All methods and procedures were carried out in accordance with relevant guidelines and regulations governing human subjects research.
Participants
Participants in the current study included 473 parents (43% mothers; 33% fathers) from the USA. Participants were stratified across six household income levels which included annual incomes of <$20,000 (n = 36), $20,000-$44,999 (n =72), $45,000-$69,999 (n = 96), $70,000-$94,999 (n = 99), $95,000-$119,999 (n = 74), and >$120,000 (n = 96). On average, participants reported having 1.5 children (SD = .77), with an average age of 10 years (SD = 6.1). Forty-two parents reported having a child with a developmental disability, 16 parents reported having a child with a physical disability, and 96 parents reported having a child with social, emotional, or behavioral challenges (e.g., ADHD, depression, anxiety). Finally, 44 parents reported having more than one child with a disability. Table 1 provides an overview of sample demographics.
Table 1
Sample Descriptive Statistics.
Descriptive Variable
|
n (%)
|
Relation to child
|
|
|
Mother
|
218 (45.61)
|
|
Father
|
166 (34.73)
|
|
Stepmother
|
2 (0.42)
|
|
Stepfather
|
2 (0.42)
|
|
Grandparent
|
2 (0.42)
|
|
Gender not specified or missing
|
88 (18.41)
|
Race/ethnicity
|
|
|
White
|
380 (79.66)
|
|
African American
|
24 (5.03)
|
|
Asian
|
34 (7.13)
|
|
Hispanic/Latino
|
28 (5.87)
|
|
Hawaiian/Pacific Islander
|
1 (0.21)
|
|
Native American
|
2 (0.42)
|
|
Biracial/multiracial
|
6 (1.26)
|
|
Other
|
2 (0.42)
|
|
Income
|
|
|
< $20,000
|
37 (7.76)
|
|
$20,000-$44,999
|
73 (15.30)
|
|
$45,000-$69,999
|
97 (20.34)
|
|
$70,000-$94,999
|
99 (20.75)
|
|
$95,000-$119,999
|
75 (15.72)
|
|
>$120,000
|
96 (20.13)
|
|
Disability Status
|
|
|
Developmental Disability
|
42 (8.81)
|
|
Physical Disability
|
16 (3.35)
|
|
Emotional/Behavioral Disability
|
96 (20.09)
|
|
More than 1 child with disability
|
44 (9.22)
|
|
|
|
|
Measures
Demographic data as well as parents’ perceptions of three areas of recess were assessed: belonging and victimization, recess policies, and recess procedures. These measures are explained in further detail below.
Demographics. Demographic information was collected for all participants, including their relation to their children, race/ethnicity, annual household income, and the age and disability status of their children. For child disability status, parents had the option to indicate that the child had more than one disability. All demographic variables were coded as categorical variables and are presented in Table 1.
Belonging and victimization. Survey items related to belonging and victimization were based on previous surveys developed and validated by McNamara [12] and colleagues, in which they modified belonging and victimization items from existing scales to be adapted to the recess environment. In the current study, belonging and victimization items were slightly adapted to reflect parents’ perceptions, rather than students (e.g., “I get along well with others during recess” was modified to “my child gets along well with others during recess”). All 12 items were measured on a 5-point Likert scale (strongly disagree to strongly agree). Results of a confirmatory factor analysis (CFA) revealed the model fit the data in the current study (χ2 = 234.52, p < .001; CFI = .968; TLI = .959; RMSEA = .098; SRMR = .042). Reliability was calculated with standardized estimates using McDonald’s [25,26] omega coefficient (ω = (Σλi)²/([Σλi]²+Σδii), where λi are the factor loadings and δii the error variances. Internal reliability for the belonging scale, as measured by ω was .78. Internal reliability for the victimization scale, as measured by ω was .85. Individual items and factor loadings for the belonging and victimization scale can be found in Table 2.
Table 2
Factor Loadings for Belonging and Victimization Scale.
