Validation of a Social Ecological Model-Based Scale for Physical Activity Among Thai Adolescents
Article information
Abstract
OBJECTIVES
The present study aims to validate a scale, based on the Social Ecological Model (SEM), that reflects the individual, social, and environmental determinants of physical activity among Thai adolescents.
METHODS
The SEM-based scale was culturally adapted for Thai adolescents through standardized translation, expert panel review, and cognitive interviewing. Data were obtained from 495 adolescents aged 12-19 years. Construct validity was examined using confirmatory factor analysis with multiple fit indices (CFI, IFI, TLI, RMSEA, SRMR), while reliability, convergent, and discriminant validity were systematically evaluated.
RESULTS
Among 495 Thai adolescents (mean age = 15.13 years; 54.1% male), females reported higher self-efficacy, perceived barriers, and family support. CFA supported the six-factor SEM model with good fit (CFI = 0.905, TLI = 0.900, RMSEA = 0.046). Most constructs showed adequate convergent validity (AVE ≥ 0.50, CR ≥ 0.79), while self-efficacy had lower AVE (0.40) but strong reliability (CR = 0.91, α = 0.91). Discriminant validity was confirmed.
CONCLUSIONS
This validated scale provides a reliable tool for Thai adolescent physical activity research in Southeast Asia, supporting evidence-based interventions and future longitudinal studies.
Introduction
The increasing prevalence of physical inactivity and sedentary lifestyles in contemporary societies has emerged as a critical global public health concern [1]. Adolescence represents a pivotal developmental stage characterized by rapid physical and psychological growth, during which regular and sufficient engagement in physical activity plays an essential role in preventing chronic diseases and enhancing overall quality of life [2]. Nevertheless, a substantial body of evidence consistently indicates that a large proportion of adolescents worldwide fail to achieve the recommended levels of physical activity [3,4].
Reflecting these worldwide trends, Thai adolescents have similarly exhibited a steady decline in physical activity levels, with girls participating at significantly lower rates than boys [5]. Notably, the growing use of digital devices, rapid urbanization, traffic congestion, and air pollution have been identified as major contributors to increasingly sedentary lifestyles, further exacerbating the rising prevalence of obesity and related health risks in this population [6].
Physical activity is a multidimensional and complex behavior shaped by the interaction of factors across multiple levels [7]. Bandura’s Social Cognitive Theory [8] provides a useful framework for explaining individual and psychological determinants, while the Social Ecological Model (SEM) offers a comprehensive approach that simultaneously accounts for influences at the individual, social, and environmental levels [9,10]. SEM encompasses personal attributes such as attitudes, beliefs, and self-efficacy [11], social influences such as family and peer support [12], and environmental factors such as access to facilities and perceptions of safety [13].
The SEM has demonstrated strong explanatory power for adolescent physical activity across diverse cultural contexts [7,14]. Nevertheless, sociocultural differences may shape the relative influence of these determinants. For instance, while school curricula in Western countries often emphasize physical education and youth sports participation, Asian societies tend to prioritize academic achievement, which may contribute to lower levels of physical activity among adolescents [15,16].
Nevertheless, existing research remains predominantly Western-centric, and studies on physical activity grounded in the SEM are still limited in the Asian context [17]. Consequently, there is a pressing need for research that elucidates how sociocultural and environmental factors differentially influence adolescent physical activity across Asian countries.
Against this backdrop, the present study aims to validate a scale, based on the SEM, that reflects the individual, social, and environmental determinants of physical activity among Thai adolescents. This validation is expected to provide a theoretical foundation for comparative studies between Korean and Thai youth and, more broadly, to generate empirical evidence for advancing adolescent physical activity promotion across diverse sociocultural contexts in Asia.
Methods
Translation and development of the scale
To validate a Social Ecological Model (SEM) scale designed to assess the physical activity characteristics of Thai adolescents, this study first reviewed conceptual frameworks derived from prior SEM-based research on physical activity. To ensure both linguistic and conceptual equivalence, an existing SEM scale was translated into Thai and subsequently back-translated into English.
