A Test of the Theory of Planned Behavior to Predict Physical Activity in an Overweight/Obese Population Sample of Adolescents From Alberta,Canada

Abstract

Purpose. To examine the utility of the theory of planned behavior (TPB) for explaining physical activity (PA) intention and behavior among a large population sample of overweight and obese adolescents (Alberta, Canada), using a web-based survey. Secondary objectives were to examine the mediating effects of the TPB constructs and moderating effects of weight status. Methods. A subsample of 427 overweight and 133 obese participants (n = 560), completed a self-administered web-based questionnaire on health and PA behaviors, including assessment of attitude, subjective norm, perceived behavioral control (PBC), and intention to participate in regular PA. Structural equation models were examined using AMOS 17.0. Results. Overall, 62% of the variance in intention was accounted for by attitude, subjective norm, and PBC; whereas 44% of the variance in PA behavior was explained by PBC and intention. When examining the TPB separately in overweight and obese subsamples, 66% and 56% of the variance for PA intention was explained for overweight and obese subsamples, respectively; and 38% and 56% of the variance in PA behavior were accounted for in the overweight and obese subsamples. Overall, attitude was the strongest predictor of PA intention, whereas PBC was the strongest predictor for PA behavior. Intention was not predictive of PA behavior. Conclusions. These results provide partial support for the utility of TPB in explaining PA behavior in a sample of overweight and obese adolescents. In particular, strong associations regarding attitude and PBC were evident across each subsample. These findings have implications for tailoring PA programs in this population.

Keywords

adolescents obese overweight physical activity theory of planned behavior web-based survey

Obesity has reached epidemic proportions (Biro & Wein, 2010; Lutz, Karloy, & Okun, 2008) in many developed nations worldwide, presenting significant liabilities to public health (Janssen, Lam, & Katzmarzyk, 2009; Wang & Lobstein, 2006). An estimated 20% of Canadian youth are classified as overweight or obese (Janssen et al., 2009; McCrindle et al., 2010). Specifically, youth obesity prevalence continues to increase, McCrindle et al. (2010) report that 13% of Canadian youth are classified as obese. Similarly, in the United States and Australia, obesity stands as the second leading source of preventable morbidity and mortality (Hamilton & White, 2008; Wang & Lobstein, 2006). Obesity has a significant burden to the health care system, with substantial direct and indirect costs (Janssen et al., 2009; Wang & Lobstein, 2006). Physical inactivity and low fitness levels are strongly associated with increased weight in children and adolescents (Burke, Beilin, & Dunbar, 2001). Of concern, many young people continue to lead an inactive lifestyle, failing to meet recommended physical activity (PA) guidelines of at least 60 minutes of moderate to vigorous activity daily (Department of Health & Aging, 2004). For example, a recent Canadian national survey has identified a considerable decline in fitness measures (e.g., aerobic fitness) in children and youth from 1981 to 2009 (Tremblay et al., 2010).

Sharp increases in the prevalence of obesity in childhood and adolescence are particularly alarming (Ogden et al., 2006), considering it is a significant predictor of obesity in adulthood and is linked to increased prevalence of morbidity and mortality, as well as numerous lifestyle diseases (Biro & Wein, 2010; Gill et al., 2009; Rey-Lopez, Vicente-Rodriguez, Biosca, & Moreno, 2008; Wang & Lobstein, 2006). Research suggests that adolescents with an elevated body mass index (BMI) experienced 30% higher rates of mortality than do young and middle-aged adults (Biro & Wein, 2010).

Numerous adverse health consequences of childhood and adolescent obesity exist (Rey-Lopez et al., 2008); physical outcomes of obesity are associated with type 2 diabetes, obesity in adulthood, cardiovascular disease and several cancers (Biro & Wein, 2010; Gill et al., 2009). Social, emotional, and physical health is generally poorer in obese adolescents in comparison with healthy weight adolescents (Wake et al., 2009).

Physical inactivity and sedentary behavior throughout childhood are independent contributors of obesity during adolescence (Rey-Lopez et al., 2008; Tremblay & Willms, 2003). It is the combination of nutrition and PA-related behaviors that affect energy balance and, therefore, weight status (Kremers et al., 2006). As such, engaging in regular PA provides protection against overweight or obesity (Rey-Lopez et al., 2008; Tremblay & Willms, 2003).

Theories are helpful for explaining and predicting health behaviors such as PA, and theory-based interventions are more efficacious than atheoretical approaches (Michie & Abraham, 2004). However, prior to the translation of theories into interventions, theories and their assumptions should be tested empirically to ensure their adequacy. The theory of planned behavior (TPB; Ajzen, 1991) is one such social-cognitive theory that has been widely examined to predict behavior in numerous populations (Conner & Sparks, 2005). However, the core tenets of the TPB have yet to be tested for PA in an overweight and obese adolescent population sample, which is the purpose of this article.

Briefly, the TPB proposes that a person’s intention to perform a behavior is the central determinant of that behavior because it reflects the level of motivation a person is willing to exert to perform the behavior (Ajzen, 1991). Intention is hypothesized to be determined by attitude, subjective norm, and perceived behavioral control (PBC). Attitude is reflected in a positive or negative evaluation of performing the behavior. Subjective norm is defined as the perceived social pressure to perform the behavior whereas PBC is defined as the perceived ease or difficulty of performing the behavior. PBC is also hypothesized to directly predict behavior. Application of the TPB identifies underlying beliefs that determine one’s attitude, subjective norm, and PBC (Ajzen, 1991) and can provide an applied basis for understanding the factors that help initiate behaviors (such as PA; Dishman, 1994; Glanz, Rimer, & Lewis, 2002). TPB has been used extensively to study a broad range of behaviors (Armitage & Conner, 2001; Godin & Kok, 1996) with a particular focus on health behaviors (Godin & Kok, 1996), including PA (Dzewaltowski, Noble, & Shaw, 1990; Hagger, Chatzisarantis, & Biddle, 2002). For example, in a review of health behaviors by Armitage and Conner (2001), attitude, subjective norm, and PBC accounted for 40% of the variance in intention, whereas intention and PBC accounted for 27% of the variance in PA behavior.

