Participants
Participants were 503 parents or caregivers of children aged 5 to 10 (M = 7.68, SD = 1.32) recruited from seven public school clusters in the North of Portugal, one of the regions with the highest prevalence of childhood OW/OB in the country [23]. Most of the participants were mothers (n = 428, 85.1%), in average 37.90 years old (SD = 5.58), with 11.54 years of schooling (SD = 3.90), and with a thin or normal weight status (n = 275, 54.7%), as per their report. Most of the children (53.5% female) were enrolled in the second, third or fourth grades (n = 302, 60.0%), and had a thin or normal weight status (BMI z-score < = 1 SD; n = 278, 55.2%), as per their caregivers’ report. Further details on the sociodemographic characteristics of the sample are presented in Table S1 in the Additional file 1.
Procedure
A cluster sampling strategy was used to obtain random lists of public school clusters that included pre-schools and elementary schools, in each of the six most populated sub-regions of the north region of Portugal, according to the information available in the Database of Contemporary Portugal (PORDATA) [24]. School clusters were then contacted, and seven agreed to take part in the study.
With the collaboration of the schoolteachers, 740 parents/caregivers were invited to participate in the study. Inclusion criteria were being a parent/caregiver of a child aged 5–10 years, and having basic language skills in Portuguese. All participants were informed about the study, and those who agreed to participate signed an informed consent form. Parents completed the printed assessment protocol during school evaluation meetings. Those who could not attend the evaluation meeting received the printed assessment protocol through their child’s classroom teacher, completed the questionnaires at home, and then returned the protocol in a sealed envelope to the research team through the teacher.
A total of 526 parents completed the printed assessment protocol (315 in the schools, in the presence of a researcher, and 211 at home by themselves). Only 503 were considered, as 23 questionnaires were returned blank, or were completed incorrectly. Even though the original study used a 2 weeks period for the test-restest reliability, in the current study, parents were invited to complete the PAPP again 1 month after the first assessment protocol. The longer period for the test-retest reliability evaluation was choosen given that in the measure’s instruction parents are asked to answer about their specific PA parenting practices over the last 30 days. If a shorter period was considered, the possibility of overestimating the temporal stability of the PAPP constructs had to be acknowledged. In addition, the literature on research methods also describes that test-retest reliability can be evaluated in intervals up to 4 weeks [25]. A total of 125 parents completed the questionnaire a second time, on average 32 days (SD = 8.72) after the first assessment wave. Data collection took place in 2019, from April to July.
Measures
Sociodemographic characteristics
Parents completed a sociodemographic questionnaire about the sex, age, education, height, and weight of their children, themselves and their partners. Children’s BMI z-scores were estimated according to the WHO reference [26]. Caregiver’s BMI scores were calculated according to the WHO guidelines [27].
Physical activity parenting practices (PAPP)
The PAPP was translated into European Portuguese and adapted to be used with parents of school-aged children aged 5–10. Details on the translation and adaptation procedures are described elsewhere [28].
The PAPP is a self-report questionnaire that evaluates the frequency of parental practices that encourage and discourage children’s engagement in PA [7]. The encouragement scale assesses parenting practices encouraging children’s PA and includes a single-factor, Engagement and Structure, with 15 items (e.g., “How often do you go on a walk with your child?”), and two single items (e.g., “How often do you not register your child for sports or dance due to lack of money?). The discouragement scale assesses parenting practices that discourage children’s engagement in PA and includes four factors, Promote Screen Time (3 items; e.g., “(…) allow your child to play a lot of videogames?“), Promote Inactivity (3 items; e.g., “(…) drive your child, when it was easy to walk?“), Psychological Control (5 items; e.g., “(…) discipline your child for being too active?“) and Restriction for Safety Concerns (4 items; e.g., “(…) let your child go outside to play around your home?”). The items are rated on a 5-point Likert scale ranging from 1 (never) to 5 (always) and reflect the parents’ assessment of how often they used each practice in the previous month.
Comprehensive feeding practices questionnaire (CFPQ)
The CFPQ is a self-report instrument, that assesses specific parenting practices related to feeding, among parents of children aged between 18 months and 8 years [17]. It was translated into European Portuguese and adapted for use with parents of children aged 5–10 years [29]. The study of its psychometric properties with a Portuguese sample of parents revealed that the questionnaire is a reliable measure to assess feeding-related parenting practices in a nine-factor model structure: Monitoring, Modeling, Promotion of Healthy Eating, Involvement, Child Control, Food as Reward, Emotion Regulation, Pressure to Eat, and Restriction for Weight Control and for Health. Using a 5-point rating scale, participants were asked to indicate how often they use a specific strategy (13 items), from 1 (never) to 5 (always), or the degree to which they agreed with a statement (30 items), from 1(disagree) to 5 (agree). In the present study, the CFPQ [29] revealed acceptable fit to data (χ2807 = 1363.12; p < .001; CFI = .91; TLI = 0.90; RMSEA = 0.04; SRMR = 0.05) and Cronbach’s alpha values ranged from .55 (Food as reward) to .87 (Monitoring).
Lifestyle behavior checklist (LBC)
The LBC [30, 31] is a self-report measure with two scales that evaluates parental perceptions of children’s problematic behaviors related to OW and OB (Problem scale) and parental self-efficacy in dealing with those problems (Confidence scale). In the present study only the confidence scale was used. The scale includes 26 statements illustrating problem behaviors related to Overeating (e.g., “Eats too quickly”), Misbehavior in Relation to Food (e.g., “Demands food”), Emotional Correlates of Being Overweight (e.g., “Complains about being teased”) and Physical Activity (e.g., “Watches too much television”), and parents are asked to rate each statement from 1 (certain I can’t do it) to 10 (certain I can do it). In the current study, the Confidence scale revealed acceptable fit to the data (χ2(154) = 533.01; p < .001; CFI = .90; TLI = .88; RMSEA = .070; SRMR = .073) and Cronbach’s alpha values ranged from .90 (Physical Activity) to .92 (Overeating and Emotional Correlates of Being Overweight).
Data analysis
Confirmatory factor analyses (CFA) were performed to evaluate the single-factor structure and the four-factor structure of the PAPP’s Encouragement and Discouragement scales, respectively. The dataset included missing data, which appeared to be missing completely at random (MCAR), as described in the results section. To deal with missing data, CFA were performed using the full information maximum likelihood estimation method. For model fit evaluation, CFI > = .95, TLI > = 0.95, RMSEA <= .06, and SRMR ≤ .10 were used as indicators of good fit to the data, CFI values from .90 to .94, TLI values from 0.90 to 0.94 and RMSEA values from .07 to .08 were used as indicators of acceptable fit [32, 33]. Additionally, CFI and TLI values between 0.80 and 0.90, and RMSEA values above 0.08 and equal to or below 0.10, were used as indicators of marginally acceptable fit [32, 33]. To improve model fit, model re-specification allowing residual covariances was performed [34].
Measurement invariance for each factor structure was tested by performing CFA multigroup comparison according to children’s BMI z-score (underweight/normal vs. overweight/obese). Cross-group constraints were set, and the more restricted models compared with the less restricted ones [35], for configural, metric, and scalar invariance. Invariance was considered when ΔCFI ≤ − .01 and ΔRMSEA ≤0.015 [35, 36].
To evaluate the factors’ reliability, Cronbach’s α values were obtained considering .70 or higher as indicative of an adequate reliability [34]. Convergent and divergent validity, and temporal stability were studied using Pearson’s correlation coefficients [37]. For test-retest reliability evaluation, correlation coefficients equal or above .70 were deemed indicative of adequate temporal stability [38].