DEVELOPMENT AND VALIDATION OF THE ARABIC VERSIONS OF THE TRANSTHEORETICAL MODEL SCALES FOR EXERCISE

High prevalence of physical inactivity in Saudi Arabia is a major public health problem that contributed to the increasing lifestyle-related diseases. Thus, interventions to promote and increase exercise are necessary. The Transtheoretical model demonstrated significant impacts in this area, however, effective interventions require psychometrically sound measures. The goal of this study is to develop, assess the psychometric properties, and confirming the factorial invariance of the Arabic versions of the Decisional Balance and Self Efficacy measures for exercise, as well as test their predicted theoretical relationships with Stages of Change with a populationbased sample of Saudi adults (N=685). Three levels of invariance were tested: Configural Invariance (nonzero factor loadings unconstrained), Pattern Identity Invariance (equal factor loadings), and Strong Factorial Invariance (equal factor loadings and measurement errors). For Decisional Balance, the two-factor uncorrelated model was the most parsimonious good-fitting model (χ (35) = 97.803, p < .001; CFI = .922; RMSEA = .076 [90% CI = .058, .093]). Internal consistency coefficient Alpha and factor rho reliability were .86 for Pros and .53 for Cons. Strong Factorial Invariance was a good fit for the model across seven grouping variables: gender, age, health status, educational level, employment status, BMI, and stage of change for exercise. The one-factor model of Self Efficacy Scale revealed an excellent fit (χ2 (8) = 16.732, p = .033; CFI = .991; RMSEA = .056 [90% CI = .015, .094]). Internal consistency coefficient alpha was .86 and factor rho reliability was .89. Strong Factorial Invariance was a good fit for the model across all seven grouping variables. Multivariate analysis by stage of change replicated expected patterns for Pros (ω= .08), Cons (ω= .02), and Self Efficacy (ω= .21). The results demonstrate the internal and external validity and measurement invariance of the Arabic versions of TTM measures for exercise, supporting their use in research and tailored interventions to increase exercise among Saudi population, as well as supporting the applicability of the Transtheoretical model to exercise behavior in Saudi Arabia.


INTRODUCTION
There is a substantial literature indicating that physical activity can improve both physical and psychological health. People who are physically active tend to live longer and have lower risk for heart disease, stroke, type 2 diabetes, and some cancers.
It can also help with weight control, and may improve academic achievement in students (CDC, 2014). Physical activity also offers positive psychological improvements by decreasing levels of anxiety, depression, and enhancing self-esteem.
According to the 2008 Physical Activity Guidelines for Americans, adults need at least 150 minutes of moderate-intensity aerobic activity or 75 minutes of vigorousintensity aerobic every week, and muscle-strengthening activities on 2 or more days a week that work all major muscle groups: legs, hips, back, abdomen, chest, shoulders, and arms (U.S DHHS, 2008).
Despite of the known health benefits of physical activity, only 21% of adults meet the 2008 Physical Activity Guidelines (CDC, 2014), and less than 5% of adults participate in 30 minutes of physical activity each day (PCFSN, 2017).
The prevalence of physical activity varies widely by country, the highest being reported in Sweden and Denmark, and the lowest in Brazil, Thailand and Kingdom of Saudi Arabia (Sisson & Katzmarzyk, 2008).
Several studies showed that the prevalence of physical inactivity is high in Saudi Arabia. Alnoza et al (2007) examined data was collected between 1995 and 2000 using the National Epidemiological Health Survey. 17395 Saudi adults aged 30-70 years participated in the study. Leisure-type and sport-related physical activities including walking were examined. They found that inactivity prevalence was very high, 98.1% for females and 93.9% for males. Also, inactivity prevalence increases with increasing age, especially in males, and decreases with increasing education levels. Amin et al (2012) conducted a study with 2176 Saudi adults aged 18-65 years to determine the prevalence and pattern of the leisure time physical activity (LTPA).
Participants were interviewed using the Global Physical Activity Questionnaire (GPAQ). Physical activity (PA) was expressed in metabolic equivalents (METs). The results revealed that the median total METs minutes/week for LTPA for both genders was (256 METs minutes/week) compare to the cut off (600 METs-minutes/day) or 150 minutes of moderate intensity 5 or more days/week. Only 19.8% of the total PA was LTPA, 50.0% of participants reported no leisure activity, and only 21.0% were considered sufficiently active. Overall, nearly 80% of participants did not achieve the recommended LTPA level with beneficial health effects. A cross sectional study (Awadalla et al, 2014) evaluated the pattern of physical activity among students of the This high prevalence of physical inactivity in Saudi Arabia is a major public health problem that contributed to the increasing lifestyle-related diseases (e.g. coronary heart disease, diabetes, hypertension, obesity, etc.). Unless concrete steps are taken to reduce physical inactivity in the Saudi population, these diseases may keep escalating to epidemic proportions, and the public health cost will be heavily burdened (Al-Hazzaa, 2004a;Al-Hazzaa, 2004b;AlNoza et al, 2007;El Bcheraoui et al, 2016).
Given this extremely high prevalence of physical inactivity in Saudi Arabia, which may be considered among the highest in the world, effective interventions to increase exercise are necessary in Saudi Arabia. One popular theoretical framework that helps researchers to better understand how people adopt and maintain regular exercise is the Transtheoretical model of behavior change (TTM). Worldwide, interventions to promote regular exercise based on the TTM have been developed and implemented and have demonstrated significant impacts in numerous applications (Grande, Cieslak & Silva, 2016;Greaney et al, 2008;Johnson et al 2008;Marcus et al., 1996Marcus et al., ,1998Sarkin et al, 2001;Steptoe et al., 1999;Woods et al., 2002;Zhu et al, 2014 Finally, to examine the expected patterns of relationships between the Decisional Balance, the Self Efficacy and the stage of change groups in this population.