The translation was conducted by a professor of health and exercise science fluent in both Thai and English, together with two professors holding doctoral degrees in health and exercise psychology and sport psychology who possessed advanced English proficiency. Following the International Test Commission [18] guidelines on test translation and adaptation, a triangulation approach was adopted to evaluate the interchangeability of the Thai and English versions of the items.
To further examine equivalence between the translated and original items, a panel of five experts—including professors and researchers specializing in health and exercise psychology, sport psychology, and health and exercise science—was convened. Items deemed inconsistent or ambiguous were retranslated and revised to enhance precision and clarity.
Subsequently, cognitive interviews were conducted with 30 Thai adolescents aged 12 to 19 years to evaluate both face validity and content validity. During this process, items identified as unclear, redundant, or insufficiently adapted to the sociocultural context of Thai adolescents were revised or supplemented accordingly. As a result, a preliminary Thai version of the SEM scale was established, consisting of 58 items grouped into six distinct factors.
Validation of a social ecological model scale for physical activity
Study participants and data collection
All study procedures were reviewed and approved by the Institutional Review Board of [blinded for review]. The research adhered to the ethical principles outlined in the Declaration of Helsinki and its subsequent revisions or equivalent ethical standards. Written informed consent was obtained from all participants prior to their involvement in the study.
To examine the validity of the SEM scale items, Thai adolescents between the ages of 12 and 19 years were recruited. Data collection was conducted using a Thai-translated and culturally adapted version of the questionnaire, administered either in person through structured interviews or via an online survey platform (Google Forms). All completed questionnaires were screened for missing responses and response irregularities, such as straight-lining. Only those that provided complete responses across all SEM domains—including self-efficacy, perceived benefits and barriers to physical activity, social support, and the physical environment—were retained for analysis. After quality control procedures, a final sample of 495 participants was included in the analytic dataset.
Measures
Self-efficacy for physical activity: Self-efficacy was measured using the Thai version of Bandura’s [19] 18-item Self-Efficacy Scale, developed through rigorous translation and cultural adaptation. Participants rated their confidence in performing regular physical activity under various conditions on a five-point Likert scale (1 = “Not at all confident” to 5 = “Completely confident”). The original scale showed high internal consistency (Cronbach’s α = 0.85-0.95) and strong test-retest reliability (0.90).
Perceived benefits and barriers to physical activity: Perceptions of benefits and barriers were assessed using the 10-item Decisional Balance Scale [20], adapted for this study. The two subscales—pros (benefits) and cons (barriers)—capture positive and negative influences on exercise participation. Items were rated on a five-point Likert scale (1 = “Not at all important” to 5 = “Extremely important”). Reported reliability of the original instrument was acceptable (α = 0.86 for pros,0.72 for cons).
Social support for physical activity: Social support was measured with a 24-item questionnaire [21], adapted for the present study. It consists of two subscales: 12 items on family support and 12 on peer support, rated on a five-point Likert scale. The original instrument demonstrated moderate to high reliability (test-retest:0.55-0.86 for family,0.78-0.81 for peers; α = 0.61-0.91 for family,0.80-0.87 for peers).
Physical environment for physical activity: Environmental factors were evaluated using a six-item scale by Ståhl et al. [22], translated and adapted for this study. Items assessed access, availability, and quality of facilities (e.g., “My community provides sufficient facilities for physical activity”), rated on a five-point Likert scale (1 = “Strongly disagree” to 5 = “Strongly agree”). The original scale showed acceptable reliability (α = 0.57-0.81).
Statistical analysis
All statistical procedures were carried out using IBM SPSS version 29.0 and AMOS version 28.0. Prior to the main analyses, the dataset was screened for missing responses and outliers, and only questionnaires with complete responses were retained. Normality was examined through skewness and kurtosis indices, and since the absolute values for all variables fell within acceptable thresholds (≤ 2 for skewness and ≤ 7 for kurtosis), the assumption of normality was deemed reasonably satisfied [23].
