Determinants of Outdoor Time in Children and Youth: A Systematic Review of Longitudinal and Intervention Studies

Spending more time outdoors can improve children’s social and cognitive development, physical activity, and vision. Our systematic review summarized the determinants of outdoor time (OT) based on the social-ecological model. We searched nine databases: MEDLINE, APA PsycINFO, Web of Science, Cochrane Central Register of Controlled Trials (CENTRAL), CINAHL, SPORTDiscus, ERIC, SocINDEX, and ProQuest Dissertations and Theses. To be included, studies needed to be quantitative and longitudinal, include ≥1 potential determinant of OT among 0- to 17-year-olds, and be published in English, French, Japanese, or Spanish. We extracted the authors, publication year, country, design, sample size, OT measures, follow-up period, potential determinants, main results, and potential moderators or mediators. Fifty-five studies examining 119 potential determinants met the inclusion criteria. OT was consistently higher in warmer seasons and among participants reporting more OT at baseline. All three interventions that included both parent sessions and additional resources to promote OT (e.g., specific advice and community guides) were effective. COVID-19 restrictions and sun safety interventions discouraging midday outdoor activities led to less OT. The quality of evidence was rated as weak for 46 studies. Most potential determinants were examined in ≤3 studies; thus, more longitudinal studies are needed to enable stronger conclusions about the consistency of evidence and meta-analyses.


Introduction
Evidence from large systematic reviews indicates that children and youth who are more physically active reap multiple benefits, including better motor and cognitive development, greater cardiovascular fitness, reduced cardiometabolic risk, and improved bone health [1,2]. However, the majority of children and youth worldwide do not meet current physical activity (PA) guidelines [3,4]. For example, Guthold et al. [4] estimated that, among 1.6 million 11-to 17-year-olds from 146 countries, 77.6% of boys and 84.7% of girls were insufficiently active. This underscores a need for interventions targeting important determinants of PA. Systematic reviews have consistently identified positive associations between time spent outdoors and children's PA [5][6][7][8]. Yet, previous research suggests that outdoor play (OP) has declined substantially over the last few decades, likely at the expense of increasing screen time [9,10], emphasizing a need to promote outdoor activities.
Given that COVID-19 restrictions were associated with a decline in PA and limited access to opportunities such as physical education and sports, promoting outdoor activities may be even more salient in pandemic and post-pandemic contexts [11]. Unfortunately, cross-sectional studies included in the scoping review by Paterson et al. [11] reported that children spent limited time outdoors during the pandemic. 21. or/12-20 22. 8 and 11 and 21 23. limit 22 to (english or french or japanese or spanish) Note: MeSH = medical subject heading; exp = used with a MeSH term to include all narrower MeSH terms; .ab, .ti, .pt, .tw = field codes for abstract, title, publication type, and text word, respectively; adj# = search for records with terms within # words of each other; * after keyword indicates truncation (e.g., adolescen* will retrieve "adolescent", "adolescents", "adolescence", etc.)

Inclusion and Exclusion Criteria
To be eligible for the review, published studies needed to include (1) participants aged 0-17 years (or parents/guardians reporting on behalf of children); (2) a measure of OT; (3) data on at least one potential determinant in relation to OT; and (4) a longitudinal and quantitative design (i.e., including intervention studies, prospective studies, and retrospective studies). Studies published in English, French, Japanese, or Spanish were eligible. Eligible measures of OT included child-or parent-reported measures, direct

Inclusion and Exclusion Criteria
To be eligible for the review, published studies needed to include (1) participants aged 0-17 years (or parents/guardians reporting on behalf of children); (2) a measure of OT; (3) data on at least one potential determinant in relation to OT; and (4) a longitudinal and quantitative design (i.e., including intervention studies, prospective studies, and retrospective studies). Studies published in English, French, Japanese, or Spanish were eligible. Eligible measures of OT included child-or parent-reported measures, direct observation, devices (e.g., accelerometers equipped with lux sensors), and any other relevant methods.
No restrictions were placed on study location(s) or type of determinants examined. Studies that did not include data for children were excluded. Literature reviews, commentaries, editorials, cross-sectional studies, qualitative studies, and articles not in the aforementioned languages were excluded.