The vast majority of studies examining the utility of the TPB in explaining PA behavior have focused on adult populations. Limited research, however, has been conducted with adolescents. Twelve nonintervention studies have examined the TPB to explain PA in adolescents (Araújo-Soares, McIntyre, & Sniehotta, 2009; Baker, Little, & Brownell, 2003; Craig, Goldberg, & Dietz, 1996; de Bruijn et al., 2006; Godin, Anderson, Lambert, & Desharnais, 2005; Hagger, Chatzisarantis, Biddle, & Orbell, 2001; Hamilton & White, 2008; Martin, Oliver, & McCaughtry, 2007; Mummery, Spence, & Hudec, 2000; Saunders, Motl, Dowda, Dishman, & Pate, 2004; Sas-Nowosielski, 2006; Trost, Saunders, & Ward, 2002); however, none of these are population-based studies, which limits our current knowledge regarding the generalizability of the TPB’s power in predicting PA in the general adolescent population. Furthermore, it appears no TPB study in an adolescent or adult population has examined the moderating effect of weight status.

The primary purpose of our study was to examine the utility of the TPB in explaining PA intention and behavior among a large population sample of overweight/obese adolescents from the province of Alberta, Canada. Secondary objectives include examining (a) mediating effects of the TPB constructs and (b) the moderating effect of weight status.

Method

The main study is reported elsewhere (Plotnikoff et al., 2011). However, this article examines the TPB by weight classification on a subset of the study population. The Web-Survey of Physical Activity and Nutrition (Web-SPAN; Storey et al., 2009) is a self-administered web-based survey of PA, diet, and smoking behaviors of adolescents. Briefly, all school boards in the province of Alberta, Canada were invited to participate. A total of 109 schools across 37 school boards agreed to participate (school board response rate of 64%, school response rate of 40%). Overall, 4,073 adolescents (out of a potential 9,071 students), consented and completed Web-SPAN during school time (111 individuals did not provide complete data and were eliminated from study analyses) yielding a student response rate of 44.8%. The current study examines a subsample of 560 overweight and obese participants.

Ethics approval for this study was obtained from the Human Research Ethics Board in the Faculty of Agricultural, Life and Environmental Sciences at the University of Alberta and by the Cooperative Activities Program in the Faculty of Education at the University of Alberta.

Measures

Theory of planned behavior

To minimize response burden, short measures of core TPB constructs (Hagger et al., 2001; Rhodes, Macdonald, & McKay, 2006; Spence et al., 2010) were used in the survey protocol, which included attitude, subjective norm, PBC, and intention. These published measures were developed in English-speaking countries (e.g., United Kingdom, Canada) with similar age groups. The measures were framed as achieving regular PA, defined as accumulating at least 30 minutes of PA on most days of the week (Public Health Agency of Canada, 2002), over the next month (Plotnikoff et al., 2011).¹

Attitude

A two-item measure (with 5-point semantic differential scales for the response options) was used to assess attitude toward PA. The scale which assessed enjoyment and importance (adapted from Rhodes, Macdonald, & McKay, 2006) included the following items: (a)“Doing regular physical activity over the next month would be”: 1 = very boring to 5 = very fun and (b) “Doing regular physical activity over the next month would be”: 1 = very unimportant to 5 = very important. The path coefficients in our study for the two attitude items were .88 and .87 and Cronbach’s α was 0.77 (α = .70 for overweight and α = .87 for obese).

Subjective norm

A single-item measure to assess subjective norm (Hagger et al., 2001) was used, which included “Most people important to me think I should take part in regular physical activity in the next month.” Response options ranged from 1 = very unlikely to 5 = very likely.

Perceived behavioral control

As a proxy to PBC, a four-item validated self-efficacy measure (Spence et al., 2010) was used, which included confidence in participating in PA over the next month when (a) tired, (b) having a lot of homework to complete, (c) approaching an adult for transportation, and (d) participating in PA on most days. Response options ranged from 1 = not at all confident to 5 = extremely confident. The path coefficients for the four PBC items in our study ranged from .71 to .87 and the scale alpha for the total sample was .84 (α = .82 for overweight and α = .87 for obese).

Intention

To measure intention, a single-item measure slightly modified from Hagger et al. (2001) was assessed, which included “I plan to be physically active on a regular basis over the next month.” Response options ranged from 1 = strongly disagree to 5 = strongly agree.

Physical activity

The Physical Activity Questionnaire for Older Children (PAQ-C; Kowalski, Crocker, & Faulkner, 1997) was used to assess PA levels of participants during the previous 7 days. The PAQ-C was specifically designed for Canadian school-aged youth, including components on PA during the school day, after school, and weekend PA. The questionnaire consists of nine items used to calculate an activity score. Responses are converted to a 5-point scale where a higher score represents greater PA levels (scores range from 1.00 to 5.00). The PAQ-C has been shown to be a reliable self-report measure of PA when test–retest reliability was examined over a period of 1 week among 84 boys (r = .75) and girls (r = .82) aged 9 to 14 years (Crocker, Bailey, Faulkner, Kowalski, & McGrath, 1997). According to Kowalski et al. (1997), the PAQ-C was a valid measure among 9- to 14-year-olds (N = 97) when compared with other PA assessment methods including an activity rating (r = .57), the Leisure Time Exercise Questionnaire (r = .41), a 7-day PA recall interview (r = .46), a Caltrac motion sensor (r = .39), and the Canadian Home Fitness Test (step test; r = .28).

In an earlier study of 459 students of this age group, no significant difference was observed between scores on the web-based PAQ-C when compared with scores on a guided self-administered paper-based PAQ-C, t(411) = −0.73, p = .46, d = −0.03 (Spence et al., 2010). Similar to Crocker t al. (1997), good test–retest reliability was found for the web-based PAQ-C completed approximately 8 days apart (r = .79).