Item
|
Loadings
|
There are lots of different games my child can play during recess
|
0.823
|
My child is threatened at recess by other children
|
0.873
|
My child has access to a variety of things to play with at recess
|
0.811
|
My child is threatened at recess by adults
|
0.912
|
My child has friends they can play with during recess
|
0.812
|
My child is not allowed to play with certain groups of children at recess
|
0.724
|
My child has been hit, kicked, or scratched at recess
|
0.715
|
My child is supported by adults during recess
|
0.684
|
My child gets along well with others during recess
|
0.684
|
My child has been in a physical fight at recess
|
0.738
|
My child is comfortable talking to teachers and staff about problems that happen at recess
|
0.822
|
My child has been teased during recess
|
0.699
|
Recess policies. Items used to assess parent perceptions of recess policies were adapted from the School Physical Activity Policy Assessment (S-PAPA [27]). This assessment examines policy related to physical activity and recess opportunities at elementary schools and includes three modules: (a) Physical Education; (b) Recess, and (c) Other Before, During, and After School Programs. In the current study, only the Recess module items were used and they were adapted to parent’s report of their beliefs about recess policies as it related to subscales of the importance of recess (e.g., “recess is an important part of the school day”); the unimportance of recess (e.g., “schools should not spend money on recess); and recess withholding (e.g., schools should not be allowed to take away recess for not completing academic work”). All 12 items were measured on a 5-point Likert scale (strongly disagree to strongly agree). Results of a confirmatory factor analysis (CFA) revealed the model fit the data in the current study (χ2 = 156.40, p < .001; CFI = .986; TLI = .981; RMSEA = .067; SRMR = .031). Internal reliability, as measured by ω, was .89 for recess importance, .84 for recess unimportance, and .74 for recess withholding. Individual items and factor loadings for the parents’ perceptions of recess policies scale can be found in Table 3.
Table 3
Factor Loadings for Parents’ Perceptions of Recess Policies.
Individual Items
|
Loadings
|
Recess is an important part of the school day IMPORT1
|
0.874
|
Schools should not be allowed to take away recess for behavior problems in the classroom WITHHOLD1
|
0.612
|
Schools should not spend money on recess
|
0.742
|
Recess is just as important as any other subject at school
|
0.799
|
Children learn important social skills during recess
|
0.866
|
Recess is good for children’s health
|
0.877
|
Physical activity is not important during the school day
|
0.787
|
Schools should be focused on academic achievement, not being physically active
|
0.801
|
Play has no place in the school day
|
0.919
|
I do not care what happens during recess at my child’s school
|
0.634
|
Schools have more important issues to focus on other than recess
|
0.715
|
Schools should not be allowed to take away recess for not completing academic work
|
0.996
|
Recess procedures. Items used to assess parent perceptions of recess procedures were adapted from the Great Recess Framework – Observational Tool (GRF-OT [28]). The GRF-OT contains 17 items that describe critical aspects of a live recess environment. Previous research has established adequate validity and reliability for this measure. In the current study, 10 items from the GRF-OT were modified to assess parent reported perceptions of student and staff engagement practices (e.g., “Teachers and staff should encourage a positive culture at recess”) and the physical environment (e.g., “The recess environment should be free of hazards”). All 10 items (three pertaining to safety and structure, three pertaining to student engagement, four pertaining to teacher engagement) were measured on a 5-point Likert scale (strongly disagree to strongly agree). Results of a confirmatory factor analysis (CFA) revealed the model fit the data in the current study (χ2 = 147.89, p < .001; CFI = .977; TLI = .969; RMSEA = .085; SRMR = .036). Internal reliability, as measured by ω, was .81 for the physical environment, and .71 for student and staff engagement. Individual items and factor loadings for the parents’ perceptions of recess procedures scale can be found in Table 4.
Table 4
Factor Loadings for Parents’ Perceptions of Recess Procedures.