Confirmatory factor analysis (CFA) was performed in AMOS employing the Maximum Likelihood (ML) estimation method. The hypothesized model specified six latent constructs: perceived benefits and barriers to physical activity, physical environmental factors, social support from family and peers, and self-efficacy.
To preserve the cultural and conceptual integrity of the scale, item deletion was deliberately minimized, with particular emphasis on maintaining the original factorial structure during its adaptation for Thai adolescents. Model respecifications were implemented by allowing error covariances within the same latent construct, guided by modification indices (MI ≥ 10). Such adjustments were applied only when theoretically justifiable and when items were conceptually or empirically prone to cooccurrence. If model fit remained unsatisfactory, items with standardized factor loadings below 0.40 were considered for removal [24-26].
The adequacy of the structural model was assessed using multiple fit indices, with the following thresholds indicating acceptable fit: Comparative Fit Index (CFI) ≥ 0.90, Incremental Fit Index (IFI) ≥ 0.90, Tucker-Lewis Index (TLI) ≥ 0.90, Root Mean Square Error of Approximation (RMSEA) between 0.05 and 0.08, and Standardized Root Mean Square Residual (SRMR) < 0.08 [27].
Construct validity was evaluated through both convergent and discriminant validity. Convergent validity was examined using standardized factor loadings, average variance extracted (AVE), and composite reliability (CR). Although Cronbach’s alpha is commonly reported, it can underestimate true reliability in models accounting for residual covariances; therefore, both Cronbach’s alpha coefficients and CR values were presented in this study [28]. Acceptable thresholds were CR ≥ 0.70 [29] and AVE ≥ 0.50 [30].
Discriminant validity was assessed by comparing the correlations among latent constructs. Following Fornell and Larcker [30], discriminant validity was considered adequate if the square root of the AVE (√AVE) of each construct exceeded its correlations with all other constructs.
Results
Demographic characteristics
The participants were Thai adolescents with a mean age of 15.13 years (SD = 1.42). The sample comprised 268 males (54.1%) and 227 females (45.9%), providing a balanced representation across sexes.
Based on the constructs of the Social Ecological Model (SEM), the characteristics of physical activity among Thai adolescents were as follows: Self-efficacy for physical activity showed a mean score of 2.60 (SD = 0.83; males = 2.40, females = 2.83). Perceived benefits to physical activity had a mean of 3.76 (SD = 0.84; males = 3.71, females = 3.81), whereas perceived barriers demonstrated a mean of 2.65 (SD = 0.89; males = 2.58, females = 2.74). Family support for physical activity yielded a mean score of 2.70 (SD = 1.06; males = 2.47, females = 2.96), while friend support showed a mean of 2.73 (SD = 0.91; males = 2.67, females = 2.80). Physical environmental factors were reported at a mean of 3.29 (SD = 0.93; males = 3.29, females = 3.29).
Regarding gender differences, females (M = 2.83) reported significantly higher self-efficacy than males (M = 2.40; t = -5.952, p < 0.001). Similarly, females (M = 2.74) exhibited greater perceived barriers compared with males (M = 2.58; t = -2.050, p = 0.041). Family support for physical activity was also significantly higher among females (M = 2.96) than males (M = 2.47; t = -5.112, p < 0.001). Full statistical details are provided in <Table 1>.
Demographic characteristics and social ecological determinant profiles.
Validation of a social ecological model scale for physical activity
Confirmatory factor analysis
After verifying data normality, confirmatory factor analysis (CFA) was performed using the Maximum Likelihood (ML) estimation method in AMOS version 28.0 <Table 2>. The hypothesized measurement model comprised six latent constructs: self-efficacy for physical activity, perceived benefits and barriers to physical activity, social support from friends and family, and physical environmental factors. The initial model failed to reach the recommended cutoff thresholds for several fit indices.
Model fit indices for social ecological determinants.
