A Systematic Review of Children’s Physical Activity Patterns: Concept, Operational Definitions, Instruments, Statistical Analyses, and Health Implications

Despite the widespread use of the expression “physical activity pattern” (PAP), there apparently is no general consensus regarding its definition. This systematic review aimed to examine available research focussing on (1) definitions of PAP, (2) instruments/techniques used to describe PAP, (3) statistical approaches used to analyse PAP, and (4) implications of PAP on children’s health. A systematic review of the available literature was done to identify studies published up to October 2019, and 76 studies were eligible. None of the studies presented a formal definition of PAP; a wide range of instruments were used to investigate children’s PAP, and most of the revised studies did not explicitly present a formal statistical model to define PAP. Twenty-four papers purported to examine associations between PAP and health indicators. The review highlights no consensus on a clear PAP definition whatever the instrument used to capture it, and we did not find any agreement regarding how best to analyse PAP. We suggest that PAP should be used when targeting the investigation of similarities/dissimilarities, as well as stabilities and/or changes in children’s PA at an intra-personal level. In sum, PAP should be used to best describe individual streams of behaviours, and not exclusively PA levels/intensities.


Introduction
Physical activity (PA) is positively associated with numerous physical, psychological, and cognitive health benefits in children and youth [1]. Despite the development of science-based global guidelines for achieving these health benefits [2], epidemiological data suggest that PA among youth is declining in many countries, and that pronounced proportions of children do not meet the recommendations [3]. A large number of studies have investigated putative correlates of children's PA, focussing mainly on understanding the differences in mean levels of PA at various intensities (i.e., sedentary, or light, moderate, vigorous) [4]. Yet, children of the same age and sex differ widely in the way they engage in the manifold expressions of their PA [5]. Indeed, not only does every child's PA differ in duration and intensity, but it also varies according to spaces/places [6], segments of the day [7], weather conditions [8], and seasons [9]. Such complexity has raised the bar and challenged researchers to also of PAP; (2) instruments and techniques used to describe PAP; (3) statistical procedures used to analyse PAP; and (4) implications of PAP for children's health.

Protocol and Registration
This systematic review was registered in the International Prospective Register of Systematic Reviews (PROSPERO; registration number CRD42018096728), and was performed following the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines [24].

Search Strategy
A systematic review of the literature was performed using four electronic bibliographic databases (PubMed, Scopus, Academic Search Complete, and SPORTSDiscus) using the following search terms, "physical activity pattern*" and "child*". Bibliographic records for the identified papers were extracted into EndNote reference manager software (version X8, Thomson ResearchSoft ® ), where duplicated results were identified and removed. Titles and abstracts of potentially relevant papers were screened, and those selected were screened independently in full text by two reviewers. To be included in the present review, eligible papers were confirmed by the two reviewers, and if discrepancies arose, they were solved by discussion between the reviewers.

Eligibility Criteria
Original peer-reviewed articles published from the year of database inception to October 6, 2019 were eligible for inclusion. Additional eligibility criteria included: (1) published in English, (2) the purpose was stated as investigating children's PAP and presented as a study aim or clearly detailed in the Methods section; (3) the sample was comprised of children aged 6-11 years or with mean age between 6.0 and 11.49 years, or in the case of samples with older/younger children, the results should be presented by age or most of the sample should be included in this age range (information clearly presented in the studies); (4) the sample should not be only comprised by children with special needs; yet, if the sample comprised children with special needs and also children with typical development, the study was included.

Methodological Quality Assessment and Network
Then, the articles selected were analysed for their data quality, taking into account seven quality criteria developed ad hoc that were adapted for previous studies [25,26] (Table 1), which were scored on a three-point scale, and the sum of these points (from 0 to 14 points), meaning the methodological quality rating (which was represented in a percentage scale). This procedure was performed by two independent reviewers, separately, and if discrepancies arose, they were solved by common agreement. A bibliometric analysis was performed by a network created using the software Gephi, aiming to show the networks defined by the authors of the revised papers (Supplementary file Figure S1). Table 1. Quality criteria developed to analyse the papers.

