Cross-sectional analyses of between-stage differences of the TTM variables
Table 2 shows the scores of studied variables by the stages. The MANOVA for T1 scores resulted in a significant main effect for the stage of change (Wilks’s λ = 0.80, F (24, 3113.03) = 8.81, p < 0.01, η2 = 0.06). Follow-up ANOVAs resulted in significant effects of the stage of experiential processes (F (4, 1490) = 48.52, p < 0.01, η2 = 0.12), behavioral processes (F (4, 1490) = 48.90, p < 0.01, η2 = 0.12), pros (F (4, 1490) = 7.44, p < 0.01, η2 = 0.02), cons (F (4, 1490) = 7.26, p < 0.01, η2 = 0.02), and self-efficacy (F (4 F (4, 1490) = 31.27, p < 0.01, η2 = 0.08). Effect size estimates ranged from small to medium. Results of Tukey follow-up tests are summarized in Table 2.
Most participants in Action and Maintenance practiced effective stress management alone (66.5%, n = 494), at home (59.2%, n = 440), and through physical activity (89.5%, n = 665). With regard to perceived effectiveness, most participants felt that specific activity they engaged in to manage stress was effective (94.9%, n = 705).
Comparisons of demographics and baseline questionnaire results between participants who completed T2 assessment and those who did not
Among 1495 participants, 946 participants completed both T1 and T2 assessments and the remaining 549 participants dropped out. Table 3 indicates the differences in demographics and baseline questionnaire results between participants who completed T2 assessment and those who did not. When compared to the participants who dropped out, the retained participants were older, more likely to be married, less likely to be a student, and more likely to have a bachelor’s or higher degree, and scored significantly higher on experiential processes and behavioral processes.
The percentages of participants who practiced effective stress management alone, at home, and through physical activity were 66.1%, 60.3%, and 89.7% in the retained group and 67.2%, 57.1%, and 89.2% in the dropout group. A series of χ2 tests i ndicated that both groups had compatible percentages with regard to being alone or with others (χ2 (1) = 0.09, p = 0.77, φ2 = 0.00), at home or outside home (χ2 (1) = 0.71, p = 0.40, φ2 = 0.00), and through physical activity or not (χ2 (1) = 0.04, p = 0.84, φ2 = 0.00). With regard to efficacy, 95.7% of the participants in Action or Maintenance at T1 in the retained group and 93.4% in the dropout group considerd that specific activity they engaged in to manage stress was effective. No significant difference was found (χ2 (1) = 1.72, p = 0.19, φ2 = 0.00).
Stage transition patterns
Stage transition patterns over the study period are shown in Table 4. The stage distribution at T1 of 946 participants was as follows: 9.9% in Precontemplation (n = 94), 18.5% in Contemplation (n = 175), 20.4% in Preparation (n = 193), 28.6% in Action (n = 271), and 22.5% in Maintenance (n = 213). One month later, at T2, the stage distribution was as follows: 5.5% in Precontemplation (n = 52), 26.4% in Contemplation (n = 250), 20.2% in Preparation (n = 191), 27.1% in Action (n = 256), and 20.8% in Maintenance (n = 197).
Table 4 shows that 25 stage transition patterns were observed. Cases in which participants progressed from the pre-Action stages to Maintenance were checked. Eight participants who were in Precontemplation at the first survey, 15 who were in Contemplation, and 21 who were in Preparation showed these patterns. These patterns could not technically occur. Data of these 44 participants was excluded from the following analyses. Data of the remaining 902 participants was analyzed in later analyses. A series of t and χ2 tests were conducted to examine differences of demographics and questionnaire scores between the participants who were analyzed (n = 902) and those who were excluded (n = 44) (data not shown). The participants who were analyzed further were more likely to be male (χ2 (1) = 3.99, p = 0.046, φ2 = 0.00). Both groups of participants showed compatible scores on perceived stress (t (44.93) = 0.39, p = 0.70, d = 0.09), experiential processes (t (944) = 0.99, p = 0.32, d = 0.15), behavioral processes (t (944) = 1.55, p = 0.12, d = 0.24), and self-efficacy (t (944) = 0.82, p = 0.41, d = 0.13), as well as being fairly matched with regard to marital status (χ2 (1) = 2.09, p = 0.15, φ2 = 0.00), education (χ2 (1) = 0.04, p = 0.84, φ2 = 0.00), the student/worker ratio (χ2 (1) = 2.25, p = 0.13, φ2 = 0.00), and age (t (944) = 0.89, p = 0.37, d = 0.14). The participants who were analyzed further scored significantly higher on pros (t (944) = 2.02, p = 0.04, d = 0.31) and cons (t (944) = 2.08, p = 0.04, d = 0.32).
Processes, pros, cons, and self-efficacy as predictors of the stage transitions
Correlations between the studied variables are shown in Table 5. Six logistic regression analyses were conducted to identify predictors of the forward stage transitions from Precontemplation, Contemplation, and Preparation and those of the backward stage transitions from Contemplation, Preparation, and Action/Maintenance. Results of these analyses are summarized in Table 6. The results of omnibus tests were found to be significant for the overall regression models predicting the forward transitions from Precontemplation (χ2(1) = 7.15, p = 0.01) and the backward transitions from Action/Maintenance (χ2(1) = 8.25, p < 0.01). None of the model's variables was predictive of the forward transitions from Contemplation and Preparation and of the backward transitions from Contemplation and Preparation.
The first analysis focused on the forward stage transitions from Precontemplation. Data of 86 participants was analyzed. The forward transitions from Precontemplation was predicted by experiential processes (Exp(ß) = 1.05, 95% C.I. 1.01−1.09, Wald(1) = 6.43, p = 0.01). Behavioral processes, pros, cons, self-efficacy, and perceived stress were not entered as significant predictors. The participants who used experiential processes more frequently were more likely to progress to the later stages than remain at Precontemplation. The model’s accuracy of the classification was 66.7%.
The second analysis focused on the forward stage transitions from Contemplation. The data of 151 participants were analyzed. No variable was entered as a significant predictor. These variables included experiential processes, behavioral processes, pros, cons, self-efficacy, and perceived stress. The model’s accuracy in terms of classification was 55.6%.
The third analysis focused on the backward stage transitions from Contemplation. Data of 76 participants were analyzed, 67 and 9 of who showed stable and backward stage transitions. No variable was entered as a significant predictor. These variables included experiential processes, behavioral processes, pros, cons, self-efficacy, and perceived stress. The model’s accuracy in terms of classification was 88.2%.
The fourth analysis focused on the forward stage transitions from Preparation. Data of 104 participants was analyzed. No variable was entered as a significant predictor. These variables included experiential processes, behavioral processes, pros, cons, self-efficacy, and perceived stress. The model’s accuracy in terms of classification was 51.0%.
The fifth analysis focused on the backward stage transitions from Preparation. The data of 121 participants were analyzed. No variable was entered as a significant predictor. These variables included experiential processes, behavioral processes, pros, cons, self-efficacy, and perceived stress. When perceived stress was entered as a predictor, it was not a significant predictor (p = 0.06). The model’s accuracy in terms of classification was 56.2%.
The sixth analysis focused on the backward stage transitions from Action and Maintenance. The data of 484 participants were analyzed. The backward stage transitions from Action and Maintenance were predicted by self-efficacy (Exp(ß) = 0.93, 95% C.I. 0.89−0.98, Wald(1) = 8.00, p < 0.01). Experiential processes, behavioral processes, pros, cons, and perceived stress were not significant predictors. The participants with lower self-efficacy were more likely to show the backward stage transitions from Action and Maintenance. The model’s accuracy in terms of classification was 64.5%.