Study Quality
We assessed study quality with the Effective Public Health Practice Project Tool (EPHPP) [27], a well-established method used in public health reviews that include few randomized controlled trials [28][29][30][31]. The tool addresses eight components: (1) selection bias; (2) study design; (3) confounders; (4) blinding; (5) validity and reliability of measurement tools; (6) withdrawals and drop-outs; (7) intervention integrity; and (8) analyses. Following recommended procedures, we graded the first six components as "strong", "moderate", or "weak" [27]. Next, we rated studies as "strong" if there were no "weak" ratings, "moderate" if there was only one "weak" rating, or "weak" if there were two or more "weak" ratings [27]. As in previous PA systematic reviews using the EPHPP, we made some adaptations to the tool [28][29][30][31]. First, for group-level interventions such as cluster randomized controlled trials, wherein all members of a school or preschool were assigned to a control or experimental group, we assessed participation rate at the group level for the first component [28,30]. Second, for the component on confounders, we expected studies to control for sex/gender, age, and socio-economic status, given that these variables were identified as correlates of PA and OT in previous reviews [5,7,19,21]. Third, like Dietz et al. [29], we considered the blinding component non-applicable in observational studies, as there are no interventions that participants/staff can be blinded to. After performing quality assessments independently for all articles, RL and MK met to discuss their assessments and resolve any discrepancies. When insufficient information was provided in the article to assess a component, we consulted previous articles or protocols from the study (when available) and/or contacted the author for correspondence. If no response was obtained, the component was conservatively rated as "weak".

Summary of Findings
Due to the methodological heterogeneity of included studies (e.g., large differences in study designs, measures of OT, potential determinants, follow-up period, and analyses; see Table 1), we considered meta-analyses inappropriate. Instead, we present a narrative synthesis, and we tabulated measures of effect size (e.g., regression coefficients and odds ratios) and statistical significance as presented in the articles. To synthesize the consistency of associations investigated in ≥3 studies as an indicator of confidence in the body of evidence for a particular determinant, we adopted a classification system used in previous systematic reviews of correlates of PA and OT [8,21,32,33]. Variables positively or negatively associated with OT in ≥60% of studies were considered "consistent" correlates and coded as + or −. When 34-59% of associations were positive or negative, we coded the variable as (+) or (−), representing a "possible" correlate. Finally, when <34% of studies supported an association, we coded the variable as 0, indicating no relationship [8]. All studies that met inclusion criteria were included in the summary of findings. As per our PROSPERO record, we intended to stratify results by gender and age group; however, because most potential determinants were examined in a few studies, we were only able to investigate gender and age by age group.   Figure 2 depicts the flow of the review process. Briefly, 4589 records were identified after the exclusion of duplicates. Of these, 4189 were excluded based on title and abstract screening. There were 3 full texts that could not be retrieved, and 346 were excluded based on inclusion/exclusion criteria, leaving 51 included articles. Three additional articles were included from twenty-six potentially relevant records identified by scanning the reference lists of included articles, and one was added from a narrative review of park prescription schemes [88]. Based on EPHPP guidelines, 16 of the 55 included papers were considered randomized controlled trials (RCTs) or controlled trials [12,[37][38][39]44,45,47,53,57,59,60,64,65,67,77,78], 3 were quasi-experimental studies [36,54,58], 32 were prospective observational studies [34,35,[40][41][42][43]46,[48][49][50][51][52]55,56,62,63,66,[68][69][70][71][72][73][74][75][80][81][82][83][84][85][86], 3 were retrospective longitudinal studies [61,76,87], and 1 was an uncontrolled pilot study [79]. Some studies combined data from control and experimental groups [35,40,82] and were considered observational because exposure to the intervention was not of substantive interest in the analyses. Table 1 summarizes the descriptive characteristics of the included studies categorized by age group at baseline (preschoolers (<5 years), children (5-11 years), and adolescents (12-17 years)) based on the mean, grade level, or midpoint of the reported age range. We classified studies that included distinct cohorts based on the highest mean age/grade reported. At baseline, the majority of cohorts had a mean age of <12 years. Only seven studies included a cohort of participants aged ≥ 12 years at baseline [81][82][83][84][85][86][87]. One study was conducted in boys only [74], and one was performed in girls only [82], whereas other studies included boys and girls. The sample size varied from 7 to 26,611, and the follow-up duration ranged from 8 weeks to 15 years. Almost all studies (n = 50) were conducted in high-income countries (e.g., USA (n = 21), Australia (n = 11), the Netherlands (n = 5), Denmark (n = 2), Singapore (n = 2), Canada (n = 1), Chile (n = 1), Finland (n = 1), Germany (n = 1), Israel (n = 1), Italy (n = 1), Spain (n = 1), and the UK (n = 1)), whereas 5 studies were conducted in upper-middle-income countries (China (n = 4) and Brazil (n = 1)). One study was conducted in eight European countries [80].   Table 2 summarizes the potential determinants of OT and the main findings for each individual study stratified by age group. Table 3 presents the summary of associations between potential determinants (organized by levels of influence of the social-ecological model) and OT. Overall, the included studies investigated 119 potential determinants representing the personal, interpersonal, community, built and natural environment, and policy levels. Table 3 is not stratified by age group, as most included studies focused on 5-to 11-year-olds, and few determinants were investigated in ≥3 studies, limiting our ability to draw conclusions about the consistency of associations. It is also noteworthy that  Table 2 summarizes the potential determinants of OT and the main findings for each individual study stratified by age group. Table 3 presents the summary of associations between potential determinants (organized by levels of influence of the social-ecological model) and OT. Overall, the included studies investigated 119 potential determinants representing the personal, interpersonal, community, built and natural environment, and policy levels. Table 3 is not stratified by age group, as most included studies focused on 5-to 11-year-olds, and few determinants were investigated in ≥3 studies, limiting our ability to draw conclusions about the consistency of associations. It is also noteworthy that only four studies assessed policies in relation to OT [57,63,74,76].