To enhance model fit, residual covariances were introduced between error terms within the same latent construct when modification indices (MI ≥ 10) indicated statistically meaningful correlations. Specifically, covariances were added for the following item pairs: one within self-efficacy, two within perceived benefits, one within perceived barriers, one within friend support, one within family support, and one within physical environmental factors.
Despite these adjustments, certain items demonstrated either insufficient standardized factor loadings ( < 0.40) or excessively high residuals. Consequently, two items from self-efficacy and one from perceived barriers were eliminated. Following these modifications, the revised model demonstrated acceptable levels of fit: CFI = 0.905, TLI = 0.900, IFI = 0.906, SRMR = 0.050, and RMSEA = 0.046.
Evaluation of convergent and discriminant validity
To assess the convergent validity of the latent constructs, the average variance extracted (AVE), composite reliability (CR), and Cronbach’s alpha coefficients were calculated. Most standardized factor loadings exceeded the recommended threshold of 0.50, indicating acceptable item quality, and all CR and Cronbach’s alpha values surpassed the 0.70 criterion, confirming satisfactory internal consistency across constructs.
The results of the AVE, CR, and Cronbach’s alpha analyses indicated that perceived benefits (AVE = 0.50, CR = 0.83, α = 0.84), perceived barriers (AVE = 0.50, CR = 0.79, α = 0.78), family support (AVE = 0.64, CR = 0.96, α = 0.96), friend support (AVE = 0.52, CR = 0.93, α = 0.93), and physical environmental factors (AVE = 0.51, CR = 0.86, α = 0.87) all met the recommended thresholds for convergent validity. Specifically, AVE values were at or above the 0.50 benchmark, while CR values consistently exceeded 0.70, suggesting that an acceptable level of convergent validity was established across these constructs.
However, the construct of self-efficacy for physical activity (SE) demonstrated a relatively low AVE value of 0.40, suggesting limited shared variance among its items. Nevertheless, the CR (0.91) and Cronbach’s alpha (0.91) for SE indicated strong internal consistency, suggesting that the construct retained adequate measurement reliability <Table 3>.
Convergent validity indices for social ecological determinants.
To evaluate discriminant validity, the square root of the AVE (√AVE) for each latent construct was compared with the inter-factor correlation coefficients (√AVE > r). This procedure confirmed the distinctiveness of the constructs. The results revealed that for perceived benefits (√AVE = 0.706), perceived barriers (√AVE = 0.697), self-efficacy (√AVE = 0.626), family support (√AVE = 0.802), friend support (√AVE = 0.723), and physical environmental factors (√AVE = 0.711), the √AVE values were greater than the corresponding inter-construct correlations. These findings provide evidence of adequate discriminant validity across all constructs <Table 4>.
Discriminant validity: Inter-factor correlations and √AVE.
Discussion
This study aimed to establish a theoretically robust and ecologically valid measurement model by systematically analyzing the individual, social, and environmental determinants of physical activity among Thai adolescents, grounded in the Social Ecological Model (SEM).
The gender-based analyses revealed that female adolescents reported significantly higher levels of self-efficacy, perceived barriers, and family support compared to their male counterparts. These findings suggest that both internal and external motivational factors—such as confidence in one’s ability to be active and encouragement from family—may play a critical role in shaping female adolescents’ participation in physical activity [31]. Conversely, female adolescents also perceived greater barriers to physical activity, highlighting how sociocultural norms and gender role expectations in Thailand and across Asia may disproportionately constrain girls’ opportunities for participation [32].
In the confirmatory factor analysis, a strategic approach was adopted to minimize item deletion in order to maintain the structural integrity of the scale and ensure cultural acceptability. Such an approach is recommended when validating theory-driven instruments in new cultural contexts, as it preserves the conceptual completeness of the original scale while allowing for the reflection of culture-specific perceptions and response patterns [33,34].
