ORIGINAL RESEARCH article

Applying social cognitive theory in predicting physical activity among chinese adolescents: a cross-sectional study with multigroup structural equation model.

\r\nJianxiu Liu,&#x;

  • 1 Division of Sports Science and Physical Education, Tsinghua University, Beijing, China
  • 2 Vanke School of Public Health, Tsinghua University, Beijing, China
  • 3 Department of Kinesiology, Hebei Institute of Physical Education, Shijiazhuang, China
  • 4 Department of Social Sciences, Hebei Sport University, Shijiazhuang, China

This cross-sectional study aimed to assess the applicability of social cognitive determinants among the Chinese adolescents and examine whether the predictability of the social cognitive theory (SCT) model on physical activity (PA) differs across gender (boys and girls) and urbanization (urban and suburban). A total of 3,000 Chinese adolescents ranging between the ages of 12–15 years were randomly selected to complete a set of questionnaires. Structural equation modeling (SEM) was applied to investigate the relationships between social cognitive variables and PA in the urbanization and gender subgroups. The overall model explained 38.9% of the variance in PA. Fit indices indicated that the structural model of SCT was good: root mean square error of approximation (RMSEA) = 0.047, (root mean square residual) RMR = 0.028, goodness of fit index (GFI) = 0.974, adjusted goodness of fit index (AGFI) = 0.960, Tucker–Lewis coefficient (TLI) = 0.971, and comparative fit index (CFI) = 0.978. Regarding the subgroup analysis, social support (critical ratios [CRs] = 2.118; p < 0.001) had a more substantial impact on the PA of adolescents in suburban areas than that in urban areas, whereas self-regulation (CRs = −2.896, p < 0.001) had a more substantial impact on the PA of adolescents in urban areas than in suburban areas. The results indicate that the SCT model predicts the PA of Chinese adolescents substantially. An SCT model could apply over a range of subgroups to predict the PA behavior and should be considered comprehensively when designing interventions. These findings would benefit PA among the Chinese adolescents, especially across genders and urbanization.

Introduction

Regular physical activity (PA) is essential for the health of adolescents, as it reduces the risk of chronic diseases, obesity, and mental health problems and improves cognitive health indicators ( Kruk, 2007 ; Lipnowski et al., 2012 ; Poitras et al., 2016 ). Despite the considerable evidence showing the benefits of PA, 80.3% of adolescents (13–15 years old) failed to meet the recommendation of 60 min of PA daily ( Hallal et al., 2012 ). Physical inactivity among adolescents is widespread worldwide ( Sisson et al., 2010 ; Patnode et al., 2011 ). In 2016, adolescents aged 11–17 years had a prevalence of physical inactivity of 81.0% (77.6% of boys and 84.7% of girls) globally ( Guthold et al., 2020 ). According to a 2017 survey in China, only 34.1% of children and adolescents met the recommendation of 60 min or more of moderate to vigorous PA (MVPA) per day ( Zhu et al., 2019 ). Evidence indicates that the habits of PA behavior established in adolescence are likely to track into adulthood ( Patnode et al., 2011 ; Hayes et al., 2019 ). Therefore, engaging in PA during this period is particularly important for the whole life span of the individual.

Evidence suggests that interventions using health behavior theories would effectively change the population-level behavior in “real world” contexts ( Hagger and Weed, 2019 ) and may help in explaining the maintenance of health-related behaviors ( Kwasnicka et al., 2016 ). Furthermore, health behavior theories provide an effective framework to understand the mediators and moderators of the behavior. Bandura (1986) proposed that human behavior, personal factors (such as, cognition), and environmental factors are affected by each other within a framework of reciprocal determinism, termed as triadic reciprocal causation. Social cognitive theory (SCT) represents a causal model in which the self-efficacy is set to the influence human behavior directly and indirectly via other mediating processes that include outcome expectations, social support, and self-regulation ( Bandura, 1991 , 1998 , 2004 ; Bandura et al., 1997 ). Self-efficacy reflects the judgment of own ability of an individual to accomplish a specific health behavior. Outcome expectancy reflects the individual’s perception of the likely social, physical, self-evaluation outcome of completing a specific health behavior. Social support is the perceived support for health behaviors from important others, such as family and friends. Self-regulation operates through a set of psychological subfunctions to influence the health behavior (e.g., self-monitoring, judgmental, and self-reactive influences).

Key SCT determinants of PA include social supports, self-efficacy, outcome expectations, and self-regulation of participating in PA ( Bandura et al., 1997 ; Anderson-Bill et al., 2011a ; Ramirez et al., 2012 ). In previous studies, SCT indicates that social and psychological determinants influence the behavior ( Bandura, 1986 , 1998 ) and that self-efficacy is the most relevant determinant to PA in SCT among all key determinants ( Dewar et al., 2013 ; Ah Hong et al., 2017 ; Beauchamp et al., 2019 ). Researchers have used SCT to predict the PA behavior in different countries (e.g., the United States, Australia, and Iran) ( Gao, 2012 ; Dewar et al., 2013 ; Bagherniya et al., 2015 ). The SCT has been widely applied to explain and change the PA behavior across a wide range of age groups and proved successful. Researchers found that the SCT serves as a good framework for researchers studying health promotion and PA in the parents of African American children ( Webber-Ritchey et al., 2018 ). Furthermore, SCT was proved to be effective when used in the intervention in persons with multiple sclerosis and breast cancer ( Auster-Gussman et al., 2021 ; Silveira et al., 2021 ). The overall predictability of SCT ranged from 5 to 52% among adolescents ( Taymoori et al., 2008 ; Lubans et al., 2011 ; Martin et al., 2011 ). However, there are not many pieces of research that use SCT to predict the PA behavior in Chinese adolescents.

The positive effects of social-cognitive determinants on PA prompted researchers to examine the invariance of SCT and the differences between the subgroups of different demographic variables. Identifying the differences in the predictive power of the SCT model and its key determinants among the demographic subgroups is essential for improving our understanding of PA behavior and designing effective interventions. One study reported the predictive power of SCT on PA among urban and underserved middle school students and found that the overall SCT predictability was 19% in urban areas and 12% in underserved areas ( Martin et al., 2011 ). There are numerous gaps in the levels of PA between adolescents living in urban and suburban areas. These differences can be partially explained by socioeconomic status, environment/equipment barriers, social support, and other factors ( Felton et al., 2002 ; Hartley, 2004 ). However, few studies have compared the difference of social-cognitive determinants in urban and suburban areas. Additionally, the predictive power of SCT’s determinants has been analyzed in several research studies. Self-efficacy, parental social support, and friend social support showed no differences between boys and girls ( Martin and McCaughtry, 2008 ). In contrast, boys are more physically active when a parent praises them for being physically active ( Ah Hong et al., 2017 ). Therefore, the predictive power of SCT and its determinants for PA in different gender groups should be further confirmed.

In terms of the estimates of invariance, few studies have confirmed the invariance of SCT across different gender or community subgroups ( Beauchamp et al., 2019 ). Examining the invariance is important to ensure that any differences reported in communities or genders are not merely a function of differences in interpretation of the measures. There is a possibility that the responses to items may be influenced by gender or region of respondents ( Duda and Hayashi, 1998 ). Investigating factorial invariance across the country, age, or gender subgroups allows the examination of the construction of questionnaires, which may result in the measurement of a latent construct similarly across samples. Furthermore, a majority of previous research studies focus on adolescents from western countries. People of Eastern cultural backgrounds have been long overlooked. Therefore, the present study aims to (a) assess the generalizability of the SCT measurements among the Chinese adolescents, (b) test the invariance of the SCT model across gender and urbanization, and (c) examine whether the predictive power of the SCT model on PA differs across the gender (boys and girls) and community subgroups (urban and suburban).

According to evidence provided by previous studies in other countries ( Dewar et al., 2013 ; Ah Hong et al., 2017 ), we hypothesize that (1) SCT could predict the PA behavior of Chinese adolescents, and that (2) there are no differences in the predictive power of SCT on PA behavior in different gender subgroups. Moreover, previous studies reported numerous gaps in PA levels among adolescents living in urban and suburban areas ( Felton et al., 2002 ; Hartley, 2004 ). Support from parents and social community members might differ between these two subgroups ( Ma et al., 2019 ). Thus, we hypothesize that (3) social support and self-regulation would differ in urban and suburban subgroups to predict PA.

Materials and Methods

Design and participants.

A cross-sectional survey was conducted in three cities in China (Beijing in the northeast, Shanghai in the southeast, and Urumchi in the west) from March to May 2018. A total of 3,000 students ranging between 12 and 15 years were randomly selected from local middle schools in each city to complete the questionnaires. The sample size in a structural equation modeling (SEM) study is usually 10 times the number of variables according to the rule of thumb. Thus, the sample size is sufficient for this study. The exclusion criteria are as follows: (a) those who were not between 12 and 15 years old; (b) those who were unable to participate in normal physical activity in the previous week due to injuries or sickness; and (c) those who completed less than 50% of their questionnaire. In total, 2,502 valid questionnaires were obtained, and the validity rate of the questionnaire was 83.4%. Missing values in the dimensions of social-cognitive variables were imputed with mean values of each dimension.

