: 10.1080/1091367x.2016.1192038 [DOI] [Google Scholar]

[pone.0316747.ref042] 42.Gao Z, Liu W, McDonough DJ, Zeng N, Lee JE. The dilemma of analyzing physical activity and sedentary behavior with wrist accelerometer data: challenges and opportunities. J Clin Med. 2021;10(24):5951. doi: 10.3390/jcm10245951 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0316747.ref043] 43.Ancoli-Israel S, Cole R, Alessi C, Chambers M, Moorcroft W, Pollak CP. The role of actigraphy in the study of sleep and circadian rhythms. Sleep. 2003;26(3):342–92. doi: 10.1093/sleep/26.3.342 [DOI] [PubMed] [Google Scholar]

[pone.0316747.ref044] 44.Maczák B, Vadai G, Dér A, Szendi I, Gingl Z. Detailed analysis and comparison of different activity metrics. PLoS One. 2021;16(12):e0261718. doi: 10.1371/journal.pone.0261718 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0316747.ref045] 45.Skeldon AC, Dijk D-J, Meyer N, Wulff K. Extracting circadian and sleep parameters from longitudinal data in schizophrenia for the design of pragmatic light interventions. Schizophr Bull. 2022;48(2):447–56. doi: 10.1093/schbul/sbab124 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0316747.ref046] 46.Danilenko KV, Stefani O, Voronin KA, Mezhakova MS, Petrov IM, Borisenkov MF, et al. Wearable light-and-motion dataloggers for sleep/wake research: a review. Appl Sci. 2022;12(22):11794. doi: 10.3390/app122211794 [DOI] [Google Scholar]

[pone.0316747.ref047] 47.Falck RS, Best JR, Li MCR, Eng JJ, Liu-Ambrose T. Revisiting the MotionWatch8©: calibrating cut-points for measuring physical activity and sedentary behavior among adults with stroke. Front Aging Neurosci. 2019;11:203. doi: 10.3389/fnagi.2019.00203 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0316747.ref048] 48.Landry GJ, Falck RS, Beets MW, Liu-Ambrose T. Measuring physical activity in older adults: calibrating cut-points for the MotionWatch 8©. Front Aging Neurosci. 2015;7. [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0316747.ref049] 49.Lin H-P, Lynk N, Moore LL, Cabral HJ, Heffernan KS, Dumas AK, et al. A pragmatic approach to the comparison of wrist-based cutpoints of physical activity intensity for the MotionWatch8 accelerometer in children. PLoS One. 2020;15(6):e0234725. doi: 10.1371/journal.pone.0234725 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0316747.ref050] 50.Kushida CA, Chang A, Gadkary C, Guilleminault C, Carrillo O, Dement WC. Comparison of actigraphic, polysomnographic, and subjective assessment of sleep parameters in sleep-disordered patients. Sleep Med. 2001;2(5):389–96. doi: 10.1016/s1389-9457(00)00098-8 [DOI] [PubMed] [Google Scholar]

[pone.0316747.ref051] 51.Cole RJ, Kripke DF, Gruen W, Mullaney DJ, Gillin JC. Automatic sleep/wake identification from wrist activity. Sleep. 1992;15(5):461–9. doi: 10.1093/sleep/15.5.461 [DOI] [PubMed] [Google Scholar]

[pone.0316747.ref052] 52.Sadeh A, Sharkey KM, Carskadon MA. Activity-based sleep-wake identification: an empirical test of methodological issues. Sleep. 1994;17(3):201–7. doi: 10.1093/sleep/17.3.201 [DOI] [PubMed] [Google Scholar]

[pone.0316747.ref053] 53.Wikland KA, Luo ZC, Niklasson A, Karlberg J. Swedish population-based longitudinal reference values from birth to 18 years of age for height, weight and head circumference. Acta Paediatr. 2002;91(7):739–54. doi: 10.1080/08035250213216 [DOI] [PubMed] [Google Scholar]