During model respecification, covariance paths between error terms were added primarily within the same latent construct to account for shared measurement error among conceptually similar items. This procedure was guided by modification indices (MI ≥ 10) and applied only when supported by strong theoretical justification, such as semantic similarity or the likelihood of simultaneous responses [25]. Nonetheless, certain items were removed when standardized regression coefficients fell below.40 or when strong conceptual redundancy between items was identified [24,25]. Specifically, within the perceived barriers construct, the item “I am too tired to exercise because of other daily tasks (PBR1)” was eliminated due to conceptual overlap with “Exercise takes too much of my time (PBR2).”
Within the self-efficacy construct, two items (SE2 and SE16) were removed. These items assessed confidence in exercising under specific conditions, such as stressful situations (SE2) and during rainy or snowy weather (SE16). SE2 yielded inconsistent responses due to its subjective interpretation and contextual ambiguity, while SE16 was deemed culturally and environmentally irrelevant given Thailand’s climatic conditions, thereby demonstrating limited conceptual validity for Thai adolescents. The deletion of these items was therefore considered a rational adjustment that enhanced both the conceptual validity and response consistency of the scale by reflecting the cultural and environmental realities of the target population.
The validation of the SEM-based scale demonstrated that most constructs satisfied the criteria for convergent validity (AVE ≥ 0.50, CR ≥ 0.70). Specifically, perceived benefits, perceived barriers, friend support, family support, and physical environment all displayed satisfactory AVE and CR values, indicating strong convergent validity and internal consistency. In contrast, self-efficacy exhibited a high composite reliability (CR = 0.91), exceeding the recommended threshold, but its AVE fell below.50, suggesting limited shared variance among items.
This indicates that while the items captured diverse facets of the construct, the overall reliability of the scale remained intact. Such findings reflect conceptual heterogeneity and the developmental characteristics of adolescence, a stage marked by psychological fluidity and contextual sensitivity, where self-efficacy as an internal motivational factor may be shaped by complex external influences, thereby limiting covariance among items [35]. Consequently, the validation and interpretation of psychosocial measures in adolescents require theoretical refinement that accounts for developmental and cultural contexts.
Discriminant validity analyses further confirmed that the square root of AVE for each construct exceeded inter-construct correlations, indicating that the factors were distinct and nonoverlapping. The presence of low but statistically significant correlations between constructs also supported the conceptual independence of internal psychological factors and external environmental determinants [36,37]. This finding demonstrates that SEM was applied in a theoretically coherent manner to structure the multilevel determinants of physical activity and that the scale achieved discriminant validity among Thai adolescents.
Overall, this study provides evidence that SEM represents a theoretically and culturally appropriate framework for adolescents in Asia. These findings lay an important foundation for cross-national comparative studies within the region. More specifically, the validated scale developed in this study offers a practical tool for policy initiatives and community-based interventions. By systematically assessing the social and physical environmental correlates of adolescent physical activity, this instrument holds particular relevance in Southeast Asian contexts where urbanization, traffic congestion, and air pollution constitute salient environmental barriers [13,38].
Nonetheless, this study is subject to several limitations. The cross-sectional design restricts the ability to draw causal inferences. Additionally, the absence of objective measures of participants’ actual physical activity levels represents a methodological shortcoming. Future research should employ longitudinal designs to examine the dynamic interplay between socioecological determinants and physical activity behaviors, while also integrating objective or observational methods—such as accelerometers or wearable devices—to enhance the validity of self-reported data.
Conclusions
The validation of the Social Ecological Model scale for Thai adolescents’ physical activity demonstrated structural validity across most latent constructs, with satisfactory levels of both convergent and discriminant validity. Although certain indicators, such as self-efficacy, showed relatively low average variance extracted (AVE) values, their composite reliability (CR) remained high, indicating sufficient internal consistency.
Notes
Acknowledgments
This work was supported by the Ministry of Education of the Republic of Korea and the National Research Foundation of Korea (NRF-2020S1A5A2A03041894)..
The funding sponsors had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, and in the decision to publish the results.