Iris Kanera
2 33 Prevalence and correlates of lifestyle behaviors Statistical analyses Analyses were conducted using SPSS 21. We used descriptive statistics to describe participant characteristics and the prevalence of health behaviors. For describing the adherence to separate recommendations, we constructed two categories ( yes, no ) for all five health behaviors. Missing values were handled according to the questionnaire manuals. For the EORTC QLQ-C30, HADS, and MAC the permitted number of missing values was one. For the SHORT SPSI-R, two missing values were permitted. The missing values were supplemented by using mean substitution, as recommended. Cases with missing values on days and time (PA), days, and number of servings (diet and alcohol) were removed from analysis. For other measures, less than 5% of the values were missing per value in a random pattern. We applied mean substitution for continuous covariates and for categorical covariates; we substituted the values of the modus. To assess the contribution of the distal and proximal factors in explaining alcohol, vegetable, and fruit consumption, and PA we conducted four sequential multiple linear regression analyses (Tabachnick & Fidell, 2001). The variables were entered in four entry steps based on the social cognitive models (e.g., Reasoned Action Approach, I-Change-Model), the theoretical framework of the present study (de Vries et al., 2003; Fishbein & Ajzen, 2010). The models prescribe an ordering of steps. This implies that socio-demographic and cancer- related factors were entered in order to control for their possible influence. Then, the psychological factors were entered in step 2 to evaluate what they add to the explanation of variance over and above the first set, the background variables. Subsequently, in step 3, the influence of attitude, social support, and self-efficacy was assessed above the two prior sets. Intention was added it in the last step, according to the assumptions of the social cognitive theories, that intention is influenced by the prior added proximal factors. To explore the correlates of smoking behavior (smoking vs quitting) among former smokers and current smokers, we conducted sequential logistic regression analysis (Tabachnick & Fidell, 2001). Never-smokers were excluded from this analysis. In the logistic regression analysis, we applied the same entry steps as described above. Results fromsequential logistic regression analysis ( N = 139) revealed large confidence intervals, due to the relative small number of participants and a large number of independent variables. Consequently, we conducted a second sequential logistic regression analysis, including fewer variables. The insignificant socio-demographic variables were removed, but core variables were entered in step 1 (age, gender, education level, type of cancer, and type of treatment). Significant psychological variables were added in entry step 2, such as the significant concepts from the EORTC QLQ-C30 (global health / QoL, cognitive functioning, social functioning, nausea / vomiting, insomnia, financial difficulties), and the subscales anxiety and depression from the HADS). In entry step 3 attitude, social support, and self-efficacy were added, and intention was added in the last step.
Made with FlippingBook
RkJQdWJsaXNoZXIy MTk4NDMw