Determinants of OT
With twenty studies, age was the most frequently examined variable. Overall, the association between age and OT is equivocal, with six studies reporting an increase with age [34,39,40,45,49,71], five showing a decrease [42,66,80,83,85], five showing mixed/inconsistent findings [43,55,70,81,82], and four reporting no associations [41,46,62,73]. However, five of the six studies reporting an increase involved younger cohorts (<5 years at baseline) [34,39,40,45,49], one of the studies coded as "mixed" reported an increase from 12 to 18 months of age followed by a steady decline until the last follow-up at 5 years [43], and one reported no changes in their younger cohort (ages 5-6 years at baseline), but a decrease in their older cohort (ages 10-12) [55]. Conversely, the average age was ≥5 years for four of the five studies reporting a decline in OT [66,80,83,85]. Five of the eleven studies that examined the association between gender and OT found that boys spent more time outdoors [50,71,81,83,85], one found similar findings in older but not younger children [55], and five found no gender differences [34,41,46,70,73]. Notably, all three studies examining gender differences in cohorts beginning in adolescence found significant differences [81,83,85]. Only two studies examined sex, reporting no differences between males and females [39,43].
Of the six studies examining parental education as a potential determinant, two found that children with more educated parents accumulated more OT [39,52], one found the opposite [70], two found no associations [41,61], and one found that children of more educated fathers increased their OT over time, whereas the mother's education was not associated with OT [43]. For household income, two studies found no association [43,61], and one study found that higher income was associated with lower odds of eliminating OP or exercise during the COVID-19 lockdown in Singapore [76]. All seven studies that examined seasonal differences found that children spent significantly more time outdoors in warmer seasons [45,52,54,55,69,71,75]. Similarly, all three studies that assessed whether OT at a previous time point predicted current OT (i.e., past behaviour) reported significant positive associations [41,70,78]. Exposure to an intervention that included both sessions with parents and additional resources promoting OP (e.g., specific advice and community guide) was consistently associated with more OT (n = 3 interventions) [36,37,58]. In contrast, exposure to a school-based sun safety intervention that discouraged OT around midday was associated with lower midday OT among intervention groups in two out of three studies [53,64,65]. Lastly, all three studies that investigated the effect of the implementation of COVID-19 restrictions found decreases in OT [63,74,76], and one of these studies also reported that OT returned to pre-COVID levels after restrictions were lifted [74].
The criteria for consistency were not satisfied for any other potential determinant. It is worth noting that two of the three school-based curricular interventions that aimed to increase OT did not yield significant differences between experimental and control groups. In general, interventions that primarily focused on other health behaviours (e.g., PA in general and obesity) were less effective, with only 2/6 showing positive results. However, small sample sizes may have limited researchers' ability to detect significant intervention effects. For example, the study by Ford et al. [59] included 28 participants, and despite a medium-to-large increase in OT following the intervention (Cohen's d = 0.71), the effect was not statistically significant (p = 0.057). Similarly, Christiana et al. [54] found that an outdoor PA prescription scheme at a single pediatric practice did not result in increased OT at the 3-month follow-up, but the intervention was very small (n = 32) and took place from August to December, and the measure of OT was very crude (frequency assessed with a Likert scale). Still, 70% of parents reported using intervention materials, and 44% believed that the prescription encouraged their child to participate in outdoor activities [54].