Simple random sampling was used in the study. First, all the schools in these cities were coded. A random number generator was used to randomly select 18 schools, 11 of them from urban areas and 7 from suburban or semirural areas. Second, the teachers in each selected school were contacted, and the purpose of the study was explained to acquire the permission for contacting their students. With approval, the survey was conducted in the classroom setting, and participants were assured that the participation was voluntary and that they were free to withdraw at any time. Before administering the survey, we explained that it was not a test and that we were interested in the actual PA and social cognitive status of participants. All answers were confidential, and all identifying information was kept anonymous to minimize the risk of social desirability bias. Participants were reminded to complete all the questions within 15–25 min. Informed consent was obtained from each student and their parents. Ethics approval to conduct the study was obtained from the ethics committee at the host university.

Instruments

Participants were asked to fill out their demographic information (e.g., age, gender, and living area), SCT-related psychological scales, and the PA questionnaire.

Self-efficacy was measured by using the Self-efficacy for Exercise Scale (SES). SES was used to test the situational confidence in persisting with exercising ( Benisovich et al., 1998 ). The head of the scale reads, “please evaluate how much confidence you have when participating in regular PA during your leisure time.” The 28-items SES uses a five-point Likert-type scale ranging from not confident at all (1) to completely confident (5), and consists of six dimensions ( Norman, 1998 ): (a) negative effects, (b) resistance from others, (c) making excuses, (d) bad weather, (e) exercising alone, and (f) inconveniency. A sample questionnaire item was “I’m tired.” This scale has been proven to have good validity and reliability in assessing psychological problems among Chinese adolescents (Cronbach’s alpha = 0.91) ( Yin, 2007 ; Ren et al., 2020 ). The reliability test of the scales in the present study is α = 0.93.

Self-regulation was measured by using the Exercise Goal-Setting Scale (EGS) and the Exercise Planning and Scheduling Scale (EPS) ( Weinberg, 2002 ). A five-point Likert scale from 1 (“does not describe”) to 5 (“describes completely”) was used for scoring both the EGS and EPS (10-items each) ( Rovniak et al., 2002 ). The EGS and EPS have been proven to be valid and reliable in assessing psychological problems among Chinese adolescents (Cronbach’s alpha = 0.88) ( Zhao, 2008 ; Xu et al., 2017 ). The head of the scale reads, “please evaluate whether the following descriptions are consistent with your when participating in PA.” A sample item was “I always schedule a week’s exercise time in advance.” The reliability test of the scales in the present study is α = 0.91.

Outcome expectation was measured by a scale developed by Lewis for the Physical Activity for Risk Reduction study and was further amended by Lewis et al. (1993) and Rogers et al. (2007) . Participants were asked to rate their agreement on statements that include 12 psychological, social, or physical benefits on a five-point Likert scale (1 = strongly disagree to 5 = strongly agree). The scale has proven good validity and reliability in assessing psychological problems among Chinese adolescents (Cronbach’s alpha = 0.94) ( Xu et al., 2020 ). The head of the scale reads, “please select the appropriate option and feeling when you participate in regular PA.” A sample item was “Regular exercise will increase my muscle mass.” The reliability test of the scales in the present study is α = 0.90.

Social support was measured by a scale constructed by Sallis et al. (1987) . The items described supportive actions and words that encourage PA. Subjects were asked to rate family and friends’ frequency to perform as the items described. The scale has been proven to have good validity and reliability among the Chinese and American children and adolescents ( Hsu Y. W. et al., 2011 ; Liang et al., 2014 ). The scale used 12 items for family members and 10 items for friends on a five-point scale ranging from 1 (never) to 5 (always). The head of the scale reads, “please rate what your family (e.g., parents, siblings, and grandparents) and friends (e.g., close friends, acquaintances, classmates) have said and done in the past 3 months concerning PA.” An example item is “Offer to exercise with me.” The reliability test of the scales in the present study is α = 0.853.

Physical activity was measured by the PA questionnaire for adolescents (PAQ-A) developed by Kowalski et al. (2004) . The PAQ-A is a self-administered, 7-day recall instrument. It was developed to measure the general levels of PA for adolescents. The questionnaire classifies PA of adolescents into different activity levels and investigates the relationship between PA and health outcomes ( MacKelvie et al., 2001 ). A five-point Likert scale from 1 (“Never”) to 5 (“Above six times a week”) was used to measure that how many times subjects participated in PA in a whole week. PA level is classified as low (≤2), moderate (>2 and ≤3), and high activity (>3), according to the mean score of nine items ( Chen et al., 2008 ). An example question is, “In the past 7 days, how many days have you exercised after school (not including weekends)?” PAQ-A has been proven to have good validity and reliability in assessing psychological problems among adolescents ( Voss et al., 2017 ).

Data Analyses

Descriptive statistics (i.e., means and SD) were computed using Stata 15.0. SEM was performed by AMOS 25.0, and the model was estimated using the maximum likelihood techniques to investigate the relationships between social cognitive variables and PA in urbanization and gender subgroups ( Byrne, 2010 ). First, a confirmatory factor analysis (CFA) was conducted considering the latent variables and observed variables following Yuan et al.’s method ( Yuan et al., 1997 ). A CFA was conducted to test the validity of the scale and the relationship between the latent variable. All standardized factor loading within this single factor should be larger than 0.5 and statistically significant. Model fit was assessed using root mean square error of approximation (RMSEA), root mean square residual (RMR), comparative fit index (CFI), normed fit index (NFI), incremental fit index (IFI), and goodness of fit index (GFI). The accepted cut-offs for the values of CFI, NFI, IFI, and GFI should be greater than 0.90; the thresholds for RMSEA and RMR should be less than 0.05.

Second, the overall fit of the resultant models was computed to assess the predictability of SCT among the Chinese adolescents. A number of the goodness of fit indices representing absolute, comparative, and residual aspects of fit were used, specifically chi-square (χ 2 ), degree of freedom ( df ), RMSEA, RMR, GFI, CFI, adjusted goodness of fit index (AGFI), Tucker–Lewis coefficient (TLI), parsimony goodness of fit index (PGFI), and parsimony normalized fit index (PNFI). The RMSEA and RMR of less than 0.05 indicate excellent model fit ( Loehlin and Beaujean, 2016 ). The GFI, AGFI, TLI, and CFI greater than 0.9 are excellent. PGFI and PNFI greater than 0.5 are excellent ( Bollen and Long, 1993 ).

Third, the invariance analysis across groups is a logical prerequisite for conducting the multigroup comparisons ( Vandenberg and Lance, 2000 ). Invariance models within the pattern of relationships among theoretical constructs (i.e., covariances) and the latent mean difference were estimated to compare if the SCT operates equivalently across the gender and urbanization subgroups. The differences between the groups can be evaluated by examining differences between the models that assume equalities among the parameters with models. A theoretical model is separately applied to each subgroup, and then the invariance analyses are conducted. Before the invariance models can be estimated, it must be established that a model without any invariances (i.e., a model that is different in each group) is reasonable. This model can be used as a basis of assessments for more constrained models. Thus, Model 1 represents unrestricted model (non-invariant, unconstrained model); Model 2 represent measurement equivalent model: (identical factor loading across the subsamples); Model 3 includes Model 2 constraints plus equal factor variances and covariances; Model 4 including Model 3 constraints plus equal paths; Model 5 includes Model 4 constraints plus equal factor residuals (“fully constrained”). Models 4 and 5 refer to the latent construct level, which deals with more substantive information about how subsamples may differ and are similar. The chi-square difference test and the TLI were applied to test the equality constraints to compare the models. If the χ 2 difference is statistically significant, then evidence of cross-group inequality exists. However, the χ 2 difference test would be too strict for a large sample study ( Quintana and Maxwell, 1999 ). TLI estimates the models for the groups separately and sums the chi-squares and the degrees of freedom. A difference of more than 0.05 in TLI or 0.01 in CFI is considered trivial in practical terms ( Cheung and Rensvold, 2002 ; Nigg et al., 2009 ). To complete the final research objective (i.e., determining the explained variance and comparing the strength of regression paths of the SCT constructs in predicting PA across democratic groups), pairwise critical ratios (CRs) for differences between parameters were examined to determine if there were significantly different regression paths for the demographic groups. A CR value larger than 1.96 indicates statistically significant differences in the latent mean.

Table 1 demonstrates the characteristics of the participants. A total of 2,502 adolescents had a mean age of 13.26 ( SD = 0.87). Most of the participants were 13 or 14 years old; 779 (31.1%) individuals were from Beijing, 388 (15.5%) were from Shanghai, and 1,335 (53.4%) were from Urumqi. Of all the participants, 48% were boys, and 52% were girls. Meanwhile, 58.6% of the participants were from urban areas, and 41.4% were from the suburbs.

www.frontiersin.org

Table 1. Descriptive statistics of participants.

Evaluation of the Measurement and Structural Models

Measurement model.

