For the purpose of the study, a 5-arm clustered Randomized Controlled Trial was conducted, which was registered at the Dutch Trial Register (NTR2297) and approved by the Medical Ethics Committee of Atrium–Orbis–Zuyd (MEC 10-N-36).
Study design
The intervention groups and the waiting list control group (who received no intervention until the end of the study period) were studied in a clustered randomized control trial (RCT). There were evaluation assessments (i.e. questionnaires) at the start (T0: also the basis for the first and second tailored advice), 3 months after baseline (T1: also the basis for the third tailored advice), 6 months after baseline (T2) and 12 months after baseline (T3) [9]. For the purpose of this study only the T0 and T3 measure were used.
Participants and procedures
Participants (Dutch speaking adults aged over fifty) were recruited via direct mailing in communities of the Municipal Health Council regions (MHC; N = 6) participating in this project. Communities were matched on their urbanity, percentage of people with a low SES, percentage of people with a high SES, percentage of immigrants, and the percentage of people aged over 50. The regions were randomly assigned to one of the five research arms: (1) basic printed intervention; (2) environmental printed intervention; (3) basic Web-based intervention; (4) environmental Web-based intervention; or (5) control group. Participants had to be over 50 years of age (no maximum age), and need to have sufficient understanding of the Dutch language. No other in- or exclusion criteria were set. Each MHC provided a list of addresses of a random sample of eligible participants living in the selected matched communities. Figure 1 provides an overview of the number of invitations that were distributed in order to reach an even number of participants aged over 50 at baseline for each condition, and of the number of participants at the enrolment and participation stages. A power calculation (effect size = 0.4, power = 80%, intracluster correlation coefficient = .1) showed that at the baseline (T0) about 420 participants were needed for each intervention condition to include a minimum of 250 participants per research condition at the 12-month (T3) assessment, bearing in mind an expected dropout rate of 40% during the 1-year follow-up based on a previous project [13]. Participants were included from November 2010 until March 2011 [10]. Invitations for the printed intervention contained an information letter, a questionnaire, a prepaid return envelope and a form for giving informed consent. Invitations to the Web-based intervention contained a similar information letter, additional information about how to fill in a Web-based questionnaire and a personal username and password to log on to the Active Plus website. For the online intervention group, informed consent was administered online.
Intervention
The Active Plus intervention is a systematically developed computer tailored, theory- and evidence-based intervention to stimulate PA among people aged over fifty [8, 9]. The intervention aimed to influence awareness, initiation and maintenance of PA by targeting pre-motivational constructs (i.e. awareness, knowledge), motivational constructs (i.e. attitude, self-efficacy, social influence, intrinsic motivation and intention) and post-motivational constructs (i.e. commitment, strategic planning, self-regulation skills, action planning and coping planning) [8].
Intervention participants received advice on three occasions, tailored to the answers they gave in previous assessments [8, 9]: (1) after the baseline assessment (immediate advice for the Web-based intervention and within two weeks for the printed intervention); (2) two months after the baseline assessment; and (3) within four months after baseline assessment feedback on progress was given based on the second assessment (again, immediate advice for the Web-based intervention and within two weeks for the printed intervention, after completing the second questionnaire). Participants in the Web-based intervention received the invitations for follow-up assessment by email, including a link to the Web-based questionnaire. Participants in the printed intervention received the invitation for follow-up assessment by mail, together with the follow-up questionnaire and a prepaid return envelope.
The intervention provides general information about the benefits of PA especially for older adults, such as the importance of physical activity to maintain a healthy cognitive state. The specific content of the basic tailored intervention advice depended on the participants’ personal characteristics (e.g. age, gender and educational level) and psychosocial characteristics (e.g. pre-motivational, motivational and post-motivational constructs as described above), and PA behaviour, and the extent to which they were planning to change their behaviour (i.e. every participant received all the tailored information, but the participants PA behaviour and stage of change determined in which advice the information was provided [8]). For example, contemplators received feedback in the first advice on their PA behaviour to raise awareness of their own physical inactivity, followed by feedback on their intention to become more physically active and on their perceived pros, cons and self-efficacy (with special attention to the most important PA barriers in this age group). Participants were stimulated using role model stories and to write down their own intrinsic PA motivation. In the second advice, contemplators received advice about their perceived social support, they received practical information about physical activity possibilities and they were encouraged to formulate action plans. The third tailored advice provided feedback about the progress in behaviour and determinant scores in the previous months [8]. The intervention with additional environmental components contained the same information as the basic tailored intervention, with additional tailored advice on local possibilities and initiatives for being physically active (e.g. walking or cycling routes in their own neighbourhood) [8]. The content of the Web-based and printed intervention were identical; however, interactive components were integrated into the Web-based version such as the use of videos instead of pictures, and Google neighbourhood maps in stead of printed neighbourhood maps [9].
Participants in the Web-based intervention received, in addition to the Web-based advice, an email with a copy (pdf-format) of their personal advice that was in the same format as the printed version. The tailored advice contained between five and eleven pages of text and illustrations, depending on (changes in) PA behaviour and determinant scores.
The different intervention conditions result in different implementation costs as well. Due to the lower reach of the Web-based intervention than of the printed intervention (12% vs. 19% response), recruitment costs for the Web-based intervention are higher than for the printed intervention. On the other hand, Web-based interventions result in lower handling, printing and postage costs. Furthermore, the provision of additional environmental information results in higher costs.
Measures
Physical activity behaviour
Total weekly days of sufficient PA and minutes of moderate to vigorous PA were assessed using the validated self-administered Dutch Short Questionnaire to Assess Health Enhancing Physical Activity (SQUASH) [14]. Total weekly days of sufficient PA was measured by a single item: ‘On how many days per week are you, in total, moderately physically active by undertaking, for example, brisk walking, cycling, chores, gardening, sports, or other physical activities for at least 30 minutes?’. Total weekly minutes of moderate to vigorous PA was calculated by multiplying the frequency (how many days per week), and duration (how many hours and minutes per day) of leisure and transport walking, leisure and transport cycling, sports, gardening, household chores and odd jobs performed with moderate or vigorous intensity. The reproducibility (rspearman = 0.58; 95% CI = 0.36–0.74) and relative validity (rspearman = 0.45; 95% CI = 0.17–0.66) of the SQUASH are reasonable for the general adult population [14]. A study of Wagenmakers et al. showed that using the SQUASH in an older population can be considered as a fairly reliable tool and that the validity was comparable to those of other questionnaires [15].
Moderators
Age, gender (0 = men; 1 = women), educational level, BMI, intention to be sufficiently physically active and the presence of a chronic physical limitation (0 = absent; 1 = present) were selected as potential moderators of the intervention effect, and were all assessed in the baseline questionnaire. Educational level was categorised as ‘low’ (primary, basic vocational or lower general school) or ‘high’ (higher general secondary education, preparatory academic education, medium vocational school, higher vocational school or university), according to the Dutch educational system. BMI was calculated by dividing self-reported weight by height in metres squared. Participants were classified as being underweight (BMI < 18.5 kg/m2), healthy weight (BMI 18.5–24.9 kg/m2) or overweight (BMI > 25 kg/m2). The participants’ intention to be sufficiently physically active was assessed using three items on a 10-point scale (e.g. ‘To what degree do you intend to be sufficiently physically active?’: 1 = absolