Temporal Trends in Children’s School Day Moderate to Vigorous Physical Activity: A Systematic Review and Meta-Regression Analysis

Abstract

Background:

Evidence from a limited sample of countries indicates that time for physical education and recess during school have declined. Schools are called to provide children with 30 minutes of moderate to vigorous physical activity (MVPA). This systematic review and meta-analysis estimated temporal trends in children’s school day MVPA.

Methods:

Three online databases were searched to identify studies with objectively measured MVPA, during school hours, in school aged children (5–18 y). Multilevel random-effects meta-analyses estimated MVPA by year, and meta-regression analyses estimated temporal trends in school day MVPA.

Results:

Studies (N = 65) providing 171 MVPA estimates, representing 60,779 unique children, from 32 countries, and spanning 2003–2019 were identified. Most studies were conducted in North America (n = 33) or Europe (n = 21). School day MVPA ranged from 18.1 (95% confidence interval, 15.1–21.1) to 47.1 (95% confidence interval, 39.4–54.8) minutes per day in any given year. Meta-regression analyses indicated that MVPA declined from 2003 to 2010 (approximately 15 min decline), plateaued from 2010 to 2015 (approximately 1 min decrease), and increased from 2015 to 2019 (approximately 5 min increase).

Conclusions:

School day MVPA decreased from 2003 to 2010 and has recently begun to increase. However, the majority of the evidence is from North America and Europe with some evidence from Oceania and very little evidence from Asia to South America.

Physical activity guidelines from around the world and from the World Health Organization call for school-aged children to engage in 60 minutes of moderate to vigorous PA (MVPA) each day.^1-4 Regular engagement in PA has been shown to improve children’s physical and social-emotional health, cognitive function, and academic achievement.^1,5 Furthermore, regular engagement in PA plays a protective role against noncommunicable diseases, such as cardiovascular disease, type 2 diabetes, obesity, and cancer.^4,6,7

Schools are a prime setting to provide children with health-enhancing levels of PA because most children around the world spend the majority of their waking hours at school,^8,9 schools have the infrastructure (eg, space, teachers) in place for children to be active, and there are a variety of traditional school day segments (eg, recess, physical education) that can be capitalized upon to provide children with opportunities to be active. In addition to traditional school segments to provide activity, there is growing work surrounding infusing PA into general education classroom time through movement integration.¹⁰ In addition, schools can provide children with PA outside of the school day through active transportation and intramural/interscholastic sports.

One of the earliest calls for schools to provide children with adequate levels of PA was made by Sallis and McKenzie¹¹ in an article published in 1991 in Research Quarterly for Exercise and Sport. Over the last 30 years, a number of organizations from around the world have echoed Sallis and McKenzie’s call to action.^9,12-15 Most recently schools have been encouraged to provide children with 30 minutes of MVPA during school hours.^16-18 A recent systematic review synthesized studies that examined objectively measured PA of school-aged children.¹⁹ This review identified a total of 91 studies published from 2003 to 2018. This study found that children fell just short of the 30 minutes per day threshold, accumulating 27.8 minutes of MVPA during the school day. However, this review was limited because it did not examine temporal trends in the accumulation of MVPA during the school day.

Despite limited empirical evidence, there is a common narrative in the literature that the amount of MVPA, and PA of any intensity, provided for children during the school day has declined over the last 2 decades.^16,20-25 This perception may be warranted as studies have shown that traditional opportunities for children to accumulate PA during the school day have declined.^20,26 A growing number of laws and regulations have mandated that schools provide requisite amounts of physical education and PA opportunities.^25,27-29 However, research has shown that schools commonly do not meet these standards likely because of limited accountability and a lack of consequences for not meeting standards.^30,31 A decline in opportunities is also likely linked to the marginalization of physical education as a subject and the fact that physical education and PA opportunities are not valued by some administrators.^32,33 However, to date, no studies have examined the temporal trend in children’s accumulation of MVPA during the school day. Thus, the purpose of this systematic review was to examine the temporal trends of children’s objectively measured MVPA accumulated during the school day.

Methods

Search Strategy

This systematic review and meta-analysis was designed, conducted, and reported in accordance with the Preferred Reporting Items for Systematic Review and Meta-Analysis statement³⁴ and Meta-analysis of Observational Studies in Epidemiology guidelines.³⁵ The current study builds upon and extends a systematic review published elsewhere¹⁹ that examined children’s accumulation of PA in a variety of structured settings including during the school day. As such, the search strategy was conducted in 2 phases with the original study conducting searches from October 2018 to February 2019 and included studies published through October 2018. Following the publication of the original study, the same search protocols were adopted and executed in January 2021 in order to update the systematic review to include studies published through December 2020. The original study was prospectively registered in the International Prospective Register of Ongoing Systematic Reviews under the number CRD 42018111804. The protocols followed in each search are presented below.

A comprehensive search across MEDLINE via PubMed, Scopus, Web of Science, and Cochrane was conducted. Searches included 4 categories that were combined using Boolean operators. The original search was completed by the second author (R.M.T.), and the follow-up search was completed by the first author (R.G.W.) in consultation with the second author (R.M.T.). The categories included outcome (ie, PA, MVPA); setting (ie, school); measure (ie, objective measure, accelerometer); and population (ie, children, adolescent). All search results were imported into EndNote X7 (Thompson Reuters, San Francisco, CA) where duplicates were removed. As reported in the original study,¹⁹ the following steps were then completed: (1) titles and abstracts were screened by 3 independent reviewers (R.M.T., M.C.M.T., and R.G.W.) to identify potential articles based on the review question; (2) studies that did not meet the eligibility criteria were removed; (3) full-text papers of potentially eligible studies were assessed; (4) the references of all included studies were reviewed to identify additional studies; (5) consensus on all full-text papers included/excluded was reached via weekly group discussion with all authors (R.M.T., R.G.W., M.C.M.T., K.B., and M.W.B.). Data including title; year of publication; authors; year/s of data collection; country and global region sample size; sex; age; race; and socioeconomic status, measurement tools and protocol (manufacturers and data reduction procedures), and MVPA estimates during the school day were extracted. Where possible and when necessary total minutes of MVPA accumulated during the school day were calculated when other metrics were provided (ie, minutes per hour, percentage, and total wear time or total time attending school). If necessary, the authors of included studies were contacted by e-mail to provide additional information.

Eligibility Criteria

Inclusion Criteria.

To be included, studies had to be published in English in a peer-reviewed scientific journal, present estimates of MVPA, measured via an ActiGraph accelerometer (ActiGraph, LLC, Pensacola, FL), during the school day, on children (5–18 y). Studies were required to use Pate et al,³⁶ Puyau et al,³⁷ Trost et al,³⁸ Van Cauwenberghe et al,³⁹ Evenson et al,⁴⁰ or Freedson et al,⁴¹ where the MVPA threshold was either 3 or 4 metabolic equivalent of tasks,^41,42 cut points to distill their data. The decision to focus on actigraph-estimated MVPA from one of the cut points above was made in order to allow for harmonization of the MVPA estimates via the Rosetta Stone equations.^43-45 A major issue in the field of PA measurement is the fact that widely disparate estimates of PA can be produced simply based upon the measurement techniques chosen.^46,47 This concern is personified by the issue of cut point nonequivalence (ie, activity intensity estimates vary between studies investigating the same population primarily because of the cut points chosen for data distillation).^48,49 In addition, the type of measurement tool chosen can greatly influence estimates of MVPA. For example, a recent systematic review of MVPA during physical education showed that estimates of MVPA are 25% higher (57.6% vs 32.6% of physical education spent in MVPA) in studies that used systematic observation as a measure when compared with accelerometer-estimated MVPA.⁵⁰ This is a substantial concern when attempting to estimate temporal trends in estimates of activity as an estimate could be artificially inflated or deflated for a given year simply because of the cut points or measurement device used in a single study. Thus, the decision was made to limit the review to studies that could be harmonized via the Rosetta Stone equations.

Exclusion Criteria.

Articles were excluded if they did not present MVPA estimates during the school day or if MVPA estimates for school day PA could not be estimated (eg, minutes per hour of MVPA along without length of school day). Studies that used accelerometer brands other than actigraph, other measurement techniques to estimate MVPA (eg, self-report, heart rate monitor, pedometer), and/or cut points other than those included in the Rosetta Stone equation study were also excluded. Studies that included children who were institutionalized, in clinical settings, or with disabilities only were also excluded.

Risk of Bias Assessment

Quality of each individual study was assessed using a risk of bias tool created for this study. The risk of bias tool consisted of 11 items and followed a root and stem style. All items were proceeded by the root “The study presented and adequately described . …” The stems included “… the design (ie, cross-sectional, RCT, longitudinal)”; “… the number of settings, classrooms, and/or sample size”; “… the demographic characteristics of the structured setting (ie, geographical location, and structured setting hours)”; “… the sample characteristics (ie, sex, age, socioeconomic status, and racial/cultural background)”; “… the brand and model of objective-measured device and device placement”; “… the physical activity/sedentary time objective-measured protocol (ie, number of days and hours using)”; “… the physical activity/sedentary time protocol measured comprising the structured setting hours”; “… the valid days and valid hours criteria”; “… the non-wear time criteria and epoch length”; “… the study presented the cut point for physical activity/sedentary time”; and “… the average of valid wear-time during setting.” Each item was evaluated on a scale of 0 to 2 with 2 indicating that the item was “presented and adequately described,” 1 indicating that the item was “not clearly described or presented,” and 0 indicating that the item was “not reported.” The final scores ranged from 0 to 22. All studies were assessed by 2 of the authors (R.M.T. and M.C.M.T.) independently. A third author (R.G.W.) was consulted if there was a disagreement on a rating item.