Moderators and Mediators
As summarized in Table 4, twelve studies examined potential moderators. Child age (n = 3), sex (n = 3), and gender (n = 2) were the only potential moderators examined in at least two studies. Age was a significant moderator in two studies [46,63]. First, Shah et al. [46] found an interaction between age and the future risk of myopia and OT. Before the age of four, there were no differences in OT between children who later became myopic and those who remained non-myopic. Then, from 4 to 8.5 years, there was a larger decline in OT among children who became myopic. Second, Li et al. [63] examined whether age, sex, and household income moderated the effect of adherence to four preventive public health measures for COVID-19 on OT. Limiting the number of visitors was associated with a significant decline in OT in children under five years of age, but not in older children [63]. Two of the three studies examining sex as a moderator found no evidence of moderation [38,63]. Conversely, van Grieken et al. [77] found that males in the intervention group had a non-significant decrease in OP, whereas females demonstrated a significant and meaningful increase (>30 min/day). Regarding gender, French et al. [83] found no evidence of moderation, whereas Miller [86] claimed that there was a significant moderation effect but did not report the direction and magnitude of the association. However, Miller [86] did report that youth with lower levels of parental monitoring perceived their neighbourhood as more supportive and reported more OT, whereas youth with high parental monitoring perceived their neighbourhood as more dangerous and reported less OT. These findings suggest that parental monitoring may be a response to a perceived lack of safety that could instill in youth concerns about neighbourhood safety.
Remmers et al. [71] examined parent-perceived responsibility towards their child's PA as a potential moderator. They noted that parent-perceived neighbourhood functionality was associated with more OP in children of parents with high perceived responsibility, while among parents with low perceived responsibility, functionality was related to less OP. They also found that the association between traffic safety and OP was stronger when parents perceived high vs. low responsibility. They suggested that "parents who feel responsible for the amount of their child PA may deliberately provide their child with the autonomy to play outside at spaces that they think are appropriate and safe" [71]. Handy et al. [61] found that the presence of children aged 6-12 in the household moderated the relationship between living in a cul-de-sac and the frequency of OP. Specifically, the presence of cul-de-sacs was supportive of OT for younger children compared to older children. However, the relationship between eleven other characteristics and OT was not modified by the presence of younger children [61]. Examining income as a potential moderator, Li et al. [63] noted that practising physical distancing led to a significant decrease in OT among children from families earning ≥ CAD 80,000, but not in those earning less. Finally, Schneor et al. [74] observed a larger decrease in OT during a full vs. partial COVID-19 lockdown. Only one study planned to examine potential mediators; however, their intervention was not effective, so the criteria for demonstrating mediation were not satisfied [78]. Table 5 summarizes the results of our quality appraisal with the EPHPP tool. Overall, 9 studies were rated as "moderate" [34,41,42,51,68,71,77,78,87], and the remainder (n = 46) were rated as "weak". The components most frequently rated as "weak" were data collection tools (n = 43), selection bias (n = 30), confounders (n = 19), blinding (n = 18), and withdrawals (n = 15). Because we excluded cross-sectional studies, all included studies were rated as "moderate" or "strong" for their study design. The components most commonly rated as "strong" were withdrawals (n = 22), confounders (n = 18), study design (n = 16), data collection tools (n = 6), selection bias (n = 4), and blinding (n = 2). Table 2. Determinants of outdoor time in children and youth.

Studies Beginning in Early Childhood (<5 Years)
Arcury (2017) [34] Gender, age, people per bedroom, number of inappropriate media (having a TV in view at meals and having a TV in the child's bedroom), number of age-appropriate toys, limiting screen time, frequency of visits to play spaces Compared to baseline, the mean time mothers estimated their child playing in the yard or park was 29.9 min/day greater at year 1 and 20.1 min/day greater at year 2 (both p = 0.001). For each additional month of age at baseline (B = 1.0 min/day, p = 0.049) and each age-appropriate toy (B = 12.3, p = 0.001), children spent more time playing in the yard or park. Each unit increase in the limiting screen time score was associated with less OP: B = −6.4, p = 0.016.
Cameron (2019) [35] Influence of peer groups (i.e., partner, friends, mothers' group, and family) on child's nutrition, TV time, and PA No association between influence of any peer group and time spent outside (all p > 0.05).
Davison (2011)  Essery (2008) [37] Effect of newsletter or booklet intervention on child feeding practices and physical activity There was a significant increase in OP reported by the newsletter (p < 0.01) and booklet (p < 0.01) groups between baseline and post-test.

Studies Beginning in Early Childhood (<5 Years)
Sääkslahti (2004) [45] Age, season, and exposure to intervention (parents of children in the intervention group received information and concrete suggestions on how, when, and where to encourage their child's PA) OP varied with intervention (p = 0.041), age (p = 0.016), and season (=12.72, p < 0.001). There were also combined relationships with age and season (p < 0.001), as well as intervention, age, and season (p < 0.001). The age-dependent increase was stronger in the intervention group. Children in the intervention group played more outdoors (p = 0.041) and less indoors (p = 0.049) than controls.
Shah (2017) [46] Age, gender, future likely myopia, and number of myopic parents Girls spent less time outdoors than boys (β = −0.04), but the difference was not significant (p = 0.14). Through the study period, children with one or two myopic parents spent an average of~0.1 SD units per day less time outdoor than children whose parents were both non-myopic (p < 0.01). OT decreased with age, but the difference was not significant (β = 0.007; p = 0.073).

Studies beginning in childhood (5-11 years)
Avol (1998) [50] Gender and ambient ozone concentration On average, boys spent~37 min longer outside in the spring than girls (p < 0.001) and~22 min more outside in the summer (p = 0.04).

Studies beginning in childhood (5-11 years)
Christiana (2017) [54] Exposure to outdoor PA prescription intervention and season No difference in frequency of OT between groups and frequency of achieving ≥60 min of outdoor PA (p ≥ 0.29).
OT declined from baseline (August) to follow-up (November/December; p < 0.01); authors attributed this finding to seasonality.