A test of the measurement model indicated a highly satisfactory fit to the data in all five measures: RMSEA = 0.047, CFI = 0.980, NFI = 0.977, IFI = 0.980, and GFI = 0.976. All the factor loading of self-efficacy, self-regulation, outcome expectation, and social support ranged from 0.50 to 0.95 ( p < 0.001), which confirmed the convergent validity of the indicators ( Anderson and Gerbing, 1988 ). Thus, the measurement model was used to test the hypothetical structural model (the figure attached in Appendix A shows the validity of dimensions).

Structural Equation Model

The structural model was tested with all the paths depicted in Figure 1 . Fit indices indicated that the structural model was good: RMSEA = 0.047, RMR = 0.028, GFI = 0.974, AGFI = 0.960, TLI = 0.971, and CFI = 0.978. The overall model explained 38.9% of the variance in PA.

www.frontiersin.org

Figure 1. The structural equation model of the overall sample. NE, negative effects; RO, resistance of others; ME, making excuses; BW, bad weather; EA, exercising alone; IC, inconveniency. * indicates p < 0.05, ** indicates p < 0.01, *** indicates p < 0.001.

Direct effects, indirect effects, and the portion of total effects mediated by other variables reflect how SCT variables influence PA ( Table 2 ). The total effect of a latent variable on PA is the combination of direct and indirect effects. Social support had the greatest total effect on PA (β total = 0.583, p < 0.001). Higher levels of social support led to the higher levels of self-efficacy, outcome expectations, and self-regulation. Self-regulation exerted a moderate total effect on PA (β total = 0.221, p < 0.001). Higher levels of self-regulatory skills, such as goal setting and planning, directly resulted in the higher PA levels. Self-efficacy had a small total effect on PA (β total = 0.153, p < 0.001). Outcome expectations exhibited a small total effect on PA (β total = 0.093, p < 0.05).

www.frontiersin.org

Table 2. The direct, indirect, and total effects of the variables in the social cognitive theory (SCT) model.

Multigroup Analysis

Multigroup SEMs were performed to examine the invariances of the SCT model across the multigroup and whether SCT operates equivalently in regression paths in predicting the PA across gender (boys and girls) and urbanization (urban and suburban).

Urbanization

The SEM was tested in urban and suburban adolescents. In both groups, the hypothesized model is a good representation of the data, as shown in Table 3 . Given that the model results in good fit indices for urban and suburban groups, a multigroup SEM with latent variables was conducted to explore which parameters could be invariant across the urbanization. Table 4 presents the constrained models (Models 2–5) and the different tests of these models with the unconstrained model (Model 1). The RMR, GFI, AGFI, TLI, CFI, and RMSEA indicate excellent model fit. By examining the differences between the constrained and unconstrained models, all models appear to be significantly different in any case at the 1% level. The TLI and IGF indices suggest negligible differences between Models 2, 3, and 4 to the unrestricted model. Thus, Model 1, the unrestricted model, will be used to assess the path coefficients in the two groups ( Figure 2 ). The architecture of the social-cognitive variables on PA was tested to see differences between the urban and suburban groups. After comparing CRs for differences between parameters, social support (CR = 2.118; p < 0.001) had a more substantial impact on the PA of adolescents in suburban areas than that in urban areas, whereas self-regulation (CR = − 2.896, p < 0.001) had a more substantial impact on the PA of adolescents in urban areas than in suburban areas.

www.frontiersin.org

Table 3. The goodness of fit index (GFI) of urban and suburban groups.

www.frontiersin.org

Table 4. Nested Models of Multigroup SEM across urban and suburban.

www.frontiersin.org

Figure 2. The multigroup structural equation modeling (SEM) across urban and suburban. *** indicates p < 0.001, otherwise p > 0.05. Bold indicates significant differences of the path coefficients (critical ratios [CRs] > 1.96).

The SEM was tested in male and female groups. The hypothesized model was a good representation of the data in both groups, as shown in Table 5 . Given that the model offered a good fit for both male and female groups, a multi-structural equation model with latent variables was carried out to study in-depth which parameters could be considered invariant across gender. Table 6 presents the fit indices of the constrained models and the different tests of these models with the unconstrained model. The RMR, GFI, AGFI, TLI, CFI, and RMSEA indicate excellent model fit. The differences between the constrained and unconstrained models indicate that all models appear to be significantly different in all cases at the 1% level. Thus, Model 1, the unrestricted model, was used to assess the path coefficients in the two groups ( Figure 3 ). The architecture of the social-cognitive variables on PA was tested to determine whether there were differences between the urban and suburban groups. After comparing CR for differences between the parameters, the only significant difference was that social support has a more substantial impact on the self-efficacy of male adolescents than on female adolescents. There was no difference in path regression from social cognitive determinants to PA for gender.

www.frontiersin.org

Table 5. The goodness of fit index.

www.frontiersin.org

Table 6. Nested Models of Multigroup SEM across male and female.

www.frontiersin.org

Figure 3. The multigroup comparison of SEM across urban and suburban. Path coefficients are reported male/female subgroup. *** indicates p < 0.001, otherwise p > 0.05. Bold indicates significant differences of the path coefficients (CRs > 1.96).

This study investigated the predictability of SCT among the Chinese adolescents. The overall model explained 38.9% of the variance in PA. Social support had the greatest effect, self-regulation exerted a moderate effect, and self-efficacy and outcome expectations had a small effect on PA. Negligible differences between the constrained and unstrained models proved the invariance of the SCT model across urbanization and gender. The findings indicate that the SCT model applies to various subgroups and should be considered comprehensively when designing interventions. Specifically, social support had a more substantial impact on the PA of adolescents in suburban areas and urban areas, whereas self-regulation had a more substantial impact on the PA in urban areas than in suburban areas. No difference in path regression from social cognitive variables to PA was found across genders.

Two studies conducted in China used the SCT model to predict the exercise frequency of breast cancer survivors ( Hsu H. T. et al., 2011 ) and obesity prevention behaviors (e.g., TV watching, water consumption, fruits and vegetables consumed, and PA) of children ( Murnan et al., 2006 ). To the best of our knowledge, ours is the first article to investigate the predictive power of SCT to PA among Chinese adolescents. Consistent with studies in other countries ( Gao, 2012 ; Dewar et al., 2013 ; Bagherniya et al., 2015 ), we found that the SCT model can predict the PA of Chinese adolescents, which confirmed the first hypothesis. The overall model explained 38.9% of the variance in PA, which can be assessed as substantial ( Cohen, 1988 ; Cohen et al., 2013 ). The results indicate that the intervention based on SCT could effectively change the PA behavior among the Chinese adolescents. In the case of self-efficacy and PA behavior, although previous studies found self-efficacy as a significant predictor of PA ( Gao, 2012 ; Dewar et al., 2013 ; Lee et al., 2018 ), the association between PA and self-efficacy was weak in our study. However, social support was the most critical factor for predicting the PA behavior among Chinese adolescents. The following factors could explain this finding: First, most adolescents in China are engaged in their academic schedule and cannot allocate their time to participate in PA by themselves though they have high self-efficacy. Second, self-efficacy has different categories (e.g., barrier self-efficacy; proxy self-efficacy; and support self-efficacy), which may confine the consistency of the results. In addition, due to the collectivist Chinese culture, their behavior may not be in the control of an individual. They often participate in PA with friends or classmates and have PA scheduled by parents. Further studies that explore the types and sources of social support relevant to adolescents are recommended.

Self-regulation had a moderate effect on PA. Self-regulation is an important set of strategies that could help students gain control over the PA behavior, indicating that individuals need improved goal setting and planning skills to engage in PA. Self-regulation was shown to be a key to success in exercising and adherence to an exercise program, preventing unwanted behavioral tendencies, and focusing on the control of PA behavior ( Bandura et al., 1997 ). Therefore, self-regulation could play a major role in achieving exercise goals, as it serves as an important psychological factor for promoting the positive health behavior, including exercise ( Bandura, 2005 ). This finding has supported the results of studies that found high self-regulation ability predicted the exercise behavior ( Anderson et al., 2006 ; Dishman et al., 2014 ; Gerber et al., 2015 ) and was consistent with the finding that people with high self-regulation abilities participated in the PA for more extended periods ( Stadler et al., 2009 ).

Previous studies did not find any relationship between outcome expectations and PA in children ( Rovniak et al., 2002 ; Gao et al., 2009 ). Our findings found that outcome expectancy had a small effect on the PA behavior, which might be due to the following reasons: first, the adolescents may not sufficiently realize the harmful consequences of physical inactivity, but they recognize the positive social, physical, and psychological expectations of being physically active. Outcome expectancy was reported to be a better predictor of PA among older adults than younger adolescents ( Anderson-Bill et al., 2011b ; Kosteli et al., 2016 ). Second, positive PA outcome expectancy of adolescents may not directly translate to the PA behavior ( Hankonen et al., 2017 ). Thus, exploring the additional constructs impacting the relationship between outcome expectancy and PA may be warranted in the future.

Measurement invariance testing for SCT has been less studied. Most researchers would assume that the instruments operate the same way and contain the same constructs across groups. However, the invariance analysis across groups is a logical prerequisite for conducting multigroup comparisons ( Vandenberg and Lance, 2000 ). If not tested, violations of measurement equivalence assumptions threaten substantive interpretations, as it shows the inability to demonstrate reliability and validity ( Vandenberg and Lance, 2000 ). Our findings indicate that the SCT model has the same construct with items associated equally with the factors for the subgroups of urbanization and gender, which indicated the invariance of SCT across urbanization and gender for PA. Future studies could compare the SCT model across different urbanization and gender subgroups.

Social support and self-regulation were found to differ in urban and suburban subgroups, which confirmed our third hypothesis. Different conclusions were obtained in previous studies concerning the predictability of social support to PA. Laird et al. (2016) found that social support did not predict PA in adolescent girls, even though parents and friends may enhance the PA behavior ( Laird et al., 2016 ). However, Martin’s study found that social support is one of the best predictors of PA among underserved middle school students but did not compare the differences between urban school students with underserved school students ( Martin et al., 2011 ). The findings of our study show that the association between social support and PA was higher in suburban than urban areas. The results indicate that Chinese adolescents in suburban areas might not receive sufficient support from their parents and friends for PA, which may be a restricted factor. The association between self-regulation and PA was higher in urban than in suburban areas, which may be explained by the higher levels of social support, quality of PA environments, and quality of PA equipment that adolescents in urban areas receive and have access to. Therefore, the strategic use of skills to remind, cue, or reinforce PA behavior may be more related to PA in urban areas than suburban areas.

The results largely confirm the second hypothesis that there are no differences in the predictive power of SCT on PA behavior in different gender subgroups. This study found no difference between the predictive power of social-cognitive variables to PA across gender subgroups, consistent with previous research ( Martin and McCaughtry, 2008 ; Chen et al., 2019 ). Results show that the social-cognitive determinants are positively related to PA for boys and girls. One study found that boys were more active and received more support from siblings than girls. The remaining variables did not vary with gender (self-efficacy, parental social support, and friend social support) ( Martin and McCaughtry, 2008 ). Other studies have conflicting results regarding the association between gender and social support. One study found that gender did not moderate the relationship between social support and PA in the Chinese adolescents ( Chen and Dai, 2016 ); however, another reported that boys in primary school were more physically active when a parent praised them for being physically active ( Ah Hong et al., 2017 ), and girls perceived more social support than boys in PA in middle school ( Martin and Smith, 2002 ). Future research should fully explore the relationship between different social support and adolescent PA.

Thus, the SCT model can predict the PA of Chinese adolescents effectively. SCT model could apply to various subgroups to predict the PA behavior and should be considered comprehensively when designing future interventions. However, results from this investigation should be viewed in light of certain limitations. Although the study used Bandura’s classical SCT model, additional relevant PA determinants were not included (e.g., school, peer, and sibling support for PA, barriers, and sociostructurally factors). Second, the self-report questionnaires may have inconsistencies with actual experiences or social desirability bias. Third, given the correlational design of the study, causality cannot be argued. Our findings proved the predictability of SCT on PA among Chinese adolescents, especially the importance of social support and self-regulation. The study investigated the invariances of an SCT model and the difference in the architecture of the social-cognitive variables across urbanization and gender subgroups. The findings indicate that future possibilities for promoting PA interventions among Chinese adolescents should include those variables. However, a questionnaire was used to measure PA due to the large sample size in this study. Future studies should consider measuring PA behavior using accelerometers that can objectively measure PA. Furthermore, future studies should look beyond SCT models and use these in conjunction with broader social-ecological models that incorporate determinants at multiple levels (policy and environment), which could further understand determinants of PA participation among adolescents ( Sallis and Owen, 2015 ).

Data Availability Statement

The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author.

Ethics Statement

The studies involving human participants were reviewed and approved by the Ethical Committee of the Medical Association of Tsinghua University approved the study (IRB#20190093). Written informed consent to participate in this study was provided by the participants’ legal guardian/next of kin.

Author Contributions

XM contributed to the idea and supervision. JL wrote the original draft preparation and edited the manuscript. MZ analyzed and interpreted the data. YZ analyzed part of the data. DW and BS revised the manuscript. All authors have read and agreed to the published version of the manuscript.

This study was supported by the National Social Science Fund of China (No. 16BTY065).

Conflict of Interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Publisher’s Note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Ah Hong, S., Peltzer, K., and Wimonpeerapattana, W. (2017). Impact of self-efficacy and parenting practice on physical activity among school children. Nagoya J. Med. Sci. 79, 339–349. doi: 10.18999/nagjms.79.3.339

PubMed Abstract | CrossRef Full Text | Google Scholar

Anderson, E. S., Wojcik, J. R., Winett, R. A., and Williams, D. M. (2006). Social-cognitive determinants of physical activity: the influence of social support, self-efficacy, outcome expectations, and self-regulation among participants in a church-based health promotion study. Health Psychol. 25, 510–520. doi: 10.1037/0278-6133.25.4.510

Anderson, J. C., and Gerbing, D. W. (1988). Structural equation modeling in practice: a review and recommended two-step approach. Psychol. Bull. 103, 411–423. doi: 10.1037/0033-2909.103.3.411

CrossRef Full Text | Google Scholar

Anderson-Bill, E. S., Winett, R. A., and Wojcik, J. R. (2011a). Social cognitive determinants of nutrition and physical activity among web-health users enrolling in an online intervention: the influence of social support, self-efficacy, outcome expectations, and self-regulation. J. Med. Internet Res. 13:e28. doi: 10.2196/jmir.1551

Anderson-Bill, E. S., Winett, R. A., Wojcik, J. R., and Williams, D. M. (2011b). Aging and the social cognitive determinants of physical activity behavior and behavior change: evidence from the guide to health trial. J. Aging Res. 2011:505928. doi: 10.4061/2011/505928

Auster-Gussman, L. A., Gavin, K. L., Siddique, J., Welch, W. A., Solk, P., Whitaker, M., et al. (2021). Social cognitive variables and physical activity during chemotherapy for breast cancer: an intensive longitudinal examination. Psychooncology 9, 1–11. doi: 10.1002/pon.5820

Bagherniya, M., Sharma, M., Mostafavi, F., and Keshavarz, S. A. (2015). Application of social cognitive theory in predicting childhood obesity prevention behaviors in overweight and obese Iranian adolescents. Int. Q. Community Health Educ. 35, 133–147. doi: 10.1177/0272684x15569487

Bandura, A. (1986). Social Foundations of Thought and Action: A Social Cognitive Theory. Englewood Cliffs, NJ: Prentice Hall.

Google Scholar

Bandura, A. (1991). Social cognitive theory of self-regulation. Organ. Behav. Hum. Decis. Process. 50, 248—-287. doi: 10.1016/0749-5978(91)90022-L

Bandura, A. (1998). Health promotion from the perspective of social cognitive theory. Psychol. Health 13, 623–649. doi: 10.1080/08870449808407422

Bandura, A. (2004). Health promotion by social cognitive means. Health Educ. Behav. 31, 143–164. doi: 10.1177/1090198104263660

Bandura, A. (2005). The primacy of self-regulation in health promotion. Appl. Psychol. 54, 245–254. doi: 10.1111/j.1464-0597.2005.00208.x

Bandura, A., Freeman, W. H., and Lightsey, R. (1997). Self-efficacy: the exercise of control. J. Cogn. Psychother. 2, 158–166. doi: 10.1891/0889-8391.13.2.158

Beauchamp, M. R., Crawford, K. L., and Jackson, B. (2019). Social cognitive theory and physical activity: mechanisms of behavior change, critique, and legacy. Psychol. Sport Exerc. 42, 110–117. doi: 10.1016/j.psychsport.2018.11.009

Benisovich, S., Rossi, J., Norman, G., and Nigg, C. (1998). “A multidimensional approach to exercise self-efficacy: relationship with exercise behaviour and attitudes towards exercise,” in Paper Presented at the Annual Meeting of the New England Psychology Association , Boston, MA. doi: 10.1186/s12913-016-1423-5

Bollen, K. A., and Long, J. S. (1993). Testing Structural Equation Models . Sage Focus Editions, Vol. 154. Newbury Park, CA: Sage.

Byrne, B. M. (2010). Structural Equation Modeling With AMOS Basic Concepts, Applications, and Programming (Multivariate Applications Series). New York, NY: Routledge.

Chen, H., and Dai, J. (2016). Does gender moderate the direct and indirect relationships between different sources of social support and adolescents’ physical activity? J. Phys. Act. Health 13, 874–881. doi: 10.1123/jpah.2015-0547

Chen, H., Dai, J., and Gao, Y. (2019). Measurement invariance and latent mean differences of the Chinese version physical activity self-efficacy scale across gender and education levels. J. Sport Health Sci. 8, 46–54. doi: 10.1016/j.jshs.2017.01.004

Chen, S. R., Lee, Y. J., Chiu, H. W., and Jeng, C. (2008). Impact of physical activity on heart rate variability in children with type 1 diabetes. Childs Nerv. Syst. 24, 741–747. doi: 10.1007/s00381-007-0499-y

Cheung, G. W., and Rensvold, R. B. (2002). Evaluating goodness-of-fit indexes for testing measurement invariance. Struct. Equ. Model. 9, 233–255. doi: 10.1207/S15328007SEM0902_5

Cohen, J. (1988). Statistical Power Analysis for the Behavioral Sciences , 2nd Edn. New York, NY: Lawrence Erlbaum Associates.