Study Characteristics

Data Extraction.

All individual MVPA estimates and study characteristics reported in the eligible studies were extracted by R.G.W. and R.M.T. and entered into a customized Excel spreadsheet (Microsoft Corporation, Redmond, WA) developed for this study. Once all data were extracted and the reported data were transformed into the metrics of interest (eg, SEs transformed into SDs, minutes per hour transformed into total MVPA accumulated during the school day), data were entered into Stata (version 14.1; StataCorp LLC, College Station, TX) for data analysis.

Data Analysis.

All analyses were performed in Stata, (version 14.1; StataCorp LLC) using a multistep process. First, all estimates were converted into Evenson et al⁴⁰ cut points using the Rosetta Stone equations.⁵¹ Evenson et al⁴⁰ cut points were chosen as studies have shown that they provide the best prediction, specificity, and sensitivity for children 5–15 years of age.^40,52 Furthermore, Evenson et al⁴⁰ cut points are widely used in the literature. Second, where necessary, estimates were converted into minutes of MVPA accumulated during the entire school day (eg, percentage of MVPA, minutes per hour converted into daily minutes). Third, meta-analytic estimates of the MVPA accumulated during the school day were estimated by data collection year using the meta summarize command in Stata. Following the overall estimates, separate meta-analytic estimates were calculated by year for girls, boys, and all children (studies that did not separate out estimates by sex), school level (primary and secondary school), as well as geographic region. The I² index was calculated to identify the heterogeneity. Values of 25%, 50%, and 75% were considered to indicate low, moderate, or high heterogeneity, respectively.⁵³ Finally, the temporal trend of school day MVPA estimates was estimated via meta regression. To account for the fact that effects were nested within study, the robumeta command was used to estimate robust SEs.⁵⁴ All models included MVPA as the dependent variable and year as the independent variable. Where appropriate, squared or root terms were included for time. All models included mean age of participants, mean participant accelerometer wear time, sex, and, consistent with our past review,¹⁹ geographic region where the study was conducted as covariates. Following the overall estimates, separate meta-analytic estimates using the same robumeta command were calculated by year for girls, boys, and girls and boys combined (studies that did not separate out estimates by sex), and by school level (primary or secondary). Meta regression was also used to estimate temporal trends in MVPA accumulated during the school day by geographic region; however, analyses were not completed for Oceania, Asia, or South America because these regions included 8 or fewer unique studies covering 6 or fewer years.

Results

A Preferred Reporting Items for Systematic Review and Meta-Analysis diagram for the literature search is presented in Figure 1. Overall, the search yielded 8871 potential articles for inclusion. After duplicates were removed, 7096 title and abstracts were screened with 6095 excluded. A total of 1001 full-text articles were retrieved and reviewed with 836 articles excluded for a variety of reasons (Figure 1). This left 65 total articles for inclusion in the meta-analysis (Table 1) with a total of 171 MVPA estimates representing 60,779 unique children. Studies included data collected from 2003 to 2019. The breakdown of the number of studies included by world region and country is presented in Figure 2. Of the 65 studies, 50.7% (n = 33) were conducted in North America, 32.3% (n = 21) in Europe, 12.3% (n = 8) in Oceania, 6.2% (n = 4) in Asia, and 4.6% (n = 3) in South America. A total of 49.2% (n = 32) of the 65 total studies were from the United States, followed by Australia with 7.7% (n = 5) of 65.

Figure 1 —

Prisma flow diagram.

Table 1.