Studies beginning in childhood (5-11 years)
Handy (2008) [61] Parental preference for and perceptions of neighbourhood characteristics ((1) nearby amenities; (2)  Exposure to school-based multicomponent intervention with specially designed curriculum ("medium" intervention) vs. exposure to multicomponent intervention plus program materials over the summer holidays and low-cost sun-protective swimwear ("high" intervention) compared to standard health curriculum ("control") Adjusted mean OT during the summer holidays between 11 am and 2 pm was highest in the control group ( (99.4, 118.5), with no differences between groups (p = 0.8).

Milne (2007) [65]
School-based sun protection curriculum over 4 years; children were encouraged to reduce sun exposure by staying indoors during the middle of the day, when solar ultraviolet radiation is highest, and to protect themselves when outdoors by using shade, clothing, hats, and sunscreen The median OT in each group (control, "moderate" intervention, and "high" intervention) was similar after 2 years. There was no association between study group and total OT at either age 10 or age 12.  Remmers (2014a) [70] Gender, age, and parental and environmental factors, including accessibility of PA-related places, attitude towards child PA, concern regarding child PA, restriction of screen time, social capital, functionality, traffic safety, attractiveness, perceived responsibility, pressure, and monitoring Children spent on average~60 more minutes in OP per week at 7 vs. 5 years of age (both boys and girls) (p < 0.01). At both time points, boys spent significantly more time in OP than girls (p < 0.01), and there were significant differences in OP duration between all seasons (p < 0.01; but season was examined as a random effect, and the direction of association was not reported).   1.05, 1.18)). As 95% CIs cross 0, differences were not significant.
Sanchez-Tocino (2019) [73] Age and gender There were no significant differences in hours spent on outdoors activities by age or gender (p > 0.05).

Walker (2021) [79] Participation in a child-centred play therapy intervention
There were no differences in OT on weekdays and weekend days between baseline and the end of the intervention (all p > 0.20)

Studies beginning in adolescence (12-17 years)
Dunton (2007) [81] Gender, age, time of week, and season Compared with girls, boys were more likely to report exercising in outdoor settings (p = 0.002) and walking in outdoor settings (p < 0.001). Walking in an outdoor setting decreased during high school (7% per year, p = 0.019), but outdoor exercising did not (p = 0.189). Students were more likely to exercise or walk outdoors on weekend days vs. weekdays (p < 0.001). Students were more likely to walk or exercise outdoors in the fall and spring seasons vs. the winter (all p < 0.05) and to walk outdoors in the spring vs. fall season (p = 0.010).

Evenson (2018) [82] Age
The number of park visits identified by GPS during the 6-day monitoring period increased from 73 to 83 (p < 0.02). Mean duration of park visits decreased from 63.9 to 38.4 min (p < 0.03).

Studies beginning in adolescence (12-17 years)
French (2013) [83] Age, ethnicity, and gender In the young cohort, OT decreased by just over 1 h/wk from baseline to follow-up, accompanied by a decline in outdoor leisure (both p ≤ 0.001). Time spent on organized outdoor sports increased (p < 0.0001). In the older cohort, there was a significant decrease in OT and outdoor sporting activities (p < 0.0001), but not in outdoor leisure time (p = 0.06). Boys spent~2.5 h/wk more outdoors than girls in both cohorts at baseline and follow-up (all p < 0.0001). The decrease in OT between baseline and follow-up was significant for girls in both cohorts (younger, p = 0.006; older p < 0.0001) and for boys in the older cohort (p = 0.001), but not in the younger cohort (p = 0.052). The decline in OT with age was seen in European Caucasian participants (younger cohort: p = 0.001; older cohort: p < 0.0001), but not in East Asian participants (younger cohort: p = 0.7; older cohort: p = 0.07).

Gopinath (2013) [84] Birth weight (categorized in quartiles)
In 12-year-olds, an increase in outdoor PA (~1 h/wk) was observed with increasing birth weight after adjustment for covariates (from the lowest to highest quartile; p trend = 0.02). Among 17-to 18-year-olds, higher birth weight was associated with higher outdoor PA (~1 h/wk, p = 0.04). In multivariable models, each SD (573.5 g) increase in birth weight was associated with a 15 min/wk increase in outdoor PA (p = 0.01).
Miller (2017) [86] Parental perceptions of neighbourhood danger, perceived neighbourhood support, parental monitoring (in general), and OT at baseline There was a positive correlation between the percent time spent outside at Times 1 and 2 (r = 0.480). Increased levels of parental monitoring at Time 1 was associated with increased OT at Time 2 (coeff = 0.7508, p = 0.0109).

Watowicz (2012) [87] Recent parental weight loss surgery
Control group participants were significantly more likely than those whose parents underwent weight loss surgery to report ≥1 h/day of OP (55.8 vs. 31.6%, p = 0.01).