Question
Answer Score

Q1
The aim(s) of the study(ies) is/are clearly set out Yes   Figure 1 shows the flow diagram of the article identification process. Using the aforementioned keywords, the electronic database search produced an initial total of 1195 results. After excluding duplicates, 595 were screened by their titles and abstracts. This process allowed a further exclusion of 483 papers, resulting in 112 papers for full textual assessment. Finally, 76 studies were found to be eligible for this review.  Figure 1 shows the flow diagram of the article identification process. Using the aforementioned keywords, the electronic database search produced an initial total of 1195 results. After excluding duplicates, 595 were screened by their titles and abstracts. This process allowed a further exclusion of 483 papers, resulting in 112 papers for full textual assessment. Finally, 76 studies were found to be eligible for this review. From the included studies, relevant data were extracted, as per our aims: sample characteristics [size, age (mean and/or range), country], methods used to measure PA and to describe PAP, as well as possible links with health risk factors. Most of the studies presented within the retrieved papers, as expected, were conducted in Europe and/or North America, with a few being carried out in Asia, Africa, Oceania, and South America. Study samples ranged from 15 [23] to 17,500 [27] subjects, and with the exception of one study, all investigated both boys and girls. From the included studies, relevant data were extracted, as per our aims: sample characteristics [size, age (mean and/or range), country], methods used to measure PA and to describe PAP, as well as possible links with health risk factors. Most of the studies presented within the retrieved papers, as expected, were conducted in Europe and/or North America, with a few being carried out in Asia, Africa, Oceania, and South America. Study samples ranged from 15 [23] to 17,500 [27] subjects, and with the exception of one study, all investigated both boys and girls.

Methodological Quality
Results regarding the papers' quality rating revealed that the highest mean score was observed for studies that used mixed instruments to determine/measure PAP (9.45 points, meaning a quality score percentage of 67.53%), followed by those that used pedometers (9.15 points, percentage of 65.31%), accelerometers (9 points, with a quality score percentage of 64.29%), heart rate monitors (8.67 points, 61.90%), and questionnaire/observation methods (8.18 points, percentage of 58.40%). All of the articles were scored as 0 for Q3, and the most of them received a score of 1 for Qs 4, 5, 6, and 7.

Physical Activity Pattern Definition
There is no consensus regarding a definition of PAP in conceptual and/or operational terms. As far as we could tell, no paper included in this review presented a formal definition. This implies that different schemes to tackle PAP were used, which varied according to the method used to measure PA. Consequently, most studies approached PAP considering children's involvement in different PA intensities (sedentary, light, moderate, vigorous, very vigorous), which were obtained according to specific accelerometer or pedometer cut-points or even based on physiological measures such as heart rate. Further, some studies also presented PAP as the type of activity children performed, as well as their sports participation.
In some cases, segments of the day as well as different days of the week, or even different seasons, were used to define PAP. As such, some researchers focused their attention on describing children's PA in different segments of the day, i.e., during school (all school time, breaks, physical education classes), before/after school, and in the evening, trying to find differences among such time segments. In contrast, others were interested in describing children's PA during week/school days and during the weekend, contrasting these days with the aim of understanding if the child's PA varied according to the days of the week. Further, others were interested in analysing children's PA in different seasons, usually fall/winter season and spring/non-winter season.

Accelerometry Data
Thirty-nine studies used accelerometers to assess PAP, from which 32 relied solely on accelerometers, while 7 studies employed them in combination with another instrument (Table 2). In general, children were monitored from 3 days to 2 weeks consecutively (in seven of the studies, children were monitored more than once [8,9,29,[52][53][54][55]), and most of the data were collected during children's awake period only; only in 4 studies [32,34,50,56] did children wear the accelerometer for 24 h/day.
When PAP was described across the day, the following approaches were mainly used: (1) describing PA per hour or the average of the measured days (at different intensities); (2) comparing PA levels taking into account different segments, namely time spent at school (during classes, physical education classes, breaks) and time out of school (before and after school, transportation mode to/from school; activities done during leisure time, activities at home, parks, sports participation); (3) describing the amount of time children spent in MVPA, and even the percentage of children complying with MVPA guidelines; (4) time spent in sedentary behaviours. Further, there was a systematic use of mean/average time in PA from the valid/measured days alongside the studies, with few studies focusing on describing PAP daily. However, when this approach was used, authors focused their efforts on children's MVPA (the time spent in each day, as well as the number of days they comply with the PA guidelines) [12,35,41,42,44]. In addition, a large number of papers also focused on comparing PAP between weekdays (or school days) and weekend days, considering the average value of the days (Monday to Friday for weekdays; Saturday and/or Sunday for weekend days) [8,12,14,29,30,33,39,41,[43][44][45]47,48,50,55,85,86], and two studies investigated the effect of an intervention program on children's PAP comparing differences in mean minutes children spent in different PA intensities and sedentariness/rest [52,54]. None of the studies investigated children's PAP by day and/or by hour at the same time (when authors focussed in daily PAP, the daily averages were used; while when the focus was on hourly PAP, averages of each intensity of PA per each hour were used).
The use of questionnaires, in association with accelerometers, allowed some authors to identify the travel mode children used to/from school/home [85,86], the type of activities usually engaged in or the activities they performed while the accelerometer was used [87], as well the estimation of the energy expenditure, based on activities recorded in the diary [88], and also the time spent in sedentary behaviours as well as the kind of sedentary activities typically engaged in [14]. Additionally, the combined use of GPS with an accelerometer provided information regarding the places and contexts that children tend to engage in PA, allowing these to be associated with PA intensities [6]. Furthermore, when accelerometers were used in combination with heart rate monitors [89], information from this last instrument was used as cut-point to classify children's PA levels considering the maximum age-related heart rate.

Pedometry Data
Pedometers were used in only 9 studies (seven [22,[57][58][59][60][61][62] of them used only the pedometer, and two [90,91] employed pedometers in combination with a questionnaire) ( Table 3). The pedometers were worn between four and seven days, only during awake periods, and in one of the studies, children were monitored across one academic semester (i.e., by using the pedometer for 7 days per month).
Similar to the accelerometer studies, authors were mainly focussed on describing PAP according to the number of steps children achieved in a typical day, using the average number of steps [22,[57][58][59][60][61][62]90,91] or classifying children according to some daily recommendation [61,90]. One study focussed only on school time, meaning that the pedometer was just used during the school period (from 8:00 to 15:00) [22].
Given that there is no universal step cut-point to classify children into different PA levels, few studies used this approach. In fact, one study classified children as low or high active, taking into account the median of their daily steps as the cut-point [90], while another used an existing cut-point recommendation [61]. The combined use of questionnaires provided authors information regarding children's travel mode [90] as well as time (hours/day) spent in different physical activities and sedentary behaviour [91].

Heart Rate Monitoring
Heart rate monitors were used in 11 studies [63][64][65][66][67][68][69][70][71]92,93], but only two used them concomitantly with a questionnaire [92,93] (Table 4). The timeframe for heart rate monitoring ranged from 3 h to 4 days. Only one study monitored children's heart rate 24 h/day [69], and in another study, children were also monitored during the weekend (two weekdays and one weekend day) [93]. Differences in PAP between seasons were also investigated in one of the papers (autumn and summer) [63].
Heart rate was also used to classify children according to their PA levels, although cut-points varied across the studies. Generally, authors reported the time, or percentage of time, children spent in different intensity activities (low, medium, high, or MVPA) across the monitored time, and the PAP was described hourly (when children used the monitor during one day [71,92] or the average value was used [63][64][65][66][67][68][69][70][71]92,93]. In addition, and when used in combination, questionnaires provided additional information regarding the type, time, period, and place of activity [92,93].
The information derived from questionnaires was related, in general, to children's PA engagement on typical days. Often, participants were asked to report time spent in activities of different intensities (light/low, moderate, vigorous, and sedentary) in the previous month [77], or in a typical week [11,27,74,75,79,80], or in the last week [74,76,78], or in a typical day [10,72,73,81], or even in periods out of school (leisure time), as well as their current PA level [78]. Furthermore, in some studies children were also asked to report the estimated time spent in screen entertainment or other sedentary behaviours [10,11,[72][73][74][75][76]78,81]. When parents were responsible for answering the questionnaire, their information was related to the time (minutes/hours), and/or frequency (days) with which their children were involved in PA of different intensities. In some reports, the information also allowed the estimation of MVPA, which was used as time (minutes/day [10,11,27,[75][76][77], or minutes/week [80]) or to classify children as having complied, or not, with the MVPA guidelines [11,72,77], and also by an estimation of metabolic equivalent [81], or even the use of a score derived from the questionnaire [83].
Given the substantial range of questionnaires used to assess PA and that the information that they provide varies substantially, it is not easy to cluster the different strategies that authors have used to determine PAP from questionnaires. However, independent of the output variable used, selected papers usually focussed on expressing PAP according to time or frequency spent in PA, compliance or not with MVPA guidelines, average differences in PA and sedentariness among different segments of the day (school segments and out of school), or between week and weekend days, travel modes, or even the child's participation in sports.

Observation
Only two studies used an observational approach to study children's PAP (Table 5). To describe PAP, Berman et al. [23] followed an observational protocol of 12 h/day divided into four-hour time blocks (8:00-12:00; 12:00-16:00; and 16:00-20:00), which were further divided into consecutive 30 min time blocks. For each one of these blocks, children's PA categories and intensity were coded, and an estimation of VO 2 (mL/min/kg) computed. Then, bouts of different intensities of PA were categorised as being "high" or "low" based on whether the estimated VO 2 was above or below anaerobic or lactate threshold during the observational period. Within the context of school physical education classes, Corbin and Pletcher [84] described PAP according to children's involvement in unorganised, low organised, and organised play situations during classes. An activity index was derived that allowed for an estimation of energy expenditure, as well as the percentage of time that children spent in different activities during physical education (sitting, walking, etc.). This approach was also used to determine the percent of time that children were actually active.

Statistical Procedures
Most of the reviewed studies did not explicitly describe the use of a formal method to determine children's PAP. From the few studies where this information was presented, the methods included profile analysis [71,92], cluster analysis [14,23], spectral analysis [23], principal component analysis [74], and ratios [40].
Profile analysis was used to compare differences between sex [92] or between experimental and control groups [71] across pre-determined heart rate categories. Berman et al. [23] used two procedures to determine PAP (cluster and spectral analysis), where the spectral analysis was used to determine the existence of recurring patterns of activity bouts in each 24-min period, while the cluster analysis allowed the authors to identify the occurrence and duration of these bouts. Cluster analysis was also used by Jago et al. [14] to identify groups of children with similar behavioural profiles regarding their PA and sedentary behaviours.
Principal components analysis was the method chosen by Antonogeorgos et al. [74] to obtain children's PAP, based on the inter-correlations between PA variables, where PAPs were apparently revealed. Using a different approach, Loprinzi et al. [40] described PAP according to ratios of different PA markers, namely MVPA/sedentary behaviour, light PA/sedentary behaviour, and total PA/sedentary behaviour, where ratios ≥ 1 implied that children were relatively more engaged in MVPA, light PA, or total PA than in sedentary behaviour.
Regarding body composition, only three studies investigated associations with PAP. Here, disparity existed between the reported results. Al-Nakeeb et al. [93] did not find significant relationships between body fat percentage and time spent in MVPA, whereas Loprinzi et al. [40] described significant differences in different body composition indicators [body mass index (BMI); waist circumference; triceps and subscapular skinfolds; android, gynoid and body fat percentage] between children who met the MVPA recommendations and engaged in more light PA than sedentary behaviour versus those who did not comply with the MVPA guidelines and engaged in less light PA than sedentary behaviour. In contrast, Bosch et al. [82] reported that actively commuting children were less likely to have high fat mass than their passive commuters peers, and those who were engaged in sports less than once a week were less prone to have high fat-free-mass when compared to daily active children.
The relationship between PAP and cardiovascular/metabolic risk markers was investigated in only two studies [51,78]. In the first, Schmidt et al. [78] used children's self-reported PA (actual and regarding the past 14 days) and categorised children into five groups based on their actual PA (i.e., inactive; relatively inactive; light PA; moderate PA; and vigorous PA). The authors also took into account the number of days that the children were involved in PA of different intensities (hard exercise, easy exercise), multimedia usage, and their annual sports participation. Results showed significant and negative associations between PAP and markers of cardiovascular, yet there were different by sex; i.e., in boys, PA was correlated with total cholesterol and triglycerides; whereas in girls, correlations were observed for body fat percentage and BMI. In Aadland et al.'s study [51], with accelerometry information, the association of children's PA volume and patterns with metabolic risk factors was investigated. It was concluded that there was a strong negative association between metabolic health and vigorous PA, yet this association was weak with moderate and light PA, and no association was observed with sedentariness. Further, it was reported that the association between metabolic health and PA seems to be determined by accelerometer epoch settings, whereby short epochs appeared more favourably associated with metabolic health than long epochs. Regarding the relationship between PA and bone mineral density, the only study included in the present review addressing this purpose did not find any significant association [83].
The majority of the reviewed studies focussed on the possible links between PAP and nutritional status. Here, results were shown to be divergent. Five [28,66,67,73,90] of the 18 studies did not find any significant association, meaning that children's PAP (levels/intensity) did not differ in groups with different nutritional statuses. It seems important to note that despite most of the studies focussing their attention towards understanding this relationship in children with normal weight and those with excess weight, Benefice et al. [66,67] sampled malnourished children and compared their PAP with normal weight children's PA and reported no differences. On the contrary, other studies reported PAP-related differences between normal weight children and their overweight/obesity peers, favouring the normal weight group (e.g., they tend to be more active or comply more with the MVPA guidelines). Further, some sex differences in this association were also observed. For example, Butte et al. [34], Sigmund et al. [89], and Williams et al. [50] reported this difference only among boys (in girls, no significant association was observed), but a slightly difference in methods/results in these studies should be highlighted. While the first two [34,89] investigated this relationship in normal weight and overweight children, the last study [50] focussed on malnourished children, with results revealing that stunted/severely stunted boys were less sedentary on both school and non-school days.

Discussion
The purpose of this review was to provide an extensive search on PAP. Herein, our review encompassed (1) the putative definitions of PAP; (2) the instrumentation and techniques used to determine PAP; (3) the statistical procedures often used to analyse PAP; and (4) the associations among PAP and child health. Overall, the reviewed papers presented a moderate methodological quality (≥ 50%). Responses to Q3 were endorsed with the lowest score, meaning that none of the studies presented a clear definition regarding the meaning of PAP. However, the general mean score observed for Q4 (1.47) and Q7 (1.39) revealed that studies clearly presented the strategies used to measure PAP (notwithstanding, it was not clear what PAP means in the studies), as well as the conclusions synthetised the major results, with most of them indicating the implications/relevance of the studies, and suggesting directions for future work.
Notwithstanding the number of published papers concerning PAP, there is apparently no consensus regarding PAP's real substratum. According to the Longman dictionary (p. 1120) [94], a pattern can be defined as "the regular way in which something happens, develops, or is done". However, as outlined in this review, most papers did not intensively examine children's PA at an intra-personal level, but typically focussed on describing/explaining/understanding their PA levels, i.e., their summaries, comparing groups and conditions, or even describing frequencies with which PA guidelines were achieved. In this context, a review by Welk et al. [95] presented children's PAP according to its "highly transitory nature". Here, the authors argued that PAP during childhood is better reflected by the accumulation of daily intermittent activity as opposed to being continuous in nature. The authors also reinforced that frequency, intensity, and duration are variables that are commonly used to characterise PAP. Welk et al. [95] also highlighted that a clear PAP definition is necessary, allowing researchers to investigate similar variables and, more importantly, avoiding the imprecise use of this term.
The instrumentation used to capture putative measurements of PAP varied substantially across studies, and all available "arsenal" was used. In fact, accelerometers, pedometers, heart rate monitors, questionnaires, and observational methods were employed. This means that, irrespective of the ways each study aims were formulated, there is no recognised agreement upon a standard instrument with which PAP could be best captured. As previously mentioned, it also reflects the absence of a unique and clear definition of PAP. For example, when PAP was investigated on a daily based routine, with the purpose to identify how PA changes between and/or within days, accelerometers or pedometers were the most used instruments. However, when researchers were interested in describing different types of activities children are usually engaged in, as well as the places where these took place, questionnaire or observational methods were used. As suggested by Welk et al. [95], given the unique aspects of children's movement patterns, these patterns influence the existing measurement techniques to determine PA, leading to a wide range of instruments used to measure PA in children, all with different strengths and weakness [96].
Given that in the most of the reviewed papers, instruments were usually used to describe children's PA levels (which has being used as a PAP definition), it would be expected that authors did not use specific statistical analyses procedures to determine PAP. In fact, few studies [14,23,40,71,74,92] mentioned the use of specific procedures to determine/analyse PAP. Yet, even in papers where specific procedures were mentioned to be used, it was possible to observe that the PAP definition was not always clear. For example, when profile analysis was used, the authors' aim was to compare differences between groups [71,92], and using cluster analysis allowed them to group children according to their behaviour similarity [14]. On the other hand, using ratios provided information regarding the proportion of the time a child is engaged in more or less PA from different intensities in comparison with time in sedentary behaviours [40]. Despite the authors' efforts to determine children's PAP, most results do not reflect what PAP should mean, i.e., the pattern of a behaviour that ought to take into account the individual child streams of PA behaviours, rather than compare it with others or even to describe if, in the total amount of the time, the child is more or less active than sedentary.
The last aim of this review was to summarise the available research concerning the relationship between PAP and health markers. It is well-known that MVPA is positively associated within numerous physical and psychological health benefits [1]. Indeed, based on such evidence, recognised international guidelines suggest the minimum amount of time children and adolescents should be active on a daily basis to improve their health [2]. Most of the studies linking PAP and health were more interested in analysing the relationship between PA levels and health (or health indicators), but not PAP per se. PA markers such as MVPA, different intensities of PA children are engaged in, and the compliance with PA guidelines were used to associate PAP with health indicators. This suggests that information is still lacking on how PAP may influence the physical and psychological health and well-being of children.
It is now evident that children who are more engaged in MVPA tend to be healthier, with a better metabolic profile than those who are less active [1]; yet, what it is not clear is how this relationship stands when considering different patterns. For example, is it best to: (1) be engaged in 60 min of continuous MVPA or in various time fractions alongside the day or (2) have a more erratic PAP profile (i.e., with random bursts of MVPA)? Such questions could probably be answered if a consensus existed on the exact meaning of PAP, which in all likelihood would guide research in a more enriching and impactful way.
The present review is not without limitations. First, we only considered children aged between 6 and 11 years; yet, including adolescent and/or adult data would probably be too extensive to synthesise in a single review. Second, the fact that we did not use other possible definition/terms during our search that may be used as a "physical activity pattern" synonym.

Conclusions
In conclusion, this systematic review shows that there is no consensus regarding a clear PAP definition whatever the instruments used to capture it. Furthermore, there apparently is no agreement on how best PAP should be analysed. Hence, there is an "urgency" for a formal and clear definition of PAP in order to guide researchers in future studies, highlighting the most useful instruments and statistical procedures to best capture these complex streams of behaviours. We then suggest that PAP can and should be used when aiming to probe similarities/dissimilarities and stabilities/changes in children PA at an intra-personal level, such that their streams of activity/sedentary behaviours may be searched for differences, or not, across a variety of conditions that apparently rule their lives, meaning that PAP should be used to best describe individual streams of personal behaviours, and not only PA intensities.
We suggest that physical activity research should focus on providing precise answers to putative links between different PAP and health risk profiles as well as health benefits. Furthermore, public health policies and guidelines should also consider these links rather than exclusively focus on average daily minutes. For example, how were these 60 min achieved? Do children maintain the same pattern of activities across the entire week, or do they vary in activities and intensities? Should they do intense or moderate activities during breaks, spread along the day (and if so, what is the duration of these breaks and in which periods of the day should they occur)? Yet, it is also important that school, families, and governments provide children opportunities to be safely active, during different moments of their days.
Author Contributions: T.N.G. conceptualised and designed the study, conducted the online search, undertook the data analysis and interpretation (papers selection and screening, and data quality), and led the writing of the manuscript. S.P. performed papers selection and screening and contributed to drafting the manuscript. M.T. performed the update of the papers to be included in the review, undertook the data analysis (data quality of the papers) and contributed to drafting the manuscript. P.T.K. and M.S. contributed to results interpretation and to drafting the manuscript. J.M. supervised the conceptualised and designed the study, and contributed to drafting the manuscript. All authors have read and agreed to the published version of the manuscript.
Funding: This research received no external funding.

Conflicts of Interest:
The authors declare no conflict of interest.