Statistical Analysis

PASW Statistics 17 (SPSS Inc. Chicago, IL) was used to calculate means, standard deviations, bivariate correlations, and regression coefficients. Structural equation models (SEMs) were examined using AMOS 17.0 (Small Waters Corp., Chicago IL). To determine if there were gender and age differences between weight status groups (i.e., overweight and obese), chi-square and student t tests were conducted. Means of completed items were imputed for items missing in scales (<5% of items missing). Group mean substitution was used where single-item measures were left uncompleted and analyses were conducted with and without missing data imputation. As the model-fit indices did not change once the missing data were imputed, analyses were conducted using this method.

Model testing

To correct for the clustering of effects at the school level, all variables were adjusted for school using multiple linear regression, and the unstandardized residuals were used in the analyses (Cohen & Cohen, 1983; Dishman et al., 2006). The proposed model was tested using maximum likelihood analysis in AMOS and the PAQ-C score was used as the dependent variable. Model fit was assessed using multiple indices, including chi-square index, goodness-of-fit index (GFI), adjusted goodness-of-fit index (AGFI), comparative fit index (CFI), and root mean square error of approximation (RMSEA). The chi-square tests the null hypothesis that the model is a good fit of the data. Although a nonsignificant chi-square result (p > .05) indicates that the model is a good fit, it is oversensitive in large samples (Bollen, 1989) and as a result additional measures are often used. The GFI provides an estimate of the proportion of variance in the variance–covariance matrix accounted for by the proposed model. The AGFI provides a GFI score adjusted for the number of parameters in the model. The CFI compares the existing model fit with a null model which assumes the indicator variables in the model are uncorrelated. GFI, AGFI, and CFI scores range from 0 to 1, with a score exceeding .9 indicating a good fit (Bollen, 1989; Jöreskog & Sörbom, 1993). The RMSEA estimates closeness-of-fit compared with the saturated model and a value of <.10 is considered to be acceptable (Keith, 1993). Hu and Bentler (1999) have proposed more conservative model-fit values (CFI and Tucker–Lewis index >.95; RMSEA < .06), but others have argued against using these higher cutoff criteria (Marsh, Hau, & Wen, 2004).

Although the univariate distribution of the data was generally normal, the multivariate kurtosis value was above the recommended value of 3. Consequently, bootstrapping procedures were used and the Bollen–Stine p value and bias-corrected regression coefficients are reported throughout (Bollen & Stine, 1992).

Mediation analyses

In the TPB, intention is thought to mediate the relationship between attitude, subjective norm, PBC, and the behavior itself. The indirect effects of intention were examined using single-mediator models and asymmetric confidence intervals tested the significance of the indirect effects (MacKinnon, Lockwood, & Williams, 2004). The unstandardized regression coefficients from PASW were used in the analyses for single models. In the first step, the pathways from subjective norm, attitude, and PBC to intention were identified (A). In the second step, the pathway from intention to PA was explored (B). The indirect (mediated) effect was then calculated by multiplying A and B. Asymmetric confidence intervals were used to test the significance of the product of coefficients (AB) using Mackinnon, Fritz, Williams, and Lockwood’s (2007) PRODCLIN (PRODuct Confidence Limits for INdirect effects) program. If zero is outside the confidence interval the mediated effect is statistically significant (MacKinnon, 2008).

Moderation analyses

Weight status (overweight or obese) was identified as a potential moderator of the TPB model. Multigroup moderation analyses were conducted using a series of models, starting from an unrestricted model to the fully constrained model (Nigg, Lippke, & Maddock, 2009). In Model 1, the relationships between variables are unconstrained and this model serves as the basis for comparison. Model 2 is the measurement equivalent model, which includes equal factor loading across subsamples (loadings from the latent factors to measured items restrained to be equal across all groups). Model 3 includes the Model 2 constraints plus equal factor covariance (additional constraints that the interrelationships of attitude, subjective norm, and PBC would be equal across groups). Model 4 includes the Model 3 constraints plus equal paths (additional constraints that the interrelations of subjective norm to intention, attitude to intention, PBC to intention, PBC to behavior, and intention to behavior would be equal across groups). To compare the models, change in CFI was used (Cheung & Rensvold, 2002). Cheung and Rensvold (2002) proposed critical values to indicate measurement invariance, which are robust for testing multigroup invariance. A ΔCFI ≤ −.01 indicates that the null hypothesis of invariance should not be rejected.

Results

Overview

The sample reflects the age and sex distribution for Alberta youth. The mean age of our sample was 13.7 years (1.5) versus 13.3 years (1.3) in the total population of Alberta of Grades 7 to 10 students (N ~ 130,000). The sample (N = 560) included 308 boys and 252 girls. The proportion of boys in our sample was 55% versus 45% (Grades 7-10) for the province. Participants provided self-reported height and weight.

Based on the International Obesity Task Force classifications (obesity defined as BMI >95th percentile, overweight as BMI >85th percentile and <95th percentile, and normal weight as BMI <85th percentile), which is based on data from six countries, 427 participants were categorized as overweight and 133 as obese (Cole, Bellizzi, & Flegal, 2000). In another population sample (N = 4,932) of Alberta youth conducted 3 years earlier (Plotnikoff et al., 2009), 7.7% of boys and 4.5% of girls were obese according to International Obesity Task Force cutoffs; the corresponding proportions of obese youth in this study were 7.6% for boys and 6% for girls.

There were no significant age or gender differences between overweight and obese participants. The descriptive statistics and bivariate correlations among study variables are reported in Table 1. Significant intercorrelations (p < .01) were observed among TPB constructs and all the constructs were associated with PA (p < .01).

Table 1.

Bivariate Correlations Among Study Variables.

	Descriptives^a				Correlations^b
Variable	N	M	SD	1	2	3	4	5
1. PAQ-C score	560	2.83	.73	—
2. Attitude	560	4.09	.90	.40**	—
3. Subjective norm	560	3.76	1.13	.30**	.41**	—
4. Perceived behavioral control	560	3.27	.91	.60**	.51**	.30**	—
5. Intention	560	4.16	.96	.42**	.69**	.37**	.52**	—

Note. M = mean; SD = standard deviation; PAQ-C score = Physical Activity Questionnaire for Older Children score.