Number of Studies Positive (%) Negative (%) Mixed (%) Null (%) Summary Code
Season ( [45] also found significant difference between spring and fall seasons, but it is not included in this table because it was not clear which season was warmer. Due to the small number of studies that were not rated as "weak" based on the Effective Public Health Practice Project tool (see Table 5), the risk of bias remains high for most potential determinants. Table 4. Moderators and mediators examined in relation to outdoor time.

Author (Year) Moderators or Mediators Examined Results
Avol (1998) [50] Asthma status (healthy, "wheezy", or asthmatic) as potential moderator Asthma status did not moderate the relationship between peak hourly ambient ozone (O 3 ) concentration and OT (all p > 0.10).
French (2013) [83] Gender and ethnicity as potential moderators Gender and ethnicity did not moderate the relationship between age and outdoor time (all p ≥ 0.2).
Händel (2017) [38] Sex, age, mother's BMI, mother's education, and mother's PA level were examined as potential moderators of the effect of the intervention.
No significant interactions were found (all p ≥ 0.2).
Handy (2008) [61] The presence of children aged 6 to 12 was examined as a potential moderator between 12 perceived neighbourhood characteristics ((1) nearby amenities; (2) neighbourhood upkeep; (3) large back yard; (4) large front yard; (5) living in a cul-de-sac rather than on a through street; (6) low traffic on neighbourhood streets; (7) parks and open spaces nearby; (8) sidewalks through neighbourhood; (9) lots of interaction among neighbours; (10) lots of people out and about in the neighbourhood; (11) low crime rate in neighbourhood; (12) safe neighbourhood for children to play) and weekly frequency of OP.
The only significant interaction in longitudinal analyses was between the change in cul-de-sac and the presence of children aged 6-12 in the household (β = 0.170; p = 0.014), suggesting that the presence of cul-de-sacs is supportive of OP for younger children.
Li (2021) [63] Child age, sex, and household income as potential moderators The effect of limiting the number of visitors on OT was significant in children < 5 years (β = −9.94, 95% CI = −17.18, −2.71, p = 0.01), but not in older children. The effect of practising physical distancing was significant for children for families earning ≥ CAD 80,000 (β = −5.00, 95% CI: −9.47, −0.53, p = 0.03), but not in those earning less. Age, sex, and household income did not moderate any other associations.
Miller (2017) [86] Gender and parental monitoring as potential moderators The effect of youth-perceived neighbourhood danger at Time 1 on OT at The author reported that "youth with lower levels of parental monitoring perceived their neighbourhood to be more supportive and spent more time outside. In contrast, youth with high parental monitoring, who perceived their neighbourhood to be more dangerous, spent less time outside". The author did not clearly discuss the moderating effect of gender.

Author (Year) Moderators or Mediators Examined Results
Remmers (2014a) [70] Perceived responsibility regarding child PA as potential moderator Perceived responsibility moderated the effect of perceived functionality on OP (Table 3). When stratified, functionality was related to more OP in children of parents with high perceived responsibility (β = 0.04; 95% CI = −0.07, 0.15), while among parents with low responsibility, functionality was related to less OP (β = −0.03; 95% C = −0.09, 0.04). Traffic safety interacted with perceived responsibility, but this effect was only significant after adjustment for main effects (Model 3) and became non-significant after adjustment for the other interaction (Model 4). Stratification showed that the association between traffic safety and OP was marginally stronger when parents perceived high (β = 0.10; 95% CI = −0.03, 0.23) vs. low responsibility (β = 0.06; 95% CI = −0.003, 0.12).
Shah (2017) [46] Age as potential moderator Age moderated the association between future risk of myopia and OT (p = 0.002), such that, before the age of 4, there was little difference in OT between children who later became myopic and those who remained non-myopic. Between the ages of 4 and 8.5 years, children who later became myopic spent progressively less time outdoors than their peers who remained non-myopic (0.1 SD unit per day difference in OT by age 8.5 years).
van  [77] Sex as potential moderator The interaction between sex and group was significant (p = 0.019 Van Stralen (2012) [78] This study investigated several hypothesized personal and environmental mediators, which included perceived pros and cons, social pressure, social support, social modelling, self-efficacy, planning skills, perceived barriers, enjoyment, and habit strength related to OP.
No statistically significant intervention effects on potential mediators were seen at Time 2; thus, criteria for mediation were not satisfied.
Watowicz (2012) [87] Length of time since parental weight loss surgery as potential moderator There were no differences in reported lifestyle behaviours (including OP) in an analysis of the subset of subjects (n = 33) for whom length of time since surgery was available (by Watowicz et al.   Note: Blinding was rated as non-applicable (N/A) for observational studies given the absence of intervention.

Discussion
Our systematic review summarized previous longitudinal and intervention studies examining the determinants of OT in children and youth. Overall, 119 determinants spanning the social-ecological model were examined across the 55 included studies (including 35 observational and 20 intervention studies). Illustrating the rapid growth in this field of research, the largest and most recent previous review on this topic included only 12 longitudinal studies [21]. Although we found that few potential determinants were included in a sufficient number of studies to draw conclusions about the consistency of associations, we identified some consistent positive or negative determinants of OT at multiple levels of influence.