Characteristics of the Included Studies

Author	Country	Year/s of measure	Setting hours	Mean wear time, min	Study design	Sample size (at the school, classroom, student levels when reported)	Sample characteristics (age, sex, socioeconomic status, and race)
Aadland et al55	Norway	2014	9 AM–2 PM	295	Cross-sectional (nested RCT)	Schools (ASK project); n = 465	10.9 y; 52% girls; NR; NR
Aibar et al56	France (F)	2011	8 AM–5 PM (F)	390.6	Cross-sectional	10 schools n = 711	14.3 y (F); 55.2% girls (F); 6.99 (0–9) SES (F); NR (F) 14.3 y (S); 45% girls (S); 6.57 (ranged: 0–9) SES (S); NR (S)
Andersen et al57	Denmark	2010	8 AM–2 PM	315.8	Longitudinal	4 schools; n = 316	10–14 y; 53.2% girls; NR; NR
Brittin et al58	United States of America	2011/2012	NR	406.7	Natural experiment	2 schools; n = 53	8.5 y, 52.8% girls; NR; 32% minority
Brusseau et al59	United States of America	2014	8 AM–3 PM	NR	Cross-sectional	3 schools; n = 395	8.4 y; 45% girls; 93.7% low income (school level); 86% minority (school level)
Burns et al60	United States of America	2016	8 AM–3 PM	NR	Cross-sectional	3 elementary school; n = 1049	8.4 y; 50.1% girls; NR; NR
Burns et al61	United States of America	2018	8 AM–3 PM	NR	Cross-sectional	5 schools; n = 2119	8.5 y; 49.1% girls; NR; NR
Burns et al62	United States of America	2019	8 AM–3 PM	421.7	Cross-sectional	5 schools; n = 435	8.4 y; 46.8% girls; 96% low income; 91% minority
Carlson et al63	United States of America	2013	NR	350	Quasi-experimental	6 district, 24 schools (97 classrooms); n = 1322	8.8 y; 53.7% girls; NR; 67.8% Latino
Carlson et al64	United States of America	2010	NR	478.8	Cross-sectional	317 census block groups in 2 US regions 942 households (subsample); n = 549	14.1 y; 49.9% girls; 64.7% had a parent with a college degree; 31.3% non-white
Carson et al65	United States of America	2010	NR	365.1	Quasi-experimental	16 schools; n = 351	11.7 y; 55.2% girls; 74.6% free or reduced lunch; 63.5% non-white
Centeio et al66	United States of America	2012	NR	397	RCT	20 urban elementary schools; n = 334	9.4 y; 57% girls; NR; 53% African American, 23% white, 2% Hispanic, and 20% other
Cinemre67	Turkey	2015	NR	415	Cross-sectional	1 elementary school; n = 40	8.3 y 0% girls; NR; NR
Cradock et al68	United States of America	2011	8 AM–3 PM	334	Quasi-experimental	6 schools; 26 classrooms; n = 393	10.2 y; 52% girls; NR; 59% black, 31% Hispanic, 7% Asian, 2% white, and 2% other
da Costa et al69	Brazil	2015	8 AM–5 PM	354.9	Cross-sectional	2 schools; n = 415	12.3 y; 54% girls; NR; NR
da Costa et al70	Brazil	2015	8 AM–5 PM	240	Quasi-experimental	2 schools; n = 270	12.6 y; 55% girls; Associação Brasileira de Empresas de Pesquisa score = 17.3; NR
Decelis etal71	Malta	2012	8:30 AM–2 PM	NR	Cross-sectional	54 schools (nationally representative); n = 769	10.8 y; 51.8% girls; NR NR
Engelen etal72	Australia	2009	9 AM–3 PM	336.5	RCT	12 primary schools; n = 206	6 y; 46.1% girls; The Index of Community Socio-Educational Advantage—1076 (980–1170); parents originating 35 different countries
Ensenyat et al73	Spain	2013	9 AM–1 PM and 3 PM–5 PM	360	RCT	7 health centers; 16 healthcare pediatric units; n = 88	9.7 y; 45.5% girls; NR; NR
Farmer et al74	New Zealand	2011	NR	NR	RCT	16 schools; n = 704	7.9 y; 51% girls; NR; 49% New Zealand, 16.8% Maori, 11.5% Pacific, 8.0% Asian, and 14.7% unknown
Fu et al75	United States of America	2019	9 AM–3 PM	360	Quasi-experimental	1 school; n = 16	7.1 y; 37.5% girls; NR; NR
Gao et al76	United States of America	2012	8 AM–3:30 PM	414.4	Cross-sectional	1 school; n = 138	8.1 y; 51.5% girls; NR; 87% white, 11.6% Hispanic, and 1.5% African American
Garn et al77	Australia	2014/2015	NR	360	Longitudinal	14 schools; n = 1767	13.0 y; 51% girls; NR; 62% English Australian while 17% Asian, 12% Middle Eastern, 6% South Pacific, and 3% other
Goh et al78	United States of America	2012	NR	NR	Quasi-experimental	1 elementary school; n = 219	8 to 11; 54.3% girls; NR; 57% white, 35% Hispanic, 5% Pacific Islander, and 3% other
Guinhouya etal79	France	2005	8:30 AM–4 PM	NR	Cross-sectional	3 elementary school; n = 93	10 y; 48.5% girls; NR; NR
Harding et al80	England	2008	NR	397.7	Longitudinal	27 secondary school (peach project); n = 363	12.0 y; 61.4%; 61.7% low SES; NR
Herrick et al81	United States of America	2009	NR	NR	Quasi-experimental	6 afterschool programs (school based); n = 100	10.4 y; 55% girls; NR; 53% Asian, 31% Latino, 3% white, 2% African American, and 11% other
Hubbard et al82	United States of America	2014	NR	385	Cross-sectional	13 schools; n = 453	9.1 y; 60.5% girls; 30.9% free or reduced lunch; 74.4% white, 10.6% Hispanic, 4.9% black/African American, 3.3% Multiracial, 2.4% Asian, and 0.9% Native American
Kallio et al83	Finland	2013–2015	NR	300	Longitudinal	9 schools; n = 970	12.5 y; 52.3% girls; NR; NR
Kim and Lochbaum84	United States of America	2012	7:30 AM–3 PM	450	Cross-sectional	1 public elementary school in a low SES neighborhood; n = 75	10.1 y; 61.5% girls; NR; 54.7% black, 29.3% Hispanic, and 16% other
Kulik et al85	United States of America	2015	NR	NR	Cross-sectional	6 schools; n = 347	9.4 y; 57.3% girls; NR; 52.3% African American, 20.6% white, 3.5% American Indian, 2.3% Hispanic, 1.2% Pacific Islander, and 20.1% other