Raw score values reported here. ^bUnstandardized residual scores used in bivariate correlations to adjust for clustering at the school level.

p < .01.

Model Testing

In the model including both overweight and obese adolescents (see Figure 1) the chi-square statistic was significant (χ² = 90.43, df = 22, p < .001), but the model represented an adequate-to-good fit to the data based on the other indices (GFI = .96, AGFI = .93, CFI = .97, and RMSEA = .08; Table 2). Standardized coefficients are reported in Figure 1. The model explained 44% of the variance in PA, but only two of the TPB pathways were statistically significant. The pathways from attitude to intention (r = .70, p < .01) and from PBC to PA (r = .63, p < .01) were both statistically significant.

Figure 1.

Standardized parameter estimates for pathways among constructs from the theory of planned behavior among overweight and obese adolescents.

Table 2.

Goodness-of-Fit Indices for the Subsamples and Multisample Nested Models.

Models	N	χ²	df	p Value ^a	GFI	AGFI	RMSEA	CFI	ΔCFI
Full model	560	90.43	22	<.001	.96	.93	.08	.97	—
Weight status
Overweight^b	427	94.85	22	<.001	.95	.90	.09	.95	—
Obese^b	133	20.99	22	.63	.97	.93	.00	1.00	—
Model 1	560	115.80	44	<.001	.96	.91	.05	.97	—
Model 2	560	124.55	51	<.001	.95	.91	.05	.97	<.01
Model 3	560	146.64	56	<.001	.84	.91	.05	.96	.01
Model 4	560	320.45	65	<.001	.89	.84	.08	.88	.08

Note. GFI = goodness-of-fit index; AGFI = adjusted goodness-of-fit index; CFI = comparative fit index; RMSEA = root mean square error of approximation. Model 1—unrestricted model; Model 2—measurement equivalent model which includes equal factor loading across subsamples; Model 3—Model 2 constraints plus equal covariances between perceived behavioral control, subjective norm, and attitude; Model 4—Model 3 constraints plus equal paths between constructs. ΔCFI = change in comparative fit index. A ΔCFI > .01 indicates that the null hypothesis of invariance should be rejected.

Bollen–Stine p value reported because of multivariate nonnormality. ^bBased on self-reported heights and weights and the International Obesity Task Force classification.

Mediation Analyses

The indirect effects of attitude, subjective norm, and PBC are reported in Table 3. Intention to be active was found to mediate the relationship between these variables and PA in the single-mediator models. The largest mediated effect was observed for attitude, suggesting that intention to be active mediates the cross-sectional relationship between attitude to be active and behavior (AB = .15; 95% CI = .10 to .21).

Table 3.

Path Coefficients, Significance Levels, and Significance of the Indirect Effects of Attitude, Subjective Norm, and Perceived Behavioral Control.

	Unstandardized Regression Coefficients and Confidence Intervals				Significance of Indirect Effect
Variables	A (SE)	95% CI	B (SE)	95% CI	AB	95% CI^a
Attitude	.73 (.03)***	[.66, .79]	.21 (.04)***	[.13, .29]	.15	[.10, .21]
Subjective norm	.32 (.03) ***	[.25, .38]	.27 (.03)***	[.21, .33]	.09	p.06, .11]
Perceived behavioral control	.55 (.04)***	[.47, .62]	.11 (.03)***	[.05, .17]	.06	[.03, .09]

Note. A = estimate of unstandardized regression coefficient of attitude, subjective norm, and perceived behavioral control predicting intention; B = estimate of the unstandardized regression coefficient of intention predicting physical activity with attitude, subjective norm, or perceived behavioral control in the model; SE = standard error; 95% CI = 95% confidence interval; AB = product of coefficients estimate.

95% asymmetric CIs of the mediated effect calculated using the PRODCLIN program.

p < .01. ***p < .001.

Moderation Analyses

The model was tested separately for overweight and obese adolescents (Table 2). Among overweight adolescents (see Figure 2), the model was found to be an adequate to good fit to the data (χ² = 94.85, df = 22, p < .001, GFI = .95, AGFI = .90, CFI = .95, and RMSEA = .09). Among adolescents classified as obese (see Figure 3), the model was an excellent fit to the data (χ² = 20.99, df = 22, p < .63, GFI = .97, AGFI = .93, CFI = 1.00, and RMSEA = .00). Table 2 presents the fit indices of the constrained models and the differences in CFI of these models with the unconstrained model. There was little difference between Models 1 and 2, which suggested that there was little difference in the way that obese and overweight participants interpreted the constructs. There was a small but significant difference between Model 3 and Model 2 (ΔCFI = .01) indicating differences in the relationships between subjective norm, PBC, and attitude among overweight and obese participants. There was a large difference between Models 3 and 4 (ΔCFI = .08), indicating large differences in the path coefficients for overweight/obese adolescents.

Figure 2.

Standardized parameter estimates for pathways among constructs from the theory of planned behavior among overweight adolescents.

Figure 3.

Standardized parameter estimates for pathways among constructs from the theory of planned behavior in obese adolescents.

Discussion

This study investigated the utility of the TPB in explaining PA intention and behavior among a large population sample of overweight/obese Canadian adolescents. Secondary study objectives were to (a) examine the mediating effects of the TPB constructs and (b) assess the moderating effects of weight status. Overall, our results partially support the TPB as an appropriate theoretical model for explaining the determinants of PA behaviors in a sample of overweight/obese adolescents. Among overweight adolescents the model was found to be an adequate-to-good fit to the data; for obese adolescents the model was an excellent fit to the data.

The model accounted for 66% and 56% of the variance in intention, and 38% and 56% of the variance in PA behavior in the overweight and obese subsamples, respectively. The reported variances are higher than those described in most TPB–PA studies in an adolescent population, which explained 6% to 37% of the variance in PA and 13% to 58% of the variance in intention (Araújo-Soares et al., 2009; Baker et al., 2003; Craig et al., 1996; de Bruijn et al., 2006; Godin et al., 2005; Hagger et al., 2001; Hamilton & White, 2008; Martin et al., 2007; Mummery et al., 2000; Saunders et al., 2004; Sas-Nowosielski, 2006; Trost et al., 2002). The higher variances explained in the current study may be attributed to the adolescent population under investigation.