REVIEW OF LITERATURE
Factorial invariance is a crucial psychometric requirement for any measure. A measurement model is called factorially invariant when the model is the same across different groups or across different time points. This methodological approach determines whether the set of items purported to assess theoretical constructs across different subgroups are empirically valid (Meredith, 1993;Cheung & Rensvold, 2002). If invariance does not hold, comparing means between different populations are meaningless because the measurement scales are fundamentally different across the two populations (Steenkamp and Baumgartner, 1998).
The Transtheoretical Model measures for exercise have been validated in several populations. One study (Plotnikoff et al., 2001)  Another study (Musser, 2003)  (3) a characteristic crossover pattern of pros and cons would be found, with an increase in pros and a decrease in cons across the stages from the precontemplation to action stages; and (4) the relationship between Self Efficacy and stage of change would show an increase in Self Efficacy across the stages.

Overview of The Transtheoretical Model
The Transtheoretical Model is a model of behavior change with a focus on dynamic variables rather than static variables. The TTM can be conceptualized as involving three dimensions: the temporal dimension, the independent variable dimension, and the intermediate variable dimension (Velicer et al., 2000).
The temporal dimension is represented by five stages of change (SOC) describing different levels of readiness to change (e.g. engage in regular exercise).
People are classified by their readiness to change into one of five stages: Precontemplation (PC), Contemplation (C), Preparation (PR), Action (A), and Maintenance (M). The independent dimension is composed of the Processes of Change (POC) that act as strategies to bring about change (Prochaska et al., 1988;Marcus & Simkin, 1993 Relationships (Marcus et al., 1992). The intermediate/outcome variable dimension (Velicer et al., 1996)

includes a series of intermediate outcome measures, including
Decisional Balance and Self-Efficacy. The Decisional Balance are Cognitive and motivational aspects of decision-making measured by the Decisional Balance Inventory (Prochaska et al., 1994;Velicer et al., 1985), which contains two constructs the Pros and Cons of engaging in a behavior (e.g. adopting and/or increasing exercise).
The Self Efficacy is a person's confidence that they can prevent or cope with the temptation to fall back into unhealthy or high-risk behavior (e.g. confidence a person has that she or he can exercise regularly), is measured using measures of confidence and situational temptations inventory (DiClemente, 1986;Velicer et al., 1990;Marcus et al., 1994).

Design
Cross-sectional measure development study.

Participants
Participants were population-based Saudi adults (N = 685), who were recruited online via emails and social media from across the kingdom of Saudi Arabia. More than half of participants were females (55.91%  Tables 3 and 4.

