Physical activity (PA) plays an important role in normal growth and development of youth and is a predictor of their health . Only about 1 in 3 young Canadians attain the recommended weekly average of at least 60 min of daily moderate-to-vigorous PA (MVPA) . The childhood to adolescence transition represents a period of individual development often marked by a decline in MVPA levels . Throughout this period, the influence of various behavioural determinants may change as individuals gain control and autonomy . Since adolescence is associated with the development of stronger social affiliations outside of the family network, it is possible that the effect of the family environment changes during this period. Still, parental support for PA can generally be perceived as social support, which is thought to promote PA through better self-efficacy . The social cognitive theory suggests that participation in PA is promoted through self-efficacy because of an enhanced capacity to make abstraction of potential barriers to PA participation . Social support could also have a positive influence on individuals’ motivation for PA, which in turn would influence behaviour .The influence of parents on adolescents’ PA levels nevertheless remains ambiguous as most studies to date concentrated on aggregated parental behaviours without accounting for possible overlapping between them [8, 9].
Notwithstanding limitations of previous studies, results to date suggest that parents can regulate offspring’s participation in PA through verbal support, co-participation, provision of resources, encouragement, guided choices, involvement and offering of rewards [8, 10,11,12]. Studies document positive associations between parental support and PA among adolescents [13, 14]. Parental support behaviours relating to services facilitating PA (e.g. transportation) can be categorized as tangible whereas providing verbal encouragements and praises in relation to PA can be categorized as intangible support [12, 15]. Although tangible and intangible parental support may have different associations with MVPA, most studies present parental support as a cumulative score combining all parental support behaviours . Further, studies to date do not fully assess whether associations between parental support behaviours and MVPA are sustained throughout adolescence. As adolescents seek to gain autonomy, it is possible that the influence of intangible parental support declines faster with age than the influence of tangible support, which adolescents continue to depend on to address external barriers to physical activity participation, including provision of transportation, equipment and registration costs . Tangible parental support may also continue to have a positive influence on adolescents’ physical activity if they depend on the participation of parents to pursue participation in the activity themselves.
Beyond parental support, parenting styles may also influence youth’s participation in PA. Parenting style refers to the emotional and relational climate created by parents and is a combination of emotional involvement (warmth) and demandingness (control) . Ambiguity exists in relation to the direction of an association between parenting style and PA [17, 18]. Whereas a cross-sectional study among children 10 to 11 years showed that low parental control is positively associated with children’s PA , a longitudinal-study showed that parental control was positively associated with PA 3 years later among elementary school students . It is also unclear if parental control is linked to adolescents’ MVPA.
Because social determinants represent a large proportion of PA socio-ecological frameworks and adolescence is a period when parents may still exert influence on their offspring’s MVPA, it is essential to gain a better understanding of the influence specific parental behaviours and parenting styles may have on MVPA throughout adolescence. Such knowledge could help inform interventions among parents to promote physically active lifestyles among adolescents. Therefore, this study aimed to 1) examine longitudinal and lagged associations between total, tangible and intangible parental support and control behaviours for PA on the MVPA of adolescents; and 2) assess the independent associations between various parental support practices (i.e. co-participation, transportation, motivational, encouragement, informational and modeling) and control (i.e. nagging and ordering) with adolescents’ MVPA.
The Monitoring Activities of Teenagers to Comprehend their Habits (MATCH) study , is an ongoing prospective study designed to describe the natural development of PA patterns of youth and identify their determinants. The MATCH study was designed to include a mix of students from schools from low, middle, and higher socioeconomic status in a variety of urban, suburban, and rural settings in French and English regions of New Brunswick. Briefly, in 2011, 806 participants were recruited from 17 schools across the province of New Brunswick, Canada when they were in Grade 5 or 6 (age 10–12 years). Participants were invited to fill three self-report questionnaires per year (every 4-months) until they completed Grade 12 (up to 24 survey cycles). The sample increased to 937 as other students from participating schools were allowed to enter the study in follow-up survey cycles. With a loss-to-follow-up proportion of less than 9% per year, 497 participants were still actively involved in the study in their last year of high school (Grade 12). On average, participants were followed up for 2.6 cycles and the most common reasons for losses to follow-up were having moved (n = 174), the school interrupting its participation (n = 191) and choosing to leave the study (n = 76). For the current analysis, we used data from cycles 9 (n = 617), 19 (n = 367) and 22 (n = 208), (when participants were approximately 12, 16 and 17 years, respectively) since information on perceived parental support and control behaviours was captured only in these cycles. Participants who participated in at least two of the three cycles which included measures of parental behaviours were included in this analysis.