The pathways from attitude to intention, and from PBC to PA, were statistically significant in both subsamples. In terms of attitude, Godin and colleagues describe similar results in an adolescent sample (Godin & Kok, 1996), with the relationship between attitude and intention positively associated across each subsample. Similar findings have been previously reported by many studies (Craig et al., 1996; Hamilton & White, 2008; Mummery et al., 2000; Sas-Nowosielski, 2006; Trost et al., 2002). Therefore, PA interventions targeting overweight and obese adolescents should provide information to enhance beliefs in favorable outcomes (e.g., fun, enjoyable, improve health) associated with participation in regular PA.

PBC was the strongest correlate of PA behavior across the two subsamples highlighting the importance of developing cognitive and behavioral strategies (e.g., scheduling activities) to overcome PA barriers (e.g., when tired, having homework), and enhance overweight/obese adolescent’s confidence in adopting and maintaining participation in regular PA.

Across each subsample, a moderate association between PBC and intention was observed. Interestingly, a stronger relationship between PBC and PA behavior (than between PBC and intention) was found across each of the two subsamples which is also consistent with the results of the overall sample (N = 4,073) from this study (Plotnikoff et al., 2011). Furthermore, the magnitude of the association between PBC and behavior was greater for the obese group versus the overweight group. However, these observations are inconsistent with previous studies on youth (Araújo-Soares et al., 2009; Baker et al., 2003; Craig et al., 1996; de Bruijn et al., 2006; Godin et al., 2005; Hagger et al., 2001; Hamilton & White, 2008; Martin et al., 2007; Mummery et al., 2000; Sas-Nowosielski, 2006; Saunders et al., 2004; Trost et al., 2002), which report that PBC is more strongly associated with intention than with behavior. Possible reasons for the differences in these findings include the uniqueness of our sample (i.e., population-based sample) and the unique target population (i.e., overweight/obese) youth. The ceiling effects and limited variability for the intention measure may have also accounted for this finding; 82% of the sample answered 3 or higher on the 5-point Likert-type scale for this measure, whereas there was greater variability for the PBC measure.

PA behavior was more strongly associated with PBC (r = .60, p < .01) than intention (r = .42, p < .01) in the study sample. The mean intention score was high and it is plausible to suggest that overweight and obese youth have good intentions to be physically active but only those with high levels of confidence are able to translate their good intentions into action. Subjective norm presented a very limited effect for each subsample, suggesting a lack of perceived pressure from peers to be active, possibly as a result of inactive peers or intimidation from active peers. The association between social norm and PA behavior is low to moderate (standardized beta coefficients ranging from .02 to .29) in the existing TPB studies (de Bruijn et al., 2006; Hamilton & White, 2008; Everson, Daley, & Ussher, 2007; Trost et al., 2002).

The relationship between intention and behavior was limited for the overweight subsample (β = .09). Furthermore, a negative association (β = −.04) between the two constructs was reported for the obese subsample. Although these pathways were not statistically significant, this interesting finding may indicate that obese adolescents have lower self-esteem, poor psychomotor skills and/or are physically limited when attempting to participate in PA because of excess weight; therefore, intention is not translated into behavior. Indeed, the obese subsample was less active (based on our PA behavior measure) than the overweight (p < .01) and normal weight (p < .001) subgroups in the total study sample.

Less time spent in PA for overweight and obese adolescents, compared with healthy weight adolescents, has been previously reported (Brownson et al., 2000; Martinez-Gonzalez, Martinez, Hu, Gibney, & Kearney, 1999). K. Ball, Crawford, and Owen (2000) suggest overweight and obesity to be related to shyness, embarrassment, and lack of motivation for engaging in PA. Furthermore, Blanchard et al. (2005) reported reduced access to PA facilities for obese individuals in contrast to individuals of healthy weight.

Intention was found to mediate the relationship between the TPB variables and PA in the single-mediator models. The largest mediated effect was observed for attitude, suggesting that intention to be active is an important mediator for the relationship between attitude to be active and PA behavior. Intention was a significant mediator in the single-mediator models, however in the full-model test, intention had a limited association with behavior for both subsamples. This may be explained in part by the fact that PBC accounted for a greater proportion of the variance for behavior, along with the strong correlation between attitude and PBC, mitigating the intention–behavior relationship in the full model.

Strengths and Limitations

There are limitations that should be acknowledged when interpreting the study results. First, the cross-sectional design prevents the assessment of temporal causality. Second, although a validated measure of PA was used, objective measures (e.g., accelerometry and pedometry) were not included. Furthermore, height and weight assessments were self-reported. However, in a sample of 459 Alberta adolescents, intraclass correlation coefficients between self-reported and measured height was .88; whereas self-reported and measured weight was .93 (Calengor, 2007). To limit response burden, the brief TPB instrument did not include the expectancy × value components for the TPB constructs as specified in the full TPB (Ajzen, 1991), and brief measures of the TPB may have potentially limited the content validity of the constructs. Finally, the temporality of the measurements maybe a limitation as some of the TPB constructs were framed “over the next month” whereas PA was assessed “within the past 7-days.” However, we do not believe that the lack of consistency in temporality of measurement would have a substantial effect on our findings and subsequent conclusions.

To date, however, this report appears to be the only testing the TPB in a large population sample of overweight/obese adolescents. There are very few TPB studies that have operationalized the TPB in the PA domain and to our knowledge there are no interventions targeting this population that have used the TPB. This study also included mediating and moderating (by weight status) results. Other study strengths include use of valid measures and a Web-based platform for data collection which facilitated access to a large number of students over a large geographic area.

Study Implications

In summary, the relationships among TPB constructs in our study are similar to those reported in other adolescent PA–TPB studies (with the exception of intention’s lack of correspondence with behavior in our sample, which is discussed below).