Measures
A demographic questionnaire, Stages of Change, Decisional Balance, and Situational Self Efficacy for Exercise were included.
Demographic Questionnaire. Questions about participant's age, gender, education level, employment, health status, height, and weight.
Stages of Change. The Stages of Change algorithm assesses the readiness of individuals to engage in regular exercise. Regular exercise is described as any planned physical activity (i.e., brisk walking, aerobics, jogging, bicycling, swimming, rowing, etc.) intended to increase physical fitness, and performed 3 to 5 times per week for 20-60 minutes per session. Exercise does not have to be painful to be effective but should be done at a level that increases breathing rate and causes sweating. Precontemplation is defined as not exercising at that level and having no intention to do so in the next 6 months. Contemplation is defined as not currently engaging in regular exercise but intending to begin regular exercise within next 6 months. Preparation is defined as not currently engaging in regular exercise but having intention to begin regular exercise within 30 days. Action is defined as currently engaging in regular exercise for less than 6 months. Maintenance is defined as engaging in regular exercise for more than 6 months. These definitions are consistent with staging algorithm recommendations (Reed et al., 1997;Schumann et al., 2002;Hellsten el al., 2008).
Decisional Balance. This scale assesses the advantages (the Pros) and disadvantages (the Cons) of engaging in regular exercise. Five items assessing the Pros of exercising (α = .90), and five items assessing the Cons of exercising (α = .67) was administered. Individuals responded on a five-point scale (1 = Not at all Important -5 = Extremely Important). Higher scores on the Pros and lower scores on the Cons items indicate that an individual perceived exercise as advantageous, while lower scores on the Pros and higher scores on the Cons would indicate that an individual perceived exercise as disadvantageous Blaney et al, 2012).
Situational Self Efficacy. A six-item Situational Self Efficacy scale (α = .82) is used to assess the confidence of individuals to engage in regular exercise across a variety of challenging circumstances. Participants rated their confidence levels for each item from 1 to 5 (1 = Not at all Confident -5 = Completely Confident). Higher scores indicated higher levels of confidence to exercise even across challenging circumstances Blaney et al, 2012).

Translation and Cross-Cultural Adaptation procedures.
The methodology used for translation and adaptation followed the published guidelines for the cross-cultural adaptation of self-reported measures by Beaton et al. (2000).

Initial translation. Forward translation of Stages of change, Decisional
Balance, and Self Efficacy scales for exercise from English into Arabic by two native Arabic-speaking translators fluent in English (T1, T2). One of these translators was familiar with the subject and the constructs that are being assessed, while the second translator was unaware of the concepts addressed.
To ensure a greater cultural fit, both translators avoided literal translation of items. For example, in the Decisional Balance scale "I would feel more comfortable with my body if exercised regularly" was translated to "I would feel more satisfied with my body if exercised regularly" since Arabs are not accustomed to hear words like "comfortable" used to describe their feeling about their bodies, unlike words like "satisfied" or "confident" which are more popular in this subject. Also "Exercise puts an extra burden on my significant other" was translated to "Exercise puts an extra burden on the most important person in my life; e.g. spouse or any beloved one" to fit Saudi culture. As with the Self Efficacy scale "when it's raining or snowing" was translated to "when it's hot or humid".

Statistical analysis
Several sets of analyses were conducted on the Arabic versions the Decisional Balance and Self Efficacy scales for exercise using EQS 6.1 and SAS 9.2 software packages. indicating a very good fit (Kline, 2011).

Confirmatory factor analyses (CFA
2. Factorial Invariance. Three levels of invariance were examined in sequential order with each level requiring more constraints: (1) configural invariance (unconstrained nonzero factor loadings); (2) pattern identity invariance (equal factor loadings); (3) strong factorial invariance (equal factor loadings and measurement errors) (Meredith & Teresi, 2006, Meredith, 1993. In addition to the model fit indices (CFI and RMSEA) described, the difference in CFI (ΔCFI) values between the higher level model and the lower level of invariance was calculated. A difference of .01 or smaller indicates that the null hypothesis of invariance should not be rejected and that the model demonstrates invariance (Cheung & Rensvold, 2002). The present study emphasized ΔCFI as the final determinant of measurement invariance due to the susceptibility of Chi-squared to sample size and model complexity that may reject null hypotheses when only trivial model differences exist (Bentler & Bonett, 1980;Cheung and Rensvold, 2002;Hu & Bentler, 1999;Kline, 2011;Wu, Li, & Zumbo, 2007 3. Scale Reliabilities. The internal consistency reliabilities of Decisional Balance and Self Efficacy scales were assessed with Cronbach's coefficient Alpha (Cronbach, 1951). In addition, factor rho reliability coefficients were calculated by using unstandardized model estimates (Kline, 2011).

Known Groups Validation. A MANOVA, with follow up ANOVAs and
Tukey tests, were conducted for each scale (Pros, Cons, and Self Efficacy) to examine functional relationships between these scales (means in standardized -score) and the stage of change groups. Also, effect sizes (ω 2 ) were calculated.