Individual Level
About half of the included studies found that boys accumulated more OT than girls, with other studies reporting no gender differences. When examining findings by age group, all three cohorts that began in adolescence found that boys spent more time outdoors than girls. In the broader literature on correlates of PA, boys are usually more active than girls [8,22], with longitudinal studies suggesting that the age-related decline in PA tends to begin earlier in girls [89]. Hence, additional efforts may be needed to promote OT among girls. In our review, five studies examined sex or gender as potential moderators, and two found significant interactions. Notably, the intervention by van Grieken et al.
[77] achieved a substantial increase in females' (but not in males') OT. Our findings suggest that the determinants of OT may vary according to sex and gender, and a better understanding of such differences could help guide future interventions.
Like the previous review by Lee et al. [21], we initially found mixed associations between age and OT. However, when considering the direction of age-related changes in OT by age group, we noted that increases were only reported in studies beginning in early childhood (≤5 years old), whereas decreases were common among older cohorts. These observations suggest that the relationship between age and OT may be curvilinear. If confirmed by future research, this would suggest a need for interventions to promote sustained engagement in outdoor activities in an effort to minimize the well-known age-related decline in PA [8,22,89] and increase in myopia prevalence [12]. In addition to promoting OT, interventions with youth may need to address important barriers to outdoor activities. For example, a large Canadian mixed-methods study indicated that the addictive nature of electronic screen devices and the belief that being indoors is safer and more comfortable were key factors associated with reduced connection to nature [90]. Echoing these findings, a quantitative study with Ecuadorian children found that connection to nature was negatively associated with screen time [91]. Conversely, connection to nature and OT are both associated with PA [5][6][7][8]91], suggesting that increasing outdoor activities could improve individual health [1,2] while helping to address the United Nations' Sustainable Development Goals [17,92].
All three studies that examined the association between OT at baseline and subsequent OT found positive associations. These findings are consistent with previous research in the fields of PA epidemiology and health psychology, showing that past behaviour is an important determinant of current behaviour [93][94][95][96]. In this regard, our findings suggest that promoting OT may yield lasting benefits, though intervention studies are needed to test this hypothesis.

Interpersonal Level
The important role of parents in supporting their child's PA is well established [97,98]. In our review, different aspects of parental influence (e.g., encouragement, co-participation, attitude, and intention) were examined as potential determinants of OT in no more than two studies, precluding strong conclusions. Yet, we found preliminary evidence that parent participation in PA/OP [41,69] and parental habit strength to increase OP [70] were associated with higher OT among primary school children. Ostrin et al. [69] notably reported a strong correlation within child-parent dyads in device-measured outdoor light exposure (i.e., lux) (r = 0.76). These findings are consistent with previous research showing consistent positive correlations in PA within parent-child dyads-especially in studies using device-based measures of PA [98].
In general, we found insufficient or inconclusive evidence regarding the associations between parental socio-demographic variables (e.g., education, income, occupation, and ethnicity) and OT (Table 3). Mostly relying on cross-sectional findings, previous reviews concluded that children from ethnic minorities and more educated parents played outside less [19,21]. It is also important to bear in mind that, in line with social-ecological theory [22], parental socio-demographic variables may interact with other levels of influence. Such interactions were not investigated in previous reviews, and we found only one study that examined income as a potential moderator [63]. Hence, more research is needed to unpack the potential role(s) of socio-demographic variables in influencing OT, as this may help in tailoring interventions to different groups.
Extending previous reviews, we were able to include 20 intervention studies. We found that all three interventions that included sessions with parents augmented by additional resources to promote outdoor activities (e.g., specific advice and community guide) were associated with higher OT [36,45,58]. While promising, two of these studies did not include a control group [36,58], and thus, further research is needed to confirm these findings.
Within the realm of interpersonal influences, nature/outdoor PA prescriptions from pediatricians and other health professionals, such as the US National ParkRx Initiative (https://www.parkrx.org/, accessed on 7 December 2022) and the Canadian PaRx (https://www.parkprescriptions.ca/, accessed on 7 December 2022), are gaining popularity to address physical inactivity and excessive screen time [88]. One outdoor PA prescription intervention met our inclusion criteria, but it was limited by a small sample size, seasonal changes, and a crude outcome measure [54]. Conversely, pilot studies from Kondo and colleagues' [88] narrative review of nature prescription programs found positive intervention effects on the park visit frequency [99,100] but did not measure OT per se. Notwithstanding the potential benefits of park prescriptions, the currently insufficient evidence of effectiveness represents a barrier for healthcare providers to dedicate their limited time [88]. This underscores a need for larger outdoor activity prescription interventions using stronger study designs and measures.