Kwon et al86	United States of America	2010	NR	390	Cross-sectional	14 public schools; n = 538	10–12 y; 53% girls; NR; 28% white, 8.5% black, 40.1% Hispanic; and 23.4% no majority race
Lin et al87	Taiwan	2012	8 AM–4 PM	NR	Cross-sectional	1 school; n = 49	9.2 y; 0% girls; NR; NR
Long et al88	United States of America	2003–2006	8 AM–3 PM	NR	Cross-sectional	2003–2006 NHANES; n = 2548	7.0 y; 50.3% girls; 28.7% of 130% of the federal poverty level; 58.1% white, 15.8% black, 14.3% Mexican American, and 11.7% other
Madsen etal89	United States of America	2013	NR	480	Cross-sectional	7 schools; n = 156	9.8 y; 40% girls; 61% free or reduced price meals; 42% Latino, 32% Asian, 12% African American, and 14% other
Madsen etal90	United States of America	2015	NR	370	RCT	6 schools; n = 450	9.5 y; 50.8% girls; NR; 49.4% Latino, 14.9% Multiracial, 9.8% Asian, 6.7% black, 6.3% white, and 12.9% other
Martin and Murtagh91	Ireland	2017	NR	333	Cross-sectional	10 schools; n = 197	8.9 y; 50.2% girls; NR; NR
McLoughlin and Graber92	United States of America	2017	NR	354.9	Cross-sectional	1 school; n = 105	10.6 y; 56.2% girls; NR; 72.6% Hispanic, 14.0% mixed race, 7.1% white, 4.4% African American, 0.9% Asian
Mooses et al93	Estonia	2014	NR	NR	Cross-sectional	13 schools; n = 244	9.1 y; 52.7% girls; NR; NR
Nettlefold et al94	Canada	2005	9 AM–3 PM	366.6	Cross-sectional	9 elementary schools; n = 379	10 y; 52.1% girls; NR; NR
Piipari et al95	United States of America	2016	NR	330	Cross-sectional	3 elementary schools; n = 200	7.8 y; 53.5% girls; NR; NR
Rajala et al96	Finland	2013	NR	300	Cross-sectional	8 cities; 8 schools; 42 classes; n = 420	13.7 y; 53.1% girls; NR; NR
Ramirez-Rico et al97	Spain	2009	Primary 9:30 AM–4 PM high-school 8:30 AM–2:30 PM	374.5	Cross-sectional	11 schools (4 primary and 7 high schools); n = 367	12 y; 62.3% girls; NR; NR
Resaland et al98	Norway	2015	9 AM–2 PM	NR	RCT	57 schools; n = 1063	10.2 y; 47.9% girls; 32% upper secondary education (parents); NR
Riley et al99	Australia	2013	9 AM–3 PM	NR	RCT	8 schools; n = 240	11.1 y; 40.9% girls; NR; NR
Salin et al100	Finland	2017	NR	300	Cross-sectional	17 schools; n = 453	11.3 y; 56.9%; NR; NR
Sayers et al101	United States of America	2007	NR	NR	Cross-sectional	3 elementary schools; n = 77	8.3 y; 48% girls; 54.5% with annual family income ≥$60,000; 64.9% white, 10.3% Asian American, 5.2% black/African American, 2.5% American Indian/Alaskan Native, and 12.9% other
Schneider et al102	United States of America	2011	NR	NR	RCT	n = 126	11.0 y; 52.0% girls; NR; 48% Latino, 19% non-Latino white, 12% African American, 10% Asian, 11% other;
Sigmund et al103	Czech Republic	2012	8 AM–1 PM	280.7	Cross-sectional	6 primary schools; n = 338	9.5 y; 50.3% girls; NR; NR
Silva et al104	Portugal	2018	8 AM–5 PM	NR	Quasi-experimental	1 school; 2 classrooms; n = 49	11.7 y; 53% girls; NR; 96% white
Sprengeler et al105	Germany	2012	NR	415	Cross-sectional	4 schools (2 in middle-income area and 2 in high-income area); 27 classrooms; n = 207	8.5 y; 47.3% girls; 75.5% high SES (ISCED); NR
Stewart et al106	New Zealand	2017	NR	268.5	Cross-sectional	7 secondary schools; n = 76	14.7 y; 40.8% girls; NR; NR
Sutherland et al107	Australia	2014	NR	NR	RCT	46 low socioeconomic elementary schools; n = 989	10.1 y; 51% girls; low SES; NR
Taylor et al108	New Zealand	2011	9 AM–3 PM	343	Cross-sectional	16 primary schools; n = 441	8 y; 46.7% girls; NR; NR
Ting et al109	Singapore	2009	7 AM–3 PM	347.8	Cross-sectional	7 secondary schools; n = 225	14.0 y; 47.6% girls; NR; NR
Tyler et al110	United States of America	2015	NR	270	Quasi-experimental	8 elementary schools; 44 classrooms; n = 719	7.7 y; 48.0% girls; 88.0% free or reduced priced lunch eligible; 88.0% African American, 7.5% white; 4.5% other
Van Kann et al111	Netherlands	2017	8:45 AM–2:15 PM	NR	Quasi-experimental	10 primary schools; n = 117	9.2 y; 59.6% girls; NR; NR
Vetter et al112	Australia	2014	6 h	360	RCT	2 primary schools; n = 172	8.4 y; 52% girls; NR; NR
Viciana et al113	Chile	2019	8:00 AM–3:15 PM	435	Cross-sectional	4 schools; n = 19	13.4 y; 42% girls; NR; NR
Wang et al114	Taiwan	2015	7:20 AM–5:00 PM	600	Cross-sectional	4 schools; n = 470	14.0 y; 50.4% girls;
Weaver et al115	United States of America	2014	NR	409.7	Cross-sectional	7 schools (2 cities); 24 classrooms; n = 323	7.5 y; 53.5% girls; 62.1% free and reduced lunch; 33.1% Hispanic, 26.4% white, 27.8%black, 9.7% Asian/Pacific/Islander, and 2.5% other
Weaver et al116	United States of America	2015	NR	353	Quasi-experimental	8 elementary schools (rural district); n = 795	7.6 y; 50.5% girls; 91% free or reduced lunch; 86% African American
Weaver et al117	United States of America	2015	NR	370.8	Quasi-experimental	4 elementary schools; 9 classrooms; n = 229	7.1 y; 58.2% girls; 32.2% free or reduced lunch; 44.0% black, 41.1% white, and 14.9% other
Webber et al118	United States of America	2003	9 AM–2 PM*	NR	RCT	35 middle schools (6 cities); n = 1566	11.8 ye; 100% girls; 35.6% reduced or free lunch; 43.1% white, 21.5% African American, 20.2% Hispanic, and 15% other
Wells etal119	United States of America	2011	NR	355	RCT	12 schools; 5 regions; n = 124	9.3 y; 56.4% girls; 68.3% free or reduced lunch; 51.5% white, 30.0 African American, 8.8% Hispanic, and 9.7% Asian