Moderate to vigorous PA
Participants self-reported their involvement in MVPA at each cycle using a two-item questionnaire . Specifically, participants were provided a definition and examples of MVPA and then asked to indicate the number of days they engaged in at least 60 min of MVPA in 1) the past week and 2) the typical week. Response options ranged from 0 to 7 and the average of the two items showed good test-retest reliability with an intraclass correlation (ICC) of 0.77 when tested among 138 adolescents aged 12.1 (standard deviation = 0.9) years . The MVPA score obtained by averaging the two items also correlated with measures of accelerometers accumulated over 5 to 7 days (r = 0.40, p < 0.001) . To account for seasonal variation in MVPA, the current analysis uses the average of participants’ MVPA scores in all three survey cycles of a given school year.
Perceived parental support
Participants reported perceived parental support at cycles 9 (age 12), 19 (age 16) and 22 (age 17) through 5 components of support for each of their parents using the Parental Support Scale . The 5 components represent parental co-participation (Did your father (mother) participate in PA or play sports with you?), transportation (Did your father (mother) bring you to a place where you can do PA or play sports?), watching (Did your father (mother) watch you do PA or play sports?), encouragement (Did you father (mother) encourage to practice PA?) and informational (Did your father (mother) tell you that PA is good for your health?). An additional item was added to capture modeling (Did your father (mother) do any PA or participate in sports?). For each item, scoring ranged from 1 (never/none) to 5 (nearly everyday). A tangible support score was computed by averaging items representing co-participation, transportation, and watching. Similarly, items representing encouragement, informational (advising) and modeling were averaged to represent intangible support. Both scores could range from 1 to 5, where higher scores reflect supportive parental practices. Average scores obtained for both parents was used for analyses. Two-week test-retest reliability of the scale had an ICC of 0.88 and a previous study demonstrated that youth’s responses on this scale are correlated (r = 0.6, p < 0.001) with responses from parents . Scale reliability in the current study was good (Cronbach’s alpha = 0.84).
Perceived parental control
Participants reported their perception of parental control at cycle 9 (age 12), 19 (age 16) and 22 (age 17) through one item representing ordering and one item representing nagging  for each of their parents. Specifically, participants responded to: “Did your father (mother) order you to do sports or PA?”, “Did your father (mother) annoy you to do sports or PA?”. For each item, scores ranged from 1 (never) to 5 (very often). Higher average scores capture perceptions of higher parental control. Internal consistency of scores in the current study was good (Cronbach’s alpha = 0.86) and a previous study supported the construct validity of this scale with reports of it being negatively correlated with measures of self-efficacy and enjoyment of physical activity .
Participants reported their gender and postal code, which was used to obtain their neighborhood average income, drawn from the 2011 National Household Survey census data expressed as tertiles herein. Schools were selected to represent living area (i.e. rural or urban) and cultural backgrounds (i.e. French or English language) .
Since missing data for MVPA and parental parameters affected less than 10% of the sample and Little’s test suggested that missing data were missing completely at random (P = 0.64), we considered missing data as inconsequential and did not pursue imputation. Complete cases for MVPA and perceived parental parameters were used in the analyses. To examine associations between parental support and MVPA over time, two-level linear mixed models accounting for clustering due to repeated measures were used. We also examined whether within-person variation in the outcome followed a linear or quadratic time trend. The quadratic term was not significant, so we did not retain it for subsequent models. In initial crude analyses, we modeled MVPA as a function of total support, tangible and intangible support, and parental control separately. Then, in partially adjusted models, MVPA was modeled as a function of each parental parameter with adjustments for time and covariates. A fully adjusted model was then computed by including all parental support and control variables and potentially confounding variables described above into one model. To test if the estimated relationship of parental support and control behaviours on MVPA changed as participants aged, we included interaction terms (e.g. parental support x age and parental control x age).
Similarly, a series of partially adjusted models and one fully adjusted model were computed to assess the independent association of each component of parental support behaviours (co-participation, transportation, watching, encouragement, informational and modeling) with MVPA. For each model above, we used an unstructured within-individual correlation structure as no prior assumption about the correlation was presumed and time points were unequally spaced for the measures of parenting. Because of previously documented gender differences in MVPA and perceived parental support , we tested interaction terms for moderation effects of parental support and parent control on MVPA by gender. These were not significant, so analyses were not stratified by gender. We used the Wald test to compare slopes associated with each individual form of parental support . Orthogonal polynomial contrasts were used to test MVPA across different covariates for linear trends . All models were carried out using mixed Stata command and were fitted using the REML method (significance level of α < 0.05).
To examine lagged associations of parental support and control on adolescents’ future MVPA, we used a cross-lagged panel design (sem Stata command) by following the crude, partially adjusted and fully adjusted sequence described above. In these models, we also adjust the estimate of the lagged variable effect for the dependent variable at the previous time point . Because the lagged-response model is recommended for use with approximately equally spaced time intervals between measurements, we built two separate sets of models (set 1: 12y vs. 16y and 12y vs. 17y; set 2: 16y vs. 17y). We used the full information maximum likelihood method and goodness of fit was assessed based on the Chi-square test (P > 0.05), root mean square error of approximation (RMSEA< 0.06), comparative fit index (CFI > 0.95), and Tucker-Lewis index (TLI > 0.95) .