I. Decisional Balance scale for Exercise
Step I: Initial Confirmatory Factor Analysis was conducted on the original 10 items from the Decisional Balance scale (N=685 Step II: Exploratory Analysis. To investigate the measurement structure of the Arabic version of the Decisional Balance scale for exercise, the sample was randomly split in two sections. First sample (N=324) was used for exploratory item analysis using principal components analysis, and the second sample (N=314) for confirmatory analysis using structural equation modeling. The initial 15 items (10 original items & 5 additional items) from the Decisional Balance scale were analyzed using principal component analysis with orthogonal varimax rotation (N=324). As expected, there were two factors with mostly simple structure. These two factors explained 44% of variance in these 15 items (Table 5 gives means and Standard deviation of these 15 items). The analysis was repeated eliminating items with loadings < .40 or with cross loadings on the non-target factor (e.g., "There is too much I would have to learn to exercise" and "Exercise increases my appetite for food"). Again, there were two factors that explained 47% of variance in these 13 items (8 items for Pros & 5 items for Cons). Finally, to obtain symmetry between Pros and Cons, three items with lower loadings were deleted from the final principal component analysis (e.g., "I would have more energy for my family and friends if I exercised regularly", "I would feel less stressed if I exercised regularly", and "Exercise helps me lose weight or maintain my current weight"). These final two factors explained 53% of variance in the final 10 items, which were used in the confirmatory factor analysis and invariance analysis (Table 6). Cronbach's coefficient alphas were calculated for each factor with values of .88 for the first factor (Pros) and .56 for the second factor (Cons). Step
Step II: Exploratory Analysis. To investigate the measurement structure of the Arabic version of the Self Efficacy scale for exercise, the sample was randomly split in two sections. First sample (N=330) was used for exploratory item analysis using principal components analysis, and the second sample (N=346) for confirmatory analysis using structural equation modeling. The initial 11 items (6 original items & 5 additional items) from the Self Efficacy scale were analyzed using principal component analysis with orthogonal varimax rotation (N=330). As expected, there was one factor that explained 59% of variance in these 11 items (Table 8 gives means and Standard deviation of these 11 items). The goal was to attain a shorter version (6 items) of the Self Efficacy scale, therefore, five items with lower loadings were deleted and a second principal component analysis was conducted. There was one factor that explained 64% of variance in these final 6 items, which were then used in the confirmatory factor analysis and invariance analysis (Table 9). Cronbach's coefficient alpha was .88.
Step III: Final Confirmatory Factor Analysis was conducted on the final 6 items from the Arabic version of the Self Efficacy scale (N=346). The one-factor model ( Figure 2) provided an excellent fit for the data, 2 (8)  in precontemplation and contemplation reported significantly lower Pros of regular exercise than those in preparation, action, and maintenance. In addition, participants in contemplation and preparation reported significantly higher Cons of regular exercise than those in maintenance. Moreover, individuals in maintenance reported significantly higher self efficacy of regular exercise then those in precontemplation, contemplation, preparation, and action. Similarly, participants in action reported significantly higher self efficacy of regular exercise then those in precontemplation, contemplation, and preparation. Finally, Individuals in precontemplation reported significantly lower self efficacy of regular exercise then those in contemplation and preparation ( Figure 3). Scale means for the Pros, Cons, and Self Efficacy are shown in Table 11.