[pone.0316747.ref054] 54.Camntech. The MotionWatch and MotionWare user guide. Version 1226a. 2019. [Google Scholar]

[pone.0316747.ref055] 55.Powers D. Evaluation: from precision, recall and F-measure to ROC, Informedness, Markedness and correlation. J Machine Learn Technol. 2011;2:37–63. [Google Scholar]

[pone.0316747.ref056] 56.Metz CE. Basic principles of ROC analysis. Semin Nucl Med. 1978;8(4):283–98. doi: 10.1016/s0001-2998(78)80014-2 [DOI] [PubMed] [Google Scholar]

[pone.0316747.ref057] 57.Pedregosa F, Varoquaux G, Gramfort A, Michel V, Thirion B, Grisel O. Scikit-learn: machine learning in python. J Machine Learn Res. 2011. [Google Scholar]

[pone.0316747.ref058] 58.Akobeng AK. Understanding diagnostic tests 3: receiver operating characteristic curves. Acta Paediatr. 2007;96(5):644–7. doi: 10.1111/j.1651-2227.2006.00178.x [DOI] [PubMed] [Google Scholar]

[pone.0316747.ref059] 59.Hosmer DW Jr, Lemenshow S, Sturdivant RX. Assessing the Fit of the Model. Applied Logistic Regression. John Wiley & Sons, Ltd; 2000. p. 143–202. [Google Scholar]

[pone.0316747.ref060] 60.Reilly JJ, Coyle J, Kelly L, Burke G, Grant S, Paton JY. An objective method for measurement of sedentary behavior in 3- to 4-year olds. Obes Res. 2003;11(10):1155–8. doi: 10.1038/oby.2003.158 [DOI] [PubMed] [Google Scholar]

[pone.0316747.ref061] 61.DuRant RH, Baranowski T, Puhl J, Rhodes T, Davis H, Greaves KA, et al. Evaluation of the Children’s Activity Rating Scale (CARS) in young children. Med Sci Sports Exerc. 1993;25(12):1415–21. doi: 10.1249/00005768-199312000-00016 [DOI] [PubMed] [Google Scholar]

[pone.0316747.ref062] 62.Finn KJ, Specker B. Comparison of Actiwatch activity monitor and children’s activity rating scale in children. Med Sci Sports Exerc. 2000;32(10):1794–7. doi: 10.1097/00005768-200010000-00021 [DOI] [PubMed] [Google Scholar]

[pone.0316747.ref063] 63.Brown WH, Pfeiffer KA, Mclver KL, Dowda M, Almeida MJCA, Pate RR. Assessing preschool children’s physical activity: the Observational System for Recording Physical Activity in children-preschool version. Res Q Exerc Sport. 2006;77(2):167–76. doi: 10.1080/02701367.2006.10599351 [DOI] [PubMed] [Google Scholar]

[pone.0316747.ref064] 64.Bernhofer EI. Investigating the concept of rest for research and practice. J Adv Nurs. 2016;72(5):1012–22. doi: 10.1111/jan.12910 [DOI] [PubMed] [Google Scholar]

[pone.0316747.ref065] 65.Tremblay MS, Aubert S, Barnes JD, Saunders TJ, Carson V, Latimer-Cheung AE, et al. Sedentary Behavior Research Network (SBRN) - Terminology Consensus Project process and outcome. Int J Behav Nutr Phys Act. 2017;14(1):75. doi: 10.1186/s12966-017-0525-8 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0316747.ref066] 66.Brønd JC, Arvidsson D. Sampling frequency affects the processing of Actigraph raw acceleration data to activity counts. J Appl Physiol (1985). 2016;120(3):362–9. doi: 10.1152/japplphysiol.00628.2015 [DOI] [PubMed] [Google Scholar]

[pone.0316747.ref067] 67.Cyriax EF. On the rotary movements of the wrist. J Anat. 1926;60(Pt 2):199–201. [PMC free article] [PubMed] [Google Scholar]

