Iris Kanera

166 Chapter 7 Measurements During online data collection, completing the questions was mandatory and resulted in a dataset without missing data due to question skipping. This was a strength of the studies. To assess PA, we used the validated IPAQ Short (Craig et al., 2003) in Chapter 2 and the validated SQUASH (Wendel-Vos et al., 2003) in Chapter 5 and Chapter 6. Measuring the specific items on the SQUASH (an instrument collecting more details than the IPAQ Short) was necessary for generating personalized tailored advice within the intervention and to provide insight into the KNW effects on several PA outcomes, which was a strength of this thesis. Furthermore, to assess dietary behavior and smoking, relevant items of the validated Dutch Standard Questionnaire on Food Consumption for the general Dutch population (Bogers et al., 2004b; van den Brink et al., 2005) and standard questions form the Dutch Measuring Instruments for Research on Smoking and Smoking Cessation were used (Mudde et al., 2006). Smoking cessation was assessed by using the seven-day point prevalence abstinence that was reported as a stable measure to predict long-term abstinence (Hughes et al., 2003; Velicer & Prochaska, 2004). In contrast, other studies among cancer survivors evaluated only quit attempts (quit smoking at least one day within the past 12 months), which might be too tolerant to measure long-term smoking cessation (Emmons et al., 2013; Ramaswamy et al., 2016). However, the use of self-report measures may be regarded as a limitation. Over- and underestimation might have occurred due to social desirability or recall bias, in particular on PA and diet outcomes. Prior research indicated that both of the used outcome measures on PA mostly overestimated PA compared to objective measures among the general population (IPAQ and SQUASH) and among cancer survivors (IPAQ; Lee et al., 2011; Nicolaou et al., 2016; Ruiz-Casado et al., 2016). A relatively small degree of overestimationwas reported on dietary outcomes among the general population (van den Brink et al., 2005; van Lee et al., 2016). However, applying a self-administered questionnaire is the most feasible and least expensive method to assess lifestyle on a large scale, although the results may be less accurate compared to objectively observed results (Helmerhorst et al., 2012; Prince et al., 2008). Inaccuracy of instruments may have had an influence on the provided feedback (MRA) concerning the respondents’level of PA and diet behavior in comparison to the norms (Chapter 4). Importantly, since possible over- or underreporting was comparably present in the entire study sample, this issue did not affect the outcomes of the effect studies when comparing relative changes over time of self-reported outcomes of the intervention and control arms (Chapter 5 and Chapter 6). Further, a limitation of this project might be that we used validated self-report questionnaires without objective validation to measure outcome behaviors (Gerritsen et al., 2015). However, objective validation of behaviors was not feasible in this large-scale research project.