Results of our study provide valuable insight for explaining adolescent PA intention and behavior, which can be translated into sport, recreation, and educational settings. The present study indicates that PBC is the strongest correlate of PA behavior; whereas attitude plays a key role in the formation of PA intention for overweight and obese adolescents. Across both samples in our study, subjective norm provided the least contribution in explaining intention to be physically active. This may suggest that overweight/obese adolescents do not perceive social pressures as an important or apparent influence when contemplating PA. Rather, it may be that overweight/obese adolescents are overly self-conscious, and therefore, practitioners should be encouraged to use individualized strategies to promote participation in PA. G. D. C. Ball et al. (2011) demonstrated that one-on-one lifestyle coaching approaches can improve measures of obesity in adolescents, and the use of patient-centered, motivation-based approaches to weight management have been recommended for this population group.

These findings have implications for tailoring and future intervention design in this population. To achieve greatest PA gains, future interventions should implement strategies that reflect our findings on the importance of attitude and especially PBC, for overweight/obese youth. Our results suggest that PA interventions aimed at overweight/obese adolescents should endeavor to provide information highlighting the benefits of participating in regular PA (Boudreau & Godin, 2007), for example, making PA fun/enjoyable, and developing strategies to show the importance of being physically active. More important, strategies in developing confidence in participating in PA when tired, busy with homework, along with parental strategies for transportation to sporting activities, are required to enhance the efficacy and control for PA behavior change. Indeed, Blanchard et al. (2005) argue that improving self-efficacy along with access to PA facilities for overweight/obese adolescents may enhance participation in PA in this target population. A continued understanding of the issues faced by overweight/obese adolescents in increasing PA levels is necessary for the development of successful PA promotion strategies.

The limited association between intention and behavior reported in our study is noteworthy. Investigations on the intention–behavior relationship reveal that intentions do not always translate into behavior (Gollwitzer & Sheeran, 2006). “Implementation intentions” are “if-then” plans specifying when, where and how one will act in order to achieve a goal, so that individuals commit themselves to acting as soon as the specified situation is encountered. Forming “implementation intentions” has been proposed as a potentially effective strategy to help people act on their positive intentions (Gollwitzer & Sheeran, 2006). Planning and goal setting have indeed, been shown to be associated with PA behavior (Hurkmans et al., 2010; Vallance, Plotnikoff, Karvinen, Mackey, & Courneya, 2010).

Future applied tests of the TPB to explain PA and programs to promote PA in this overweight/obese population should assess and operationalize implementation intentions through action planning and coping planning. Specifically, action planning should consider (a) which physical activities to participate in, (b) where to participate in the physical activities, (c) on which days to participate in the physical activities, and (d) the duration of participating in physical activities. Coping planning, a barrier-focused strategy which requires individuals to plan how they would overcome certain behavioral obstacles, should consider (a) what to do if obstacles arise, (b) how to cope with setbacks, (c) what to do in difficult situations, (d) which options will result in action, and (e) when extra effort is needed to prevent lapses.

Other social cognitive theories/integrated models along with multilevel ecological models are recommended to be tested in future research studies in the overweight/obese adolescent population to further provide an evidence base for designing programs for practitioners.

Footnotes

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article:

RCP is supported by a Senior Research Fellowship Salary Award from the National Health and Medical Research Council, Australia.

Notes

References

Ajzen

(1991). The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50, 179-211.

Araújo-Soares

McIntyre

Sniehotta

F. F.

(2009). Predicting changes in physical activity among adolescents: The role of self-efficacy, intention, action planning and coping planning. Health Education Research, 24, 128-139.

Armitage

C. J.

Conner

(2001). Efficacy of the theory of planned behavior: A meta-analytic review. British Journal of Social Psychology, 40, 471-499.

Baker

C. W.

Little

T. D.

Brownell

K. D.

(2003). Predicting adolescent eating and activity behaviors: The role of social norm and personal agency. Health Psychology, 22, 189-198.

Ball

G. D. C.

Mackenzie-Rife

K. A.

Newton

M. S.

Alloway

C. A.

Slack

J. M.

Plotnikoff

R. C.

Goran

M. I.

(2011). One-on-one lifestyle coaching for managing adolescent obesity: Findings from a pilot, randomized controlled trial in a real-world, clinical setting. Paediatrics and Child Health, 16, 346-350.

Ball

Crawford

Owen

(2000). Obesity as a barrier to physical activity. Australian and New Zealand Journal of Public Health, 24, 331-333.

Biro

F. M.

Wein

(2010). Childhood obesity and adult morbidities. American Journal of Clinical Nutrition, 91(Suppl.), 1499s-1505s.

Blanchard

C. M.

McGannon

K. R.

Spence

J. C.

Rhodes

R. E.

Nehl

Baker

Bostwick

(2005). Social ecological correlates of physical activity in normal weight, overweight, and obese individuals. International Journal of Obesity, 29, 720-726.

Bollen

K. A.

(1989). Structural equations with latent variables. New York, NY: John Wiley.

10.

Bollen

K. A.

Stine

R. A.

(1992). Bootstrapping goodness-of-fit measures in structural equation models. Social Research Methods, 21, 205-229.

11.

Boudreau

Godin

(2007). Theory of planned behavior to predict exercise intention in obese adults. Canadian Journal of Nursing Research, 39, 112-125.

12.

Brownson

R. C.

Eyler

A. A.

King

A. C.

Brown

D. R.

Shuy

Y. L.

Sallis

J. F.

(2000). Patterns and correlates of physical activity among US women 40 years and older. American Journal of Public Health, 90, 264-270.

13.

Burke

Beilin

L. J.

Dunbar

(2001). Family lifestyle and parental body mass index as predictors of body mass index in Australian children: A longitudinal study. International Journal of Obesity and Related Metabolic Disorders, 25, 147-157.

14.

Calengor

K. E.

(2007). Nutrition, physical activity and related health behaviours in youth as assessed by a web-survey (Unpublished doctoral dissertation). University of Alberta, Edmonton, Canada.

15.

Cheung

G. W.

Rensvold

R. B.

(2002). Evaluating goodness-of-fit indexes for testing measurement invariance. Structural Equation Modeling, 9, 233-255.