Abbreviations: ASK, active smart kids; ISCED, International Standard Classification of Education; NHANES, National Health and Nutrition Examination Survey; NR, not reported; PA, physical activity; RCT, randomized controlled trial; SES, socioeconomic status; SEIFA, Socio-economic Indexes for Areas.

Figure 2 —

Estimates of MVPA by region and country. MVPA indicates moderate to vigorous physical activity.

Risk of bias ratings by study on individual indicators are presented in Figure 3. The overall score ranged from 12 to 22 with a mean score of 18.2 (SD = 2.3) and a median score of 18.5. Across studies, the number of schools, classrooms, and sample size; sample characteristics (ie, sex, age, socioeconomic status, and racial/cultural background); the protocol used to measure MVPA during the school hours; the valid days and valid hours criteria; the nonwear time criteria and epoch length; and the average of valid wear time during school was frequently not presented and adequately described. A total of 60 of the 65 included studies neglected to clearly state at least one of their accelerometer processing procedures (ie, mean wear time, nonwear criteria, epoch length, valid days, and hours criteria).

Figure 3 —

Risk of bias. RCT indicates randomized controlled trial.

The summary estimates of MVPA by year are presented in Figure 4 along with estimates of heterogeneity (I²). Individual study estimates along with pooled yearly estimates are presented in Supplementary Table S1 (available online). The meta-regression estimates are presented in Figures 5-7. Heterogeneity between studies within a year that included more than one study ranged from 4.2% to >95%. There was a statistically significant curvilinear trend in MVPA from 2003 to 2019 (P < .001). When considering all studies, MVPA during the school day declined from 2003 to 2010, plateaued from 2010 to 2015, and increased from 2015 to 2019. A similar pattern was observed for both Europe (P < .001) and North America (P = .001), for studies that estimated boys (P = .076) and girls (P = .014) separately, and for both primary and secondary school children, with MVPA declining in the early 2000s, plateauing from 2010 to 2015, and then increasing again. For studies that combined boys and girls MVPA, there was a slight increasing linear trend (P = .029) in MVPA from 2003 to 2019.