Cohen, J., Cohen, P., West, S. G., and Aiken, L. S. (2013). Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences. New York, NY: Taylor and Francis. doi: 10.4324/9780203774441

Dewar, D. L., Plotnikoff, R. C., Morgan, P. J., Okely, A. D., Costigan, S. A., and Lubans, D. R. (2013). Testing social-cognitive theory to explain physical activity change in adolescent girls from low-income communities. Res. Q. Exerc. Sport 84, 483–491. doi: 10.1080/02701367.2013.842454

Dishman, R. K., Jackson, A. S., and Bray, M. S. (2014). Self-regulation of exercise behavior in the TIGER study. Ann. Behav. Med. 48, 80–91. doi: 10.1007/s12160-013-9573-8

Duda, J. L., and Hayashi, C. T. (1998). “Measurement issues in cross-cultural research within sport and exercise psychology,” in Advances in Sport and Exercise Psychology Measurement , ed. J. L. Duda (Morgantown, WV: Fitness Information Technology), 471–483. doi: 10.1186/s12913-016-1423-5

Felton, G. M., Dowda, M., Ward, D. S., Dishman, R. K., Trost, S. G., Saunders, R., et al. (2002). Differences in physical activity between black and white girls living in rural and urban areas. J. Sch. Health 72, 250–255. doi: 10.1111/j.1746-1561.2002.tb07338.x

Gao, Z. (2012). Urban Latino school children’s physical activity correlates and daily physical activity participation: a social cognitive approach. Psychol. Health Med. 17, 542–550. doi: 10.1080/13548506.2011.647699

Gao, Z., Lee, A. M., Solmon, M. A., and Zhang, T. (2009). Changes in middle school students’ motivation toward physical education over one school year. J. Teach. Phys. Educ. 28, 378–399. doi: 10.1186/s13054-016-1208-6

Gerber, M., Lindwall, M., Brand, S., Lang, C., Elliot, C., and Pühse, U. (2015). Longitudinal relationships between perceived stress, exercise self-regulation and exercise involvement among physically active adolescents. J. Sports Sci. 33, 369–380. doi: 10.1080/02640414.2014.946072

Guthold, R., Stevens, G. A., Riley, L. M., and Bull, F. C. (2020). Global trends in insufficient physical activity among adolescents: a pooled analysis of 298 population-based surveys with 1⋅ 6 million participants. Lancet Child Adolesc. Health 4, 23–35.

Hagger, M. S., and Weed, M. (2019). DEBATE: do interventions based on behavioral theory work in the real world? Int. J. Behav. Nutr. Phys. Act. 16:36. doi: 10.1186/s12966-019-0795-4

Hallal, P. C., Andersen, L. B., Bull, F. C., Guthold, R., Haskell, W., and Ekelund, U. (2012). Global physical activity levels: surveillance progress, pitfalls, and prospects. Lancet 380, 247–257. doi: 10.1016/S0140-6736(12)60646-1

Hankonen, N., Heino, M. T. J., Kujala, E., Hynynen, S.-T., Absetz, P., Araújo-Soares, V., et al. (2017). What explains the socioeconomic status gap in activity? Educational differences in determinants of physical activity and screentime. BMC Public Health 17:144. doi: 10.1186/s12889-016-3880-5

Hartley, D. (2004). Rural health disparities, population health, and rural culture. Am. J. Pub. Health 94, 1675–1678. doi: 10.2105/ajph.94.10.1675

Hayes, G., Dowd, K. P., MacDonncha, C., and Donnelly, A. E. (2019). Tracking of physical activity and sedentary behavior from adolescence to young adulthood: a systematic literature review. J. Adolesc. Health 65, 446–454. doi: 10.1016/j.jadohealth.2019.03.013

Hsu, H. T., Dodd, M. J., Guo, S. E., Lee, K. A., Hwang, S. L., and Lai, Y. H. (2011). Predictors of exercise frequency in breast cancer survivors in Taiwan. J. Clin. Nurs. 20, 1923–1935. doi: 10.1111/j.1365-2702.2010.03690.x

Hsu, Y. W., Chou, C. P., Nguyen-Rodriguez, S. T., McClain, A. D., Belcher, B. R., and Spruijt-Metz, D. (2011). Influences of social support, perceived barriers, and negative meanings of physical activity on physical activity in middle school students. J. Phys. Act. Health 8, 210–219. doi: 10.1123/jpah.8.2.210

Kosteli, M. C., Williams, S. E., and Cumming, J. (2016). Investigating the psychosocial determinants of physical activity in older adults: a qualitative approach. Psychol. Health 31, 730–749. doi: 10.1080/08870446.2016.1143943

Kowalski, K. C., Crocker, P. R., and Donen, R. M. (2004). The Physical Activity Questionnaire for Older Children (PAQ-C) and Adolescents (PAQ-A) Manual , Vol. 87. Saskatoon, SK: College of Kinesiology, University of Saskatchewan.

Kruk, J. (2007). Physical activity in the prevention of the most frequent chronic diseases: an analysis of the recent evidence. Asian Pac. J. Cancer Prev. 8, 325–338.

PubMed Abstract | Google Scholar

Kwasnicka, D., Dombrowski, S. U., White, M., and Sniehotta, F. (2016). Theoretical explanations for maintenance of behaviour change: a systematic review of behaviour theories. Health Psychol. Rev. 10, 277–296. doi: 10.1080/17437199.2016.1151372

Laird, Y., Fawkner, S., Kelly, P., McNamee, L., and Niven, A. (2016). The role of social support on physical activity behaviour in adolescent girls: a systematic review and meta-analysis. Int. J. Behav. Nutr. Phys. Act. 13:79. doi: 10.1186/s12966-016-0405-7

Lee, C. G., Park, S., Lee, S. H., Kim, H., and Park, J. W. (2018). Social cognitive theory and physical activity among Korean male high-school students. Am. J. Mens Health 12, 973–980. doi: 10.1177/1557988318754572

Lewis, C. E., Raczynski, J. M., Heath, G. W., Levinson, R., Hilyer, J. C. Jr., and Cutter, G. R. (1993). Promoting physical activity in low-income African-American communities: the PARR project. Ethn. Dis. 3, 106–118.

Liang, Y., Lau, P. W. C., Huang, W. Y. J., Maddison, R., and Baranowski, T. (2014). Validity and reliability of questionnaires measuring physical activity self-efficacy, enjoyment, social support among Hong Kong Chinese children. Prev. Med. Rep. 1, 48–52. doi: 10.1016/j.pmedr.2014.09.005

Lipnowski, S., LeBlanc, C., Society, C. P., Living, H. A., and Committee, S. M. (2012). Healthy active living: physical activity guidelines for children and adolescents. Paediatr. Child Health 17, 209–210. doi: 10.1093/pch/17.4.209

Loehlin, J. C., and Beaujean, A. A. (2016). Latent Variable Models: An Introduction to Factor, Path, and Structural Equation Analysis. New York, NY: Taylor & Francis.

Lubans, D. R., Okely, A. D., Morgan, P. J., Cotton, W., Puglisi, L., and Miller, J. (2011). Description and evaluation of a social cognitive model of physical activity behaviour tailored for adolescent girls. J. Health Educ. Res. 27, 115–128. doi: 10.1093/her/cyr039

Ma, S., Jiang, M., Wang, F., Lu, J., Li, L., and Hesketh, T. (2019). Left-behind children and risk of unintentional injury in rural China-a cross-sectional survey. Int. J. Environ. Res. Public Health 16:403. doi: 10.3390/ijerph16030403

MacKelvie, K. J., McKay, H. A., Khan, K. M., and Crocker, P. R. (2001). A school-based exercise intervention augments bone mineral accrual in early pubertal girls. J. Pediatr. 139, 501–508. doi: 10.1067/mpd.2001.118190

Martin, J. J., and McCaughtry, N. (2008). Predicting physical activity in inner-city Hispanic American children. Hisp. Health Care Int. 6, 150–158.

Martin, J. J., McCaughtry, N., Flory, S., Murphy, A., andWisdom, K. (2011). Using social cognitive theory to predict physical activity and fitness in underserved middle school children. Res. Q. Exerc. Sport 82, 247–255. doi: 10.1080/02701367.2011.10599752

Martin, J. J., and Smith, K. (2002). Friendship quality in youth disability sport: perceptions of a best friend. Adapt. Phys. Activ. Q. 19, 472–482. doi: 10.1123/apaq.19.4.472

Murnan, J., Sharma, M., and Lin, D. (2006). Predicting childhood obesity prevention behaviors using social cognitive theory: children in China. Int. Q. Community Health Educ. 26, 73–84. doi: 10.2190/rv25-395t-25n3-pn14

Nigg, C. R., Lippke, S., andMaddock, J. E. (2009). Factorial invariance of the theory of planned behavior applied to physical activity across gender, age, and ethnic groups. Psychol. Sport Exerc. 10, 219–225. doi: 10.1016/j.psychsport.2008.09.005

Norman, G. J. (1998). A Cluster Analytic Test of the Transtheoretical Model Applied to Exercise Behavior . Doctoral dissertation. South Kingston, RI: University of Rhode Island.