Individual Items
|
Loadings
|
Schools should have dedicated outdoor space to recess
|
0.834
|
Schools should have grass and natural areas for children to play in during recess
|
0.889
|
The recess environment should be free of hazards
|
0.744
|
Children should be able to do what they want at recess
|
|
Recess should include a variety of activities for children to play
|
0.846
|
Children should be physically active during recess
|
0.699
|
Teachers and school staff should encourage a positive culture at recess
|
0.868
|
Teachers and school staff should supervise recess
|
0.752
|
Teachers and school staff should play alongside children at recess
|
0.216
|
Schools should teach children conflict resolution skills during recess
|
0.546
|
Procedures
All study procedures were approved by the Institutional Review Board at the first author’s institution. Before any data were collected, informed consent was provided by all participants. Data were collected through Prolific [29] (www.prolific.co), an online platform designed for researchers to recruit potential study participants. The Prolific platform explicitly informs registered users that data collected will be used in research, has clear guidelines as to what is and is not allowable (e.g., direct identifiers are not allowed within data collection), and requires a minimum payment for participants at a rate of $6.50 USD per hour. Prolific also allows researchers to recruit participants from pre-screened, unique niche populations. For the current study, only parents were included in the recruitment process. Previous research has suggested that Prolific produces high quality data that is comparable to, or of a higher standard as compared to other online recruitment methods [30]. This may be due to Prolific’s implementation of quality checks limiting random responses and bot accounts, such as verifying phone numbers, limiting number of accounts using the same internet protocol (IP) address or internet service provider (ISP), restricting new accounts based on IP and ISP, and investigating suspicious accounts reported by researchers [31].
Data Analysis
Prior to data analysis, all data were screened for patterns of missingness. Data were also screened for careless responses, with no issues flagged. Specifically, data were checked to ensure that all participants responded uniquely to an open-ended question, no participant response time was below two-seconds per item, and no participant response time was below 2 SDs of the mean [32]. To account for missing data present at the item level, models were estimated using the full available information, based on algorithms implemented in Mplus. Data were then analyzed using latent variable modeling in Mplus v8.4 with the weighted least square mean and variance adjusted estimator. As a first step, we used CFA to test the fit of the measurement model of each assessment (as described under each measure above). Next, we created a household disability index (HDI) to account for how many disabilities each parent reported across all children living in the home. For example, if a participant listed Child 1 with a physical disability and developmental disability, and Child 2 with a physical disability, the corresponding HDI would be a score of “3” for that participant. HDI was then used as a predictor of parent perceptions of recess using a latent variable model framework along with household income and race/ethnicity. Model 1 examined the relationship between HDI and parent perceptions of belonging and victimization at recess. Model 2 examined the relationship between HDI, income, race/ethnicity and parent perceptions of importance of recess, un-importance of recess, and recess withholding. Finally, Model 3 examined the relationship between HDI, income, race/ethnicity, and parent perceptions of the physical environment and staff and student engagement at recess.
Decisions about model fit were made using the Chi Square (χ2) statistic, the Root Mean Square Error of Approximation (RMSEA), the Comparative Fit Index (CFI), the Tucker Lewis Index (TLI), and the Standard Root Mean Square Residual (SRMR). While the χ2 statistic is the most commonly reported measure used in establishing model fit [33], this value is sensitive to sample size, and a non-significant χ2 value is often difficult to obtain even when the model is a good fit using other criteria or assessment [34]. As such, it is typical to use model fit indicies that are less dependent on variations in sample size such as the RMSEA, CFI, TLI, and SMRM (Marsh et al., 2004). Cut-off values > .90 for the CFI and TLI have been conisdered indicative of adequte model fit, while values ≥ .95 are preferred for an acceptable modle fit, and cut-off values <.08 have been conisdered indicative of adequte model fit for the SRMR and RMSEA, while values of ≤ .06 for the RMSEA are preferred for an acceptable model fit [34,35].