Figure 4 —

School day estimates of MVPA overall and by region, sex, and school level. MVPA indicates moderate to vigorous physical activity.

Figure 5 —

Bubble plot and meta-regression estimated temporal trends in school day MVPA overall by region. CI indicates confidence interval; MVPA, moderate to vigorous physical activity.

Figure 7 —

Bubble plot and meta-regression estimated temporal trends in school day MVPA by school level. CI indicates confidence interval; MVPA, moderate to vigorous physical activity.

Discussion

This systematic review examined the temporal trend of MVPA during the school day from 2003 to 2019. Findings indicate that early this century MVPA during the school day declined (from 2003 to 2010), then plateaued, and has recently begun to increase again. The findings of this systematic review do not support the narrative that school-aged children are engaging in less MVPA during the school day today compared with the prior 2 decades.^16,120 While data do show children’s MVPA during the school day declined in the early 2000s, recent studies show MVPA accumulated during school has increased, with most recent estimates indicating children engage in approximately 20 to 30 minutes of MVPA during the school day. It is worth noting that, as shown in Figures 4 and 5, early studies that indicated children were engaging in more MVPA during the school day in 2003 and 2005 were smaller and had more variability in their estimates, indicating that they may have been lower quality. This same pattern emerges when the estimates from North America and Europe are considered independently.

The only subgroup analysis where this U-shaped trend was not observed was for studies, which combined girls and boys MVPA data. In fact, there was a slight but consistent increasing trend in MVPA accumulation for this subgroup. This is likely due to the fact that the earliest study included in this subgroup analysis was conducted in 2007, after the declines in MVPA were observed in the overall analysis and the other subgroup analyses. This conclusion is also supported by the fact that trends of school day MVPA accumulation from 2007 to 2020 look nearly identical for studies that combined boys and girls when compared with studies that analyzed boys and girls separately. Thus, this finding may be an artifact of the limitations of the available studies.

It is possible that these findings reflect an increase in study quality and/or a change in the methods with which children’s PA was collected during the school day. For instance, the use of accelerometers was relatively novel in 2003, the first year included in the current review. A PubMed search of “children” and “accelerometer” yields 20 studies in 2003 and 383 studies in 2020, an almost 20-fold increase. However, risk of bias indicators does not indicate that study quality has improved from 2003 to 2020. For instance, the mean risk of bias score by year ranged between 14 and 20 and did not show an increase over years. In fact, the 2 years with the lowest mean risk of bias scores (ie, indicating high risk of bias) were 2012 and 2019. Furthermore, the current study employed the Rosetta Stone equations to harmonize MVPA estimates.⁴³ Although there is a small degree of error when applying the conversion equations, applying these conversion equations minimizes the possibility that changes in MVPA estimates over time are due to the application of different accelerometer cut points in different studies.

Reasons for the decline in MVPA in schools in the early 2000s may be partially driven by global educational policy discourse that was increasingly focused on standards-based instruction and high stakes testing in schools.^121,122 For instance, the No Child Left Behind Act was passed in the United States in 2002.¹²³ This law tied federal funding for schools to student achievement on standardized testing. Similar trends of increased high stakes testing and national assessments that were tied to national curriculum standards, teacher pay, and school funding were instituted in the United Kingdom and Australia¹²⁴ and in other countries around the world¹²⁵ at the turn of the century. As a result, opportunities for students to attend physical education and recess were increasingly removed from schools^9,26 as schools prioritized academic time in an effort to maximize students’ performance on these high stakes, standardized tests. It is unclear why recent estimates show an upward trend in children’s MVPA during the school day. One plausible explanation is that the efforts of public health professionals, researchers, and administrators to introduce PA interventions in schools may be having an impact. For instance, a number of large-scale PA interventions have been tested and disseminated in schools in recent years.^126,127 Furthermore, in the United States, the Healthy Hunger Free Kids Act of 2010¹²⁸ required that schools create a wellness plan, and as part of this plan, schools were required to articulate strategies for increasing PA of children during the school day. Similarly, the European Union has endorsed similar policies for schools in its member states.¹²⁹ These efforts may be generating the intended effect of increasing children’s PA during the school day.

A notable finding of this systematic review is that the vast majority of studies were conducted in North America or Europe (54 of 65 total studies). This is not surprising given that North America and Europe consist of mostly high-income countries and are responsible for the bulk of health research.^130,131 However, this is troubling as North America and Europe make up only 14.3% of the world’s population,¹³² and PA trends have been shown to differ by country and region.^133-136 The lack of studies from other regions may be related to the higher concentration of middle- to low-income countries in these regions. It has been noted that middle- to low-income countries lack the research infrastructure to conduct health research studies. For example, Africa has 198 researchers per million inhabitants while the United Kingdom and the United States have more than 4000 per million inhabitants.¹³⁶ The dearth of studies from middle- to low-income countries is serious limitation of the literature to date.