16.

Cohen

(1983). Applied multiple regression/ correlation analysis for the behavioral sciences. Hillsdale, NJ: Lawrence Erlbaum.

17.

Cole

T. J.

Bellizzi

M. C.

Flegal

K. M.

(2000). Establishing a standard definition for child overweight and obesity worldwide: International survey. British Medical Journal, 320, 1240-1243.

18.

Conner

Sparks

(2005). Theory of planned behavior and health behavior. In Conner

Norman

(Eds.), Predicting health behavior (pp. 170-222). London, England: Open University Press.

19.

Craig

Goldberg

Dietz

W. H.

(1996). Psychosocial correlates of physical activity among fifth and eighth graders. Preventive Medicine, 25, 506-513.

20.

Crocker

P. R.

Bailey

D. A.

Faulkner

R. A.

Kowalski

K. C.

McGrath

(1997). Measuring general levels of physical activity: Preliminary evidence for the Physical Activity Questionnaire for Older Children. Medicine & Science in Sports & Exercise, 29, 1344-1349.

21.

de Bruijn

Kremers

S. P. J.

Lensvelt-Mulders

de Vries

Van Mechelen

Brug

(2006). Modelling individual and physical environmental Factors with adolescent physical activity. American Journal of Preventive Medicine, 30, 507-512.

22.

Department of Health & Aging. (2004). Australia’s physical activity recommendations for 5-12 year olds. Canberra: Department of Health & Aging.

23.

Dishman

R. K.

(1994). The measurement conundrum in exercise adherence research. Medicine & Science in Sports & Exercise, 26, 1382-1390.

24.

Dishman

R. K.

Hales

D. P.

Pfeiffer

K. A.

Felton

G. A.

Saunders

Ward

D. S.

. . . Pate

R. R.

(2006). Physical self-concept and self-esteem mediate cross-sectional relations of physical activity and sport participation with depression symptoms among adolescent girls. Health Psychology, 25, 396-407.

25.

Dzewaltowski

D. A.

Noble

J. M.

Shaw

J. M.

(1990). Physical activity participation: Social cognitive theory versus the theories of reasoned action and planned behavior. Journal of Sport & Exercise Psychology, 12, 388-405.

26.

Everson

E. S.

Daley

A. J.

Ussher

(2007). Brief report: The theory of planned behaviour applied to physical activity in young people who smoke. Journal of Adolescence, 30(2), 347-51.

27.

Gill

T. P.

Baur

L. A.

Bauman

A. E.

Steinbeck

K. S.

Storlien

L. H.

Fiatarone Singh

M. A.

. . . Caterson

I. D.

(2009). Childhood obesity in Australia remains a widespread health concern that warrants population-wide prevention programs. Medical Journal Australia, 190, 146-148.

28.

Glanz

Rimer

B. K.

Lewis

F. M.

(2002). Theory, research, and practice in health behavior and health education. In Glanz

Rimer

B. K.

Lewis

F. M.

(Eds.), Health behavior and health education: Theory, research, and practice (3rd ed., pp. 22-39). San Francisco, CA: Jossey-Bass.

29.

Godin

Anderson

Lambert

Desharnais

(2005). Identifying factors associated with regular physical activity in leisure time among Canadian adolescents. American Journal of Health Promotion, 20, 20-27.

30.

Godin

Kok

(1996). The theory of planned behavior: A review of its applications to health related behaviors. American Journal of Health Promotion, 11, 87-98.

31.

Gollwitzer

P. M.

Sheeran

(2006). Implementation intentions and goal achievement: A meta-analysis of effects and processes. Advances in Experimental Social Psychology, 38, 70-110.

32.

Hagger

M. S.

Chatzisarantis

Biddle

S. J. H.

Orbell

(2001). Antecedents of children’s physical activity intention and behavior: Predictive validity and longitudinal effects. Psychology & Health, 16, 391-407.

33.

Hagger

M. S.

Chatzisarantis

N. L. D.

Biddle

S. J. H.

(2002). A meta-analytic review of the theories of reasoned action and planned behavior in physical activity: Predictive validity and the contribution of additional variables. Journal of Sport & Exercise Psychology, 24, 3-32.

34.

Hamilton

White

K. M.

(2008). Extending the theory of planned behavior: The role of self and social influences in predicting adolescent regular moderate-to-vigorous physical activity. Journal of Sport & Exercise Psychology, 30, 56-74.

35.

Bentler

P. M.

(1999). Cutoff criteria for fit indices in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1-55.

36.

Hurkmans

E. J.

Maes

de Gucht

Knittle

Peeters

A. J.

Ronday

H. K.

Vliet Vlieland

T. P. M.

(2010). Motivation as a determinant of physical activity in patients with rheumatoid arthritis. Arthritis Care & Research, 62, 371-377.

37.

Janssen

Lam

Katzmarzyk

P. T.

(2009). Influence of overweight and obesity on physician costs in adolescents and adults in Ontario, Canada. Obesity Reviews, 10, 51-57.

38.

Jöreskog

K. G.

Sörbom

(1993). LISREL8: Structural equation modeling with the SIMPLIS command language. Hillsdale, NJ: Lawrence Erlbaum.

39.

Keith

T. Z.

(1993). Latent variable structural equations modeling: LISREL in special education research. Remedial and Special Education, 14, 36-46.

40.

Kowalski

K. C.

Crocker

P. R. E.

Faulkner

R. A.

(1997). Validation of the Physical Activity Questionnaire for Older Children. Pediatric Exercise Science, 9, 174-186.

41.

Kremers

S. P. J.

de Bruijn

G. J.

Visscher

T. L. S.

van Mechelen

de Vries

N. K.

Brug

(2006). Environmental influences on energy balance-related behaviors: A dual-process law. International Journal of Behavioral Nutrition and Physical Activity, 3, 9-12.

42.

Lutz

R. S.

Karloy

Okun

M. A.

(2008). The why and the how of goal pursuit: Self determination, goal process cognition, and participation in physical exercise. Psychology of Sport and Exercise, 9, 559-575.