Community Level
In contrast, school-based sun safety interventions that discouraged OT during midday were generally associated with lower OT [53,65]. Well-intended public health campaigns focusing on different issues have collectively recommended keeping children indoors for most of the day to avoid (1) sun exposure between 10:00 am and 4:00 pm to prevent skin melanoma; (2) ozone exposure, which tends to be highest in commuting periods immediately before and after school; and (3) exposure to insect-borne diseases from dusk to dawn [7]. Notwithstanding these issues, researchers have called for a better balance between risks and benefits and identified often overlooked risks associated with being indoors, including physical inactivity, excess screen time, an increased risk of myopia, exposure to an obesogenic food environment, indoor air pollution, allergens, and cyberpredators [7,14]. Parents, educators, pediatricians, and other health professionals could play an important role in reframing perceptions of risks associated with OT [14,101].

Natural Environment Level
We found that children consistently spent more time outside during the warmer months. These findings are consistent with previous reviews on PA and OT [21,102]. While seasonal variations cannot be modified in the short term, interventions could be developed to promote outdoor activities in colder seasons with appropriate clothing. Such interventions may be particularly useful for new or recent immigrants who may be at higher risk of physical inactivity [103].

Policy Level
Only four studies examined policy-level determinants, but all of them found significant associations with OT. Cortinez-O'Ryan et al. [57] found that an evening street closure intervention occurring twice a week for 12 weeks was associated with significant increases in measures of frequency and duration of OP. In contrast, all three studies that examined the effect of COVID-19 restrictions found a significant decrease in OT. Furthermore, in a small sample of Israeli boys, Schneor et al. [74] reported that the effect of a full lockdown was greater than that of a partial lockdown and that device-measured OT returned to prepandemic levels after restrictions were removed. Our findings are consistent with reviews that reported large decreases in PA and increases in screen time during the pandemic [9,104]. Given the potential of PA to reduce the risk and severity of COVID-19 infections [105] and the lower risk of transmission in outdoor environments, more efforts should be invested in promoting OT and PA during future pandemics.

Limitations of Included Studies
Based on the EPHPP tool, quality was rated as "weak" for most included studies. It is worth noting that some components of the EPHPP are rated quite severely [27,30]. For example, the withdrawals component is rated as "weak" whenever attrition exceeds 40%, regardless of the follow-up duration. Similarly, a participation rate below 60% is considered "weak" for the selection bias component, regardless of the complexity and/or length of the study. The blinding of participants was impossible or impractical in most intervention studies. Hence, the summary quality ratings should be interpreted with caution. That said, given the large number of "weak" ratings for data collection tools, future studies should use measures of OT that have been shown to be reliable and valid. In our review, most studies used child or parent reports and did not provide information about the psychometric properties of their measures. Some studies also appeared underpowered to examine determinants of OT. Finally, most of the studies were conducted in high-income countries, so the transferability of the findings to lower-and middle-income countries is unclear.

Strengths and Limitations of the Review
The strengths of the review include the focus on longitudinal studies that can establish a temporal sequence, the consideration of the entire pediatric population, and the robust screening and quality assessment process. Furthermore, we were able to consider articles published in English, French, Japanese, and Spanish, although we may still have missed articles published in other languages. Given that most potential determinants have been examined in only one or two studies, we could not conduct the planned analysis of determinants stratified by gender and age group. Similarly, the heterogeneity of the exposure and outcome measures among the included studies precluded meta-analyses. Future studies examining the same potential determinants are needed to draw stronger conclusions. We also noted that OT was measured in a variety of ways, with some studies focusing only on OP, others on outdoor PA, and others on all activities that took place outdoors (Tables 1 and 2). Standard definitions of terms related to OT and play have recently been proposed [18], and their uptake in future studies could help minimize methodological heterogeneity.

Conclusions
Previous studies have examined 119 potential determinants of OT representing multiple levels of influence of the social-ecological model, but few determinants have been examined frequently enough to draw strong conclusions. Nevertheless, we found consistent evidence that children spend more time outdoors in warmer seasons, that OT at baseline predicted subsequent OT, and that COVID-19 restrictions and sun safety interventions discouraging midday outdoor activities were associated with less OT. The association between age and OT appears to be curvilinear, with most increases reported in early childhood, followed by decreases in late childhood and adolescence. About half of studies examining gender differences reported that boys spent more time outdoors (especially in adolescence), and the remainder found no differences. Interventions that include both parent sessions and additional resources to promote OT seem promising, but more robust research is needed to confirm their effectiveness. Given that most included studies were rated as "weak", there remains a need for stronger studies to improve the quality of evidence. Longitudinal studies investigating the determinants of OT among underrepresented populations, notably adolescents and children and youth living in low-and middle-income countries, are also needed. Finally, future studies should consider using both subjective and device-based measures of OT, such as accelerometers and other devices equipped with light sensors that can reduce social desirability and recall biases associated with subjective measures [69,74,106].