The limited number of studies from middle- to low-income countries included in this review indicate that children in these countries are accumulating less MVPA during the school day. For example, 3 studies were included from Brazil with these studies indicating that children accumulated 12.2 minutes of MVPA during the school day.^69,70 This is considerably lower than the estimates of MVPA from the high-income countries included in this review. However, this finding is likely nuanced by the differences in the schools themselves, their surrounding neighborhoods, and the local and national policies that govern them. It is imperative therefore to overcome barriers to conducting research in middle- and low-income countries in order to gain a full understanding of the trends of school day PA of children around the world. In order to do this, similar to the SUNRISE study of 24-hour movement behaviors in preschoolers,¹³⁷ future research should make a concerted effort to conduct studies in underrepresented and frequently middle- and low-income countries located in Oceania, Asia, and South America. The recently established Active Healthy Kids Global Alliance¹³⁸ may be one framework that could facilitate objective measurement of PA in schools in these regions.

Another limitation of the current literature is the multitude of ways in which MVPA, and other PA intensity levels are measured. This is a major limiting factor to the field of PA measurement because disparate estimates of PA can be produced simply based upon the measurement techniques chosen.^46,47 This concern is personified by the issue of cut point nonequivalence (ie, activity intensity estimates vary between studies investigating the same population primarily because of the cut points chosen for data distillation).^48,49 This is a substantial concern when attempting to estimate temporal trends in estimates of activity, and this concern is amplified when there are very few studies in any given year (ie, an estimate could be artificially inflated or deflated for a given year simply because of the cut points or method used in a single study). For this reason, this study only included MVPA estimates produced by ActiGraph accelerometers that used a limited number of cut points because harmonization techniques exist across these estimates.⁴⁵ However, there are no comparable methods for harmonizing estimates produced by different objective measurement tools (eg, heart rate monitor, pedometer, systematic observation). Furthermore, there are no methods for harmonizing estimates of different PA intensity levels (eg, sedentary, light, moderate, vigorous, or total PA). Thus, as suggested by a panel of experts in PA measurement,⁴⁶ efforts should be made to standardize measurement protocols in the future. Furthermore, more harmonization methods should be created and used by researchers when comparing estimates of the same activity metric but produced by different protocols.

This systematic review has several strengths. The number of studies that include 60,779 unique children, from 32 countries, and span 2003–2019 lend credibility to the findings. Furthermore, the meta-regression accounted for the clustered estimates within studies. Furthermore, estimates were harmonized using the Rosetta Stone equations. Thus, cut point nonequivalence was accounted for in the analysis.

However, the findings must also be considered with respect to the limitations of the study. The first limitation is that the majority of studies in this systematic review were conducted in North America or Europe with most of these conducted in the United States (n = 32). Thus, estimates may only reflect specific countries in these regions, most of which are high-income countries. Furthermore, while cut point nonequivalence was accounted for, studies that used a different cut point besides one of the 7 included in the Rosetta Stone equation were not included. Furthermore, studies that did not use ActiGraph accelerometers (eg, Actical, pedometer, heart rate monitor) to measure MVPA were not included. This decision was made because ActiGraph accelerometers are the most widely used measure of PA in children¹⁹ and are commonly used in large-scale epidemiological studies of youth PA and MVPA.^139,140 Furthermore, there are no Rosetta Stone equations available for translating estimates of MVPA produced by alternative methods into MVPA estimates produced by ActiGraph accelerometers. Another limitation of the systematic review is the high variability between the studies and the low scores on the risk of bias tool. Several methodological issues have been identified that explain high variability in estimates between studies such as study design issues, inclusion criteria, inconsistent measurement protocols between studies, and varying data reduction processes.^19,141,142 This study is also limited by the timeframe from which studies were drawn (ie, 2003–2020). This is likely because accelerometers were not broadly used to measure children’s PA before the turn of the century. Studies that have examined temporal trends in fitness have indicated that the largest decreases in children’s cardiorespiratory fitness occurred between 1970 and 2000. These decreases in cardiorespiratory fitness may be at least partially explained by changes in school day MVPA that happened prior to 2003 and are absent from the current review.¹⁴³ Another limitation is that the temporal trends in MVPA accumulation may be influenced by unmeasured confounders at the school and child level (eg, socioeconomic status).

Conclusions

Contrary to the prevailing narrative that MVPA accumulated during the school day is declining, the current systematic review found that school day MVPA has increased recently. However, the majority of evidence is from North America and Europe with some evidence from Oceania and very little evidence from Asia and South America. Future studies should prioritize objectively measuring children’s MVPA during the school day in regions other than North America and Europe. Furthermore, efforts to standardize PA measurement protocols in research and methods for harmonizing PA metrics produced by different methods are needed.

Figure 6 —

Bubble plot and meta-regression estimated temporal trends in school day MVPA by sex. CI indicates confidence interval; MVPA, moderate to vigorous physical activity.