CONCLUSION
The purpose of this study was three-fold: (a) to translate the Stages of change, the Decisional Balance, and the Self Efficacy scales for exercise into Arabic language, (b) to assess the factorial invariance and the reliability of the Arabic versions of the Decisional Balance and Self Efficacy scales for exercise in Saudi sample, and (c) to examine the expected patterns of relationships between the Decisional Balance, the Self Efficacy and the stage of change groups in this population.
The overall psychometric properties of these scales revealed that they were reliable and valid instruments that were invariant across subgroups varying in gender, age, health status, educational level, employment status, BMI, and stage of change.
The establishment of factorial invariance indicated that these constructs were being measured similarly across these demographic subgroups.
Decisional Balance. This study replicated the two-factor (pros and cons) uncorrelated measurement structure for the Decisional Balance scale in this Saudi sample, consistent with prior results (Almohammadi et al., under review manuscript;Blaney et al.,2012;Geller et al., 2012;Nigg et al., 1998;Paxton et al., 2008) showing that the pros and cons were orthogonal. Also, the scales showed good internal slightly higher than .01. This indicates that there might be some small differences in the factor model within these subgroups. These differences may due to sample fluctuation, but future investigation is needed to determine the source of these differences. However, this violation appears minor since the overall fits of these models were still very good (e.g. CFI and RMSEA). Therefore, strong factorial invariance should not be rejected. The results indicate that there is a consistent relationship between the two subscales (Pros and Cons), and the ten items that measure these factors.
Self Efficacy. This study confirmed the one-factor model for the Self Efficacy scale for exercise in this sample, replicating the underlying structure found in previous studies (Almohammadi et al., under review manuscript;Benisovich et al, 1998;Blaney et al.,2012;Geller et al., 2012;Paxton et al., 2008 indicates that there might be some small differences in the factor model within these subgroups. These differences may due to sample fluctuation, but future examination is needed to pinpoint the cause of these differences. Again, this violation appears minor since the overall fits of these models were still very good (e.g. CFI and RMSEA).
Therefore, strong factorial invariance should not be rejected.
Overall, the results suggest that participants in different subgroups did not respond differently to the Decisional Balance and Self Efficacy scales items. This consistency in the Measurement model is essential to valid research and effective interventions especially with population-based sample where variation is inevitable.
As expected, the results also found that Decisional Balance varied across stage of change groups, and the overall η 2 of .24 could be interpreted as a large multivariate effect size (Cohen, 1992). Participants in the preparation, action, and maintenance stages endorsed the Pros of exercising more highly compared to those in precontemplation and contemplation, with ω 2 of .08 representing a medium effect of stage of change. The Cons of exercising were rated as less important by participants in maintenance compared to those in contemplation and preparation, with ω 2 of .02 representing a small effect of stage of change. Although the magnitude of the Cons stage effect was small (ω 2 = .02), it was not surprising since all cons items used in this study had relatively low saturations (.40 to .54), and this pattern was observed in a previous study (e.g. Blaney et al, 2012). Further investigation into the costs of increasing regular exercise in this population is needed to lead to better measure of cons of exercise.
The overall patterns for Pros and Cons across the stages of change were consistent with the theoretical predictions of the TTM predictions and previous literature (Hall & Rossi, 2008;Prochaska et al, 1994), supporting the external validity of this exercise Decisional Balance instrument.
Similarly, Self Efficacy varied across stage of change. As predicted, self efficacy increased gradually across stages (Rossi & Redding, 2001;Velicer et al, 1990). Participants' confidence to engage in regular exercise was lower in the earlier stages of change and increased as individuals progressed to the later stages. These results are consistent with TTM predictions and replicated previous studies Blaney et al, 2012;Sarkin et al, 2001), supporting the external validity of this exercise Self Efficacy instrument.
Lastly, the observed findings should be interpreted in light of limitations. One limitation of this study is that this was a cross-sectional sample; future research is needed to examine the stability of these measures in samples over time. Further, a nonclinical, population-based sample was used in this study; scales should undergo additional validation to be utilized with individuals with illnesses related to insufficient physical activity (e.g. heart diseases, pre-diabetes and diabetes, and obesity). Also, the new additional items from the Arabic versions of the Decisional Balance and Self Efficacy scales were only examined within this Saudi sample; future research is needed to examine cross-cultural invariance of these new items. Another limitation was that the majority of participants were from the western region of Saudi Arabia; future research would benefit from a more diverse sample of Saudi adults to help rule out regional differences in exercise behavior. Finally, the generalizability of the measurement properties of Decisional Balance and Self Efficacy instruments is limited to the adult population from which the validation sample was drawn.
To conclude, the Transtheoretical Model of behavior change has proven to be effective across multiple behaviors, including exercise and physical activity. Stages of change, Decisional Balance, and Self Efficacy are key constructs within this strong framework, and investigators utilize these measures in TTM-based tailored interventions to promote and increase exercise. This study supported the underlying structure, internal consistency reliability, external validity, and measurement invariance of these measures in a population-based Saudi sample.  ‬ 13,11,9,7,5 ‫السلبيات‬ ,) (12,10,6,4,2)   (1) All numbers were retrieved from the website of the General Authority for Statistics: Kingdom of Saudi Arabia at https://www.stats.gov.sa/en.
(2) is defined as any bodily movement produced by skeletal muscles that causes acceleration of breathing and heartbeats such as running, brisk walking, cycling, swimming, and any sport activities like football, handball, and basketball...etc. Such activity should be performed at least 5 times per week for 30 minutes per session.
(3) Only 25.10% of Saudi males & 7.30% of Saudi females exercise for 150 minutes per week.