43.

MacKinnon

D. P.

(2008). Introduction to statistical mediation analysis. New York, NY: Taylor & Francis.

44.

MacKinnon

D. P.

Fritz

M. S.

Williams

Lockwood

C. M.

(2007). Distribution of the product confidence limits for the indirect effect: Program PRODCLIN. Behavior Research Methods, 39, 384-389.

45.

MacKinnon

D. P.

Lockwood

C. M.

Williams

(2004). Confidence limits for the indirect effect: Distribution of the product and resampling methods. Multivariate Behavioral Research, 39, 99-128.

46.

Marsh

H. W.

Hau

K. T.

Wen

(2004). In search of golden rules: Comment on hypothesis-testing approaches to setting cutoff values for fit indexes and dangers in overgeneralizing Hu and Bentler’s 1999 findings. Structural Equation Modeling, 11, 320-41.

47.

Martin

J. J.

Oliver

McCaughtry

(2007). The theory of planned behavior: Predicting physical activity in Mexican American children. Journal of Sport & Exercise Psychology, 29, 225-238.

48.

Martinez-Gonzalez

M. A.

Martinez

J. A.

F. B.

Gibney

M. J.

Kearney

(1999). Physical inactivity, sedentary lifestyle and obesity in the European Union. International Journal of Obesity and Related Metabolic Disorders, 23, 1192-1201.

49.

McCrindle

B. W.

Manlhiot

Millar

Gibson

Stearne

Kilty

. . . Dobbin

S. W.

(2010). Population trends toward increasing cardiovascular risk factors in Canadian adolescents. Journal of Pediatrics, 157, 837-843.

50.

Michie

Abraham

(2004). Identifying techniques that promote health behavior change: Evidence based or evidence inspired? Psychology & Health, 19, 29-49.

51.

Mummery

W. K.

Spence

J. C.

Hudec

J. C.

(2000). Understanding physical activity intention in Canadian school children and youth: An application of the theory of planned behavior. Research Quarterly for Exercise and Sport, 71, 116-124.

52.

Nigg

C. R.

Lippke

Maddock

J. E.

(2009). Factorial invariance of the theory of planned behavior applied to a physical activity across gender, age, and ethnic groups. Psychology of Sport and Exercise, 10, 219-225.

53.

Ogden

C. L.

Carroll

M. D.

Curtin

L. R.

McDowell

M. A.

Tabak

C. J.

Flegal

K. M.

(2006). Prevalence of overweight and obesity in the United States, 1999-2004. Journal of the American Medical Association, 95, 1549-1555.

54.

Plotnikoff

Karunamuni

Spence

Storey

Forbes

Raine

. . . McCargar

(2009). Chronic disease-related lifestyle risk factors in a sample of Canadian adolescents. Journal of Adolescent Health, 44, 606-609.

55.

Plotnikoff

R. C.

Lubans

D. R.

Costigan

S. A.

Trinh

Spence

Downs

McCargar

(2011). Test of the theory of planned behavior to explain physical activity in a large population sample of adolescents from Alberta, Canada. Journal of Adolescent Health, 49, 547-549.

56.

Public Health Agency of Canada. (2002). Canada’s physical activity guide for children and youth. Retrieved from http://www.phac-aspc.gc.ca/hp-ps/hl-mvs/pag-gap/cy-ej/index-eng.php

57.

Rey-Lopez

J. P.

Vicente-Rodriguez

Biosca

Moreno

L. A.

(2008). Sedentary behavior and obesity development in children and adolescents. Nutrition, Metabolism & Cardiovascular Diseases, 18, 242-251.

58.

Rhodes

R. E.

Macdonald

H. M.

McKay

H. A.

(2006). Predicting physical activity intention and behaviour among children in a longitudinal sample. Social Science & Medicine, 62, 3146-3156.

59.

Sas-Nowosielski

(2006). Application of the theory of planned behavior in predicting leisure time physical activity of Polish adolescents. Human Movement, 7, 105-110.

60.

Saunders

R. P.

Motl

R. W.

Dowda

Dishman

R. K.

Pate

R. R.

(2004). Comparison of social variables for understanding physical activity in adolescent girls. American Journal of Health Behavior, 28, 426-436.

61.

Spence

Blanchard

Clark

Plotnikoff

Storey

Kennedy

McCargar

(2010). The role of self-efficacy in explaining gender differences in physical activity among adolescents: A multilevel analysis. Journal of Physical Activity & Health, 7, 176-183.

62.

Storey

Forbes

L. E.

Fraser

S. N.

Spence

J. C.

Plotnikoff

R. C.

Raine

K. D.

. . . McCargar

L. J.

(2009). Diet quality, nutrition and physical activity among adolescents: The Web-SPAN (Web-Survey of Physical Activity and Nutrition) project. Public Health Nutrition, 12, 2009-2017.

63.

Tremblay

M. S.

Shields

Laviolette

Craig

Janssen

Connor

S. G.

(2010). Fitness of Canadian children and youth: Results from the 2007-2009 Canadian Health Measures Survey. Health Reports, 21, 1-14.

64.

Tremblay

M. S.

Willms

J. D.

(2003). Is the Canadian childhood obesity epidemic related to physical inactivity? International Journal of Obesity, 27, 1100-1105.

65.

Trost

S. G.

Saunders

Ward

D. S.

(2002). Determinants of physical activity in middle school children. American Journal of Health Behavior, 26, 95-102.

66.

Vallance

Plotnikoff

R. C.

Karvinen

K. H.

Mackey

J. R.

Courneya

K. S.

(2010). Understanding physical activity maintenance in breast cancer survivors. American Journal of Health Behavior, 34, 225-236.

67.

Wake

Canterford

Patton

Hesketh

Hardy

Williams

. . . Carlin

J. B.

(2009). Comorbidities of overweight/obesity experienced in adolescence: Longitudinal study. Archives of Disease in Childhood, 95, 162-168.

68.

Wang

Lobstein

(2006). Worldwide trends in childhood overweight and obesity. International Journal of Pediatric Obesity, 1, 11-25.