Patnode, C. D., Lytle, L. A., Erickson, D. J., Sirard, J. R., Barr-Anderson, D. J., and Story, M. (2011). Physical activity and sedentary activity patterns among children and adolescents: a latent class analysis approach. J. Phys. Act. Health 8, 457–467. doi: 10.1123/jpah.8.4.457

Poitras, V. J., Gray, C. E., Borghese, M. M., Carson, V., Chaput, J. P., Janssen, I., et al. (2016). Systematic review of the relationships between objectively measured physical activity and health indicators in school-aged children and youth. Appl. Physiol. Nutr. Metab. 41, S197–S239. doi: 10.1139/apnm-2015-0663

Quintana, S. M., and Maxwell, S. E. (1999). Implications of recent developments in structural equation modeling for counseling psychology. Couns. Psychol. 27, 485–527. doi: 10.1177/0011000099274002

Ramirez, E., Kulinna, P. H., and Cothran, D. (2012). Constructs of physical activity behaviour in children: the usefulness of social cognitive theory. Psychol. Sport Exerc. 13, 303–310. doi: 10.1016/j.psychsport.2011.11.007

Ren, Z., Hu, L., Yu, J. J., Yu, Q., Chen, S., Ma, Y., et al. (2020). The influence of social support on physical activity in Chinese adolescents: the mediating role of exercise self-efficacy. Children 7:23. doi: 10.3390/children7030023

Rogers, L. Q., Courneya, K. S., Shah, P., Dunnington, G., and Hopkins-Price, P. (2007). Exercise stage of change, barriers, expectations, values and preferences among breast cancer patients during treatment: a pilot study. Eur. J. Cancer Care 16, 55–66. doi: 10.1111/j.1365-2354.2006.00705.x

Rovniak, L. S., Anderson, E. S., Winett, R. A., and Stephens, R. S. (2002). Social cognitive determinants of physical activity in young adults: a prospective structural equation analysis. Ann. Behav. Med. 24, 149–156. doi: 10.1207/S15324796ABM2402_12

Sallis, J. F., Grossman, R. M., Pinski, R. B., Patterson, T. L., and Nader, P. R. (1987). The development of scales to measure social support for diet and exercise behaviors. Prev. Med. 16, 825–836. doi: 10.1016/0091-7435(87)90022-3

Sallis, J. F., and Owen, N. (2015). “Ecological models of health behavior,” in Health Behavior: Theory, Research, and Practice , eds K. Glanz, B. K. Rimer, and K. V. Viswanath (San Franciso, CA: Jossey-Bass/Wiley), 43–64.

Silveira, S. L., Froehlich-Grobe, K., and Motl, R. W. (2021). Developing a community-engaged wheelchair exercise program for persons with MS: community advisory board formation and feedback. Disabil. Rehabil. Assist. Technol. 12, 1–8. doi: 10.1080/17483107.2021.2010819

Sisson, S. B., Broyles, S. T., Baker, B. L., and Katzmarzyk, P. T. (2010). Screen time, physical activity, and overweight in U.S. youth: national survey of children’s health 2003. J. Adolesc. Health 47, 309–311. doi: 10.1016/j.jadohealth.2010.02.016

Stadler, G., Oettingen, G., and Gollwitzer, P. M. (2009). Physical activity in women: effects of a self-regulation intervention. Am. J. Prev. Med. 36, 29–34. doi: 10.1016/j.amepre.2008.09.021

Taymoori, P., Rhodes, R. E., and Berry, T. R. (2008). Application of a social cognitive model in explaining physical activity in Iranian female adolescents. Health Educ. Res. 25, 257–267. doi: 10.1093/her/cyn051

Vandenberg, R. J., and Lance, C. E. (2000). A review and synthesis of the measurement invariance literature: suggestions, practices, and recommendations for organizational research. Organ. Res. Methods 3, 4–70.

Voss, C., Dean, P. H., Gardner, R. F., Duncombe, S. L., and Harris, K. C. (2017). Validity and reliability of the Physical Activity Questionnaire for Children (PAQ-C) and Adolescents (PAQ-A) in individuals with congenital heart disease. PLoS One 12:e0175806. doi: 10.1371/journal.pone.0175806

Webber-Ritchey, K. J., Taylor-Piliae, R. E., and Loescher, L. J. (2018). Physical activity in parents of young African American children: the application of social cognitive theory. Res. Theory Nurs. Pract. 32, 63–81. doi: 10.1891/0000-000Y.32.1.63

Weinberg, R. S. (2002). “Goal setting in sport and exercise: research to practice,” in Exploring Sport and Exercise Psychology , 2nd Edn, eds J. L. Van Raalte and B. W. Brewer (Washington, DC: American Psychological Association), 25–48.

Xu, H., Su, C., Ji, Y., Yin, F., Yang, Y., Yang, S., et al. (2020). Predicting physical exercise changes in Chinese rural adolescents: the application of the health action process approach model. Psychol. Health Med. 25, 639–651. doi: 10.1080/13548506.2019.1709653

Xu, X., Pu, Y., Sharma, M., Rao, Y., Cai, Y., and Zhao, Y. (2017). Predicting physical activity and healthy nutrition behaviors using social cognitive theory: cross-sectional survey among undergraduate students in Chongqing, China. Int. J. Environ. Res. 14:1346. doi: 10.3390/ijerph14111346

Yin, B. (2007). Research on Chinese College Students’ Physical Activity With Trans-Theoretical Model . Doctoral dissertation. Shanghai: East China Normal University.

Yuan, K.-H., Bentler, P. M., and Kano, Y. J. B. (1997). On averaging variables in a confirmatory factor analysis model. Behaviormetrika 24, 71–83.

Zhao, Y. (2008). Physical Activity Behavior of Chinese University Students: Social Cognitive Determinants . Master’s thesis. Beijing: Beijing Sport university.

Zhu, Z., Tang, Y., Zhuang, J., Liu, Y., Wu, X., Cai, Y., et al. (2019). Physical activity, screen viewing time, and overweight/obesity among Chinese children and adolescents: an update from the 2017 physical activity and fitness in China-the youth study. BMC Public Health 19:197. doi: 10.1186/s12889-019-6515-9

www.frontiersin.org

Appendix Figure 1 The confirmatory factor analysis (CFA) of the model and the validity of scales in different dimensions

Keywords : social cognitive theory, physical activity, Chinese, adolescents, structural equation model

Citation: Liu J, Zeng M, Wang D, Zhang Y, Shang B and Ma X (2022) Applying Social Cognitive Theory in Predicting Physical Activity Among Chinese Adolescents: A Cross-Sectional Study With Multigroup Structural Equation Model. Front. Psychol. 12:695241. doi: 10.3389/fpsyg.2021.695241

Received: 14 April 2021; Accepted: 27 December 2021; Published: 15 March 2022.

Reviewed by:

Copyright © 2022 Liu, Zeng, Wang, Zhang, Shang and Ma. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY) . The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

*Correspondence: Xindong Ma, [email protected]

† These authors have contributed equally to this work and share first authorship

Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.

U.S. flag

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

  • Publications
  • Account settings
  • Advanced Search
  • Journal List
  • Int J MS Care
  • v.23(1); Jan-Feb 2021

Social Cognitive Theory and Physical Activity in Older Adults with Multiple Sclerosis

Associated data, background:.

The expanding population of older adults with multiple sclerosis (MS) likely experiences many of the same benefits of physical activity (PA) as younger and middle-aged adults with MS. However, participation in PA is exceedingly low in this segment of the MS population. This study examined variables from social cognitive theory (SCT) as correlates of PA in older adults with MS to inform the subsequent development of behavioral interventions.

Older adults with MS (≥60 years of age, N = 180) completed an online survey including demographic and clinical characteristics, SCT variables (exercise self-efficacy, exercise goal setting, social support, and outcome expectations), and PA (total PA [TPA] and moderate-to-vigorous PA [MVPA]).

Bivariate correlation analyses indicated that all SCT variables were significantly associated with TPA and MVPA (all P ≤ .001). Hierarchical linear regression analyses indicated that disability status was a significant correlate of TPA ( β = −0.48, R 2 = 0.23) and MVPA ( β = −0.44, R 2 = 0.19) in step 1; disability and self-efficacy were significant correlates of TPA (disability β = −0.20, self-efficacy β = 0.59, R 2 = 0.50) and MVPA (disability β = −0.16, self-efficacy β = 0.60, R 2 = 0.47) in step 2; and disability, self-efficacy, and exercise goal setting were significant correlates of TPA (disability β = −0.21, self-efficacy β = 0.50, exercise goal setting β = 0.14, R 2 = 0.55) and MVPA (disability β = −0.17, self-efficacy β = 0.51, exercise goal setting β = 0.15, R 2 = 0.51) in step 3.

Conclusions:

These results suggest that behavioral interventions focusing on self-efficacy and exercise goal setting as targets from SCT may be appropriate for increasing PA in older adults with MS.

Of the nearly 1 million adults with multiple sclerosis (MS) in the United States, an estimated 43% are 55 years or older. 1 This finding reflects a graying of the population of adults with MS that coincides with the demographic shift of adults in the general population, and it portends considerable risk of negative effects of MS and aging on walking performance 2 , 3 and physical 3 , 4 and cognitive 2 , 5 function as well as health-related quality of life. 6

To date, there are few options for managing the consequences of aging with MS, 7 but some data support physical activity (PA) behavior as being associated with better walking performance, 2 , 3 , 8 physical function, 2 , 3 and cognitive function 2 in this particular segment of the MS population. Nevertheless, the level of PA is exceedingly low among older adults with MS. 9 Older adults with MS consistently engage in less moderate-to-vigorous PA (MVPA) than do younger and middle-aged adults with MS, 8 , 10 and the MS population as a whole is considerably less physically active than is the general population of adults without MS. 11 This observation supports the importance of identifying theory-based, modifiable correlates of PA behavior for informing the design of behavioral interventions in older adults with MS.

Social cognitive theory (SCT) 12 represents a common approach for identifying modifiable variables as correlates of PA in those with MS. 13 It posits self-efficacy as a proximal determinant of health behaviors and further posits a causal structure among self-efficacy, outcome expectations, perceived impediments/facilitators, and goal setting with health behaviors. 14 Furthermore, SCT recognizes key sources or factors for targeting self-efficacy, including mastery experiences and vicarious experiences, that may translate into health behavior change. There is a wealth of cross-sectional and longitudinal data indicating that SCT variables correlate with PA in young and middle-aged adults with MS. 15 – 20 Importantly, SCT has further informed the design and delivery of behavioral interventions that have increased PA participation in young and middle-aged adults with MS. 21 , 22 This suggests that adopting SCT for examining correlates of PA represents an important first step in the subsequent design and delivery of behavioral interventions for older adults with MS.

The present study examined variables from SCT as correlates of PA in older adults with MS. We first examined SCT variables as correlates of total PA (TPA) and MVPA in the sample of older adults. We then examined the hierarchical associations among SCT variables and TPA and MVPA. We expected bivariate associations among SCT variables and PA outcomes and further expected that self-efficacy would be the primary correlate of PA, followed by other SCT variables, even after controlling for disability status as a confounder.

Participants

This cross-sectional study recruited participants from across the United States via an e-mail distribution from the National Multiple Sclerosis Society. Participants were invited to complete an online survey that focused on correlates of PA behavior. To be included, participants self-reported a diagnosis of MS and age of 60 years or older.

Demographic and Clinical Characteristics

Participants self-reported sex, age, race (selected from a predetermined list), type of MS, and disease duration (years). Disability status was measured using the Patient-Determined Disease Steps (PDDS) scale. 23

Physical Activity

Participants completed the Godin Leisure-Time Exercise Questionnaire (GLTEQ) as a validated self-measure of PA for persons with MS. 24 – 27 Participants recorded the number of bouts (≥15 minutes) of mild (eg, easy walking, yoga), moderate (eg, fast walking, easy bicycling or swimming), and strenuous (eg, jogging, running, vigorous swimming or bicycling) activity completed during a typical week. The number of bouts was multiplied by weights of 3, 5, and 9 for mild, moderate, and strenuous activity, respectively, and then summed into a total leisure-time PA score (GLTEQ-total); higher scores reflect more TPA. The GLTEQ was further scored by summing the weighted values for moderate and strenuous activity only, which provided a health contribution score (GLTEQ-HCS) that corresponds with health-promoting PA (ie, MVPA); higher scores reflect more MVPA.

Self-efficacy

Self-efficacy was measured by the Exercise Self-Efficacy Scale (EXSE). The EXSE contains six items regarding one’s confidence for accumulating 30 minutes or more of MVPA on most days of the week in 1-month increments across the next 6 months. 20 , 28 The items were rated on a scale ranging from 0 (not confident at all) to 100 (highly confident) and were averaged into a total exercise self-efficacy score, with higher scores reflecting greater self-efficacy.

Outcome Expectations

The Multidimensional Outcome Expectations for Exercise Scale (MOEES) assessed beliefs about the benefits of regular participation in exercise and PA. 29 The MOEES contains 19 items across three subdomains of physical, social, and self-evaluative outcome expectations. The items were rated on a 5-point scale ranging from 1 (strongly disagree) to 5 (strongly agree), and item scores were summed, yielding a total score reflecting perceptions of benefits of regular exercise and PA.

Social Support

The abbreviated Social Provisions Scale (SPS) measured social support for PA. 30 , 31 This version of the SPS contains six-items that measure current relationships/support for PA behaviors, and the items were rated on a 4-point scale ranging from 1 (strongly disagree) to 4 (strongly agree). The item scores were summed into a total score, with higher scores reflecting greater perceived support for PA.

Goal Setting

Goal setting for PA was measured using the ten-item Exercise Goal-Setting Scale (EGS). 32 The items reflect goal-setting behaviors for PA and include statements such as “I often set exercise goals,” “I usually set dates for achieving my exercise goals,” and “I usually achieve the exercise goals I set for myself.” Items were rated on a 5-point scale ranging from 1 (does not describe) to 5 (completely describes), and the scores for each item were summed into a total score reflecting goal setting for exercise and PA engagement.

All study procedures were approved by the University of Alabama at Birmingham institutional review board. After receiving the initial e-mail invitation that provided a brief overview of the survey, interested participants were prompted to access the questionnaire online. Participants provided informed consent and confirmed both age and MS diagnosis before accessing the survey. Participants then completed the questionnaires online using survey software (Qualtrics, Provo, UT) within 1 week of starting the research process. The use of internet-based self-report measures increases the accessibility of the survey and, therefore, the generalizability of the data. In addition, all measures included in the survey were validated for use in persons with MS. Participants who were interested in receiving compensation provided an address for receipt of remuneration ($10 USD).

Data Analysis

Data were analyzed using IBM SPSS Statistics for Windows, Version 25.0 (IBM Corp, Armonk, NY), and descriptive statistics are presented as mean ± SD unless otherwise noted (eg, number [%]). We examined the associations between SCT variables (EXSE, EGS, SPS, MOEES) and PA (GLTEQ-total, GLTEQ-HCS) using Spearman rank-order correlations. We further examined the association between disability status (PDDS scale) and PA (GLTEQ-total, GLTEQ-HCS) because previous research has demonstrated a relationship between these two variables 18 and disability might represent a confounder. The magnitude of correlations was interpreted as small, moderate, and large based on values of 0.1, 0.3, and 0.5, respectively. 33 We then conducted hierarchical linear regression analyses whereby we regressed PA (GLTEQ-total and GLTEQ-HCS) on disability status (PDDS scale) in step 1, exercise self-efficacy (EXSE) in step 2, and goal setting (EGS), social support (SPS), and outcome expectations (MOEES) in step 3. We examined the β-coefficients for identifying the independent contributions of the variables in the model as well as model fit based on R 2 and change in R 2 (Δ R 2 ) per step of the model.

Participant Characteristics

The demographic and clinical characteristics of the 180 participants are presented in Table 1 . The mean ± SD age of the sample was 65.2 ± 4.7 years, and the participants were mostly female and White. The sample had a mean ± SD disease duration of 20.8 ± 9.9 years and presented with a relapsing-remitting course of MS with moderate disability (median PDDS scale score = 4.0). This sample is generally consistent with the demographic characteristics of older adults with MS who are primarily female and White race. 1 , 34 Descriptive statistics for the measures of PA and SCT variables are presented in Table 2 .

Demographic and clinical characteristics of the 180 study participants

Note: Values are given as mean ± SD, number (percentage), or median [interquartile range].

Abbreviations: PDDS, Patient-Determined Disease Steps; PPMS, primary progressive multiple sclerosis; RRMS, relapsing-remitting multiple sclerosis; SPMS, secondary progressive multiple sclerosis.

Mean scores of PA and SCT variables

Abbreviations: EGS, Exercise Goal-Setting Scale; EXSE, Exercise Self-Efficacy Scale; GLTEQ-HCS, Godin Leisure-Time Exercise Questionnaire health contribution score; GLTEQ-t, Godin Leisure-Time Exercise Questionnaire total leisure-time activity score; MOEES, Multidimensional Outcomes Expectations for Exercise Scale; PA, physical activity; SCT, social cognitive theory; SPS, Social Provisions Scale.

Associations Between PA and SCT Variables

Results from the Spearman rank-order correlation analyses are presented in Table 3 . Disability status and all SCT variables were significantly associated with TPA (GLTEQ-total) and MVPA (GLTEQ-HCS) (all P ≤ .001). The associations between exercise self-efficacy (EXSE) and disability status with PA were large in magnitude, whereas all other associations were between small and moderate in magnitude.

Correlations between PA and SCT variables a

Abbreviations: EGS, Exercise Goal-Setting Scale; EXSE, Exercise Self-Efficacy Scale; GLTEQ-HCS, Godin Leisure-Time Exercise Questionnaire health contribution score; GLTEQ-t, Godin Leisure-Time Exercise Questionnaire total leisure-time activity score; MOEES, Multidimensional Outcomes Expectations for Exercise Scale; PA, physical activity; PDDS, Patient-Determined Disease Steps; SCT, Social Cognitive Theory; SPS, Social Provisions Scale.

Regressions by Outcome Measure

Total physical activity (glteq-total).

Results of the regression are presented in Table S1 , which is published in the online version of this article at ijmsc.org . Disability status (PDDS scale) was included in step 1 of the hierarchical linear regression as a possible confounding variable 18 and accounted for 23% of the variance in TPA ( R 2 = 0.23, β = −0.48). Exercise self-efficacy was included in step 2 and explained an additional 28% of the variance in TPA ( R 2 = 0.50, Δ R 2 = 0.28, β = 0.59). The remaining SCT variables (EGS, SPS, and MOEES) were included in step 3. This model explained an additional 4% of the variance ( R 2 = 0.55, Δ R 2 = 0.04), and disability status (β = −0.21), exercise self-efficacy (β = 0.50), and exercise goal setting (β = 0.14) were significant correlates of TPA scores.

Moderate-to-Vigorous Physical Activity (GLTEQ-HCS)

Results of the regression are presented in Table S2 . Variables were entered into the regression in the same manner as GLTEQ-total. Disability status initially explained 19% of the variance in MVPA ( R 2 = 0.19, β = −0.44). Exercise self-efficacy explained an additional 28% of the variance in MVPA ( R 2 = 0.47, Δ R 2 = 0.28, β = 0.60), and including all SCT variables explained a total of 51% of the variance in MVPA ( R 2 = 0.51, Δ R 2 = 0.04). The final model included disability status (β = −0.17), exercise self-efficacy (β = 0.51), and exercise goal setting (β = 0.15) as significant correlates of MVPA scores.

Older adults with MS are an expanding segment of the MS population, 1 and these individuals experience a fast rate of disability progression 35 that manifests as problems with walking 2 , 3 , 8 and reduced physical 3 , 4 and cognitive 2 , 5 function. There are few options available for managing the consequences of aging with MS. There is evidence of an association between PA and improved symptoms in this particular population, 2 , 3 , 8 yet levels of PA are exceedingly low in older adults with MS. 9 We conducted bivariate correlation and linear regression analyses based on SCT for examining modifiable correlates of PA in older adults with MS. The analyses indicated that self-efficacy was significantly and independently associated with PA and consistently explained the largest portion of variance in PA outcomes, even after accounting for disability status and other SCT variables. The only other SCT variable explaining variance in PA was goal setting. These results support continued research on self-efficacy, as well as goal setting, for informing behavioral interventions that target PA and its secondary benefits in older adults with MS.

The results of the present study align with previous research on associations between self-efficacy and PA participation in comparatively younger adults with MS. 15 – 17 , 20 Importantly, self-efficacy is a modifiable construct, and SCT identifies specific sources or influences of self-efficacy that can be targeted with behavioral interventions. 12 The most effective source of self-efficacy is mastery experiences. By experiencing success, an individual’s confidence in their ability to make a change and adopt the new health behavior is increased. Vicarious experiences, or seeing someone similar to you succeed, is another strategy for increasing self-efficacy via social support. Another major source of self-efficacy is social persuasion and verbal encouragement, which suggests that developing interventions that provide strong social support are a powerful precursor and driver of behavior change. Last, a person’s emotional state or affective experiences can influence self-efficacy. To positively affect self-efficacy, it is important to reduce stress and minimize negative emotional responses. 12 Specific behavior change techniques that target these sources of self-efficacy include action planning that involves explicit instructions of how to achieve goals, providing positive feedback, and focusing on small successes. 36 Future research should consider focusing on the sources of self-efficacy using theory-based approaches to inform the design of behavioral interventions for increasing PA in older adults with MS. 13

According to the causal model proposed in SCT, self-efficacy may be associated with PA through goal setting; however, goal setting itself influences health behavior independent of self-efficacy. 14 This is supported by the present results that demonstrated a relationship between goal setting and PA (TPA and MVPA) independent of disability status, self-efficacy, and other SCT variables. Together with previous research in MS that has demonstrated an independent association between goal setting and PA, 17 , 19 , 21 , 37 this finding suggests that goal setting may be another target of behavioral interventions aimed at increasing PA in older adults with MS. Goals set the course for behavior change and provide incentives and motivation. 14 Important characteristics of successful goal-setting behavior include setting goals that are clearly defined, attainable, measurable, and time limited. 38 Behavioral interventions that incorporate goal-setting practices may have a positive effect on PA behavior in older adults with MS.

The present study is not without important limitations. The cross-sectional research design precludes any inferences of causality. Another limitation is the homogeneity of the primarily female and White sample; this may limit the generalizability of the results to males and other races with MS. Future research may benefit from including a more diverse sample. Also note that all data were based on self-report measures and, therefore, are inherently associated with certain limitations compared with objective measures. In addition, the present study considered only variables from SCT as possible correlates of PA. Future studies may consider examining determinants of PA proposed by alternative behavior change models (eg, the transtheoretical model 39 or the health action process approach 40 ).

Overall, the primary result of this cross-sectional study indicated that self-efficacy was significantly and independently associated with both total leisure-time PA and health-promoting PA (ie, MVPA) in older adults with MS. Exercise goal setting was further identified as an independent correlate of PA. These SCT variables are modifiable through focal intervention and may represent targets of behavioral change for increasing PA in older adults with MS. Future research may design and test behavioral interventions based on SCT for increasing PA in the expanding segment of older adults living with MS.

PRACTICE POINTS

  • The expanding population of older adults with MS likely experiences many of the same benefits of physical activity (PA) as younger and middle-aged adults with MS. However, participation in PA is exceedingly low in this particular segment of the MS population.
  • The present results suggest that behavioral interventions focusing on self-efficacy and exercise goal setting as targets from social cognitive theory may be appropriate for increasing PA in older adults with MS.

Supplementary Material

Financial Disclosures: The authors declare no conflicts of interest.

Funding/Support: This research was supported by the National Multiple Sclerosis Society (grant MB 0029). Dr Baird is supported by the National Institutes of Health (training grant 2T32HD071866-06).

IMAGES

  1. Social Cognitive Theories Free Essay Example

    social cognitive theory research paper

  2. Psycho Educational and Social-Ecological models of Social Cognitive

    social cognitive theory research paper

  3. Sale > history of social cognitive theory > in stock

    social cognitive theory research paper

  4. (PDF) Social Cognitive Theory

    social cognitive theory research paper

  5. (PDF) Contributions of Social Cognitive Theory in the context of

    social cognitive theory research paper

  6. SCT

    social cognitive theory research paper

VIDEO

  1. Social Psychology: Chapter 2 (Methodology) Part 2

  2. Social Cognitive Theory

  3. Social Cognitive Theory

  4. What is Social Cognitive Theory?

  5. AEN2233: Social Cognitive Theory: New media and Changes in news consumption- FAKE NEWS

  6. Social Learning Theory , Cognitive theory

COMMENTS

  1. Social cognitive theory.

    Abstract The present chapter traces the evaluation of social cognitive theory toward a psychology of human agency. The nature of human agency and the mechanisms through which it operates is analyzed in the context of major changes in the conception of human behavior.

  2. Frontiers

    Social cognitive theory (SCT) represents a causal model in which the self-efficacy is set to the influence human behavior directly and indirectly via other mediating processes that include outcome expectations, social support, and self-regulation (Bandura, 1991, 1998, 2004; Bandura et al., 1997). Self-efficacy reflects the judgment of own ...

  3. The perspectives of social cognitive career theory approach in

    Social cognitive career theory Derived from Albert Bandura's Self-Efficacy Theory and General Social Cognitive Theory (Lent, 2013 ), SCCT develops into a comprehensive career theory that argues that an individual's career path results from the interaction between multiple career elements since it was proposed by Lent et al. ( 1994 ).

  4. Social Cognitive Theory and Physical Activity in Older Adults

    The present study examined variables from SCT as correlates of PA in older adults with MS. We first examined SCT variables as correlates of total PA (TPA) and MVPA in the sample of older adults. We then examined the hierarchical associations among SCT variables and TPA and MVPA.

  5. Applications and applicability of Social Cognitive Theory in

    Abstract. Social Cognitive Theory is a theory that derives from Psychology yet has been adopted in other interdisciplinary subject areas, including information science. The origins and key concepts of Social Cognitive Theory are presented, and the value of its contributions to information science research analysed, with particular reference to ...