[pone.0316747.ref068] 68.Beck F, Marzi I, Eisenreich A, Seemüller S, Tristram C, Reimers AK. Determination of cut-off points for the Move4 accelerometer in children aged 8-13 years. BMC Sports Sci Med Rehabil. 2023;15(1):163. doi: 10.1186/s13102-023-00775-4 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0316747.ref069] 69.Tsanas A. Investigating wrist-based acceleration summary measures across different sample rates towards 24-hour physical activity and sleep profile assessment. Sensors (Basel). 2022;22(16):6152. doi: 10.3390/s22166152 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0316747.ref070] 70.Orme M, Wijndaele K, Sharp SJ, Westgate K, Ekelund U, Brage S. Combined influence of epoch length, cut-point and bout duration on accelerometry-derived physical activity. Int J Behav Nutr Phys Act. 2014;11(1):34. doi: 10.1186/1479-5868-11-34 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0316747.ref071] 71.Kim Y, Beets MW, Welk GJ. Everything you wanted to know about selecting the “right” Actigraph accelerometer cut-points for youth, but…: a systematic review. J Sci Med Sport. 2012;15(4):311–21. doi: 10.1016/j.jsams.2011.12.001 [DOI] [PubMed] [Google Scholar]

PERMALINK

Behaviour-based movement cut-off points in 3-year old children comparing wrist- with hip-worn actigraphs MW8 and GT3X

Daniel Jansson

Rikard Westlander

Jonas Sandlund

Christina E West

Magnus Domellöf

Katharina Wulff

Roles

Abstract

Background

Methods

Results

Conclusion

Introduction

Challenges for compositional analysis

Rational of the study

The present study

Materials and methods

Participants

Procedures

Fig 1. Top: Representative example of an area graph derived from wrist activity patterns of a child performing six investigator-observed behaviours (dashed boxes) shown for a MotionWatch 8 device attached to the non-dominant wrist.

Statistical analysis

Fig 2. Schematic description of ROC analysis procedures: A) One versus one (OvsO) and B) one versus the rest (OvsR) in order of descending intensities.

Results

Table 1. Descriptive characteristics of the participants (n = 30). Shown are means ± SD.

Table 2. Mean ± SD of activity counts (counts/30 seconds) for each activity ranked from highest to lowest intensity measured at wrist-level and hip-level with ActiGraph GT3X and MotionWatch 8. Mean ± SD of the duration spent in activity is consistently similar.

Comparing MW8 with GT3X revealed significant correlations in activity counts at their respective positions, despite different operating scales

Fig 3. Relationship between activity counts/30s in 3-year-old children (n.

Fig 4. Boxplots for six different intensity levels measured in 3-year old children with A) wrist-worn MotionWatch 8; B) hip-worn MotionWatch 8; C) wrist-worn ActiGraph GT3X; and D) hip-worn ActiGraph GT3X.

Collapsing behavioural classes indicated outstanding accuracy thresholds for ‘mobile PA’ and excellent accuracy thresholds for ‘stationary PA’ intensities

Table 3. ROC-AUC analysis for wrist-worn MotionWatch 8 and ActiGraph GT3X devices after collapsing categories. See details for collapsed categories in Fig 2.

Table 4. ROC-AUC analysis for hip-worn MotionWatch 8 and ActiGraph GT3X devices after collapsing categories. See details for collapsed categories in Fig. 2.

Discussion

Observational method

Types and terminology of movement behaviours

Sensory modes and body positions

Within model comparison of cut-off points

Strength and limitations

Conclusions

Supporting information

Acknowledgments

Data Availability

Funding Statement

References

Decision Letter 0

Duncan S Buchan

Roles

Author response to Decision Letter 1

Decision Letter 1

Duncan S Buchan

Roles

Acceptance letter

Duncan S Buchan

Roles

Associated Data

Supplementary Materials

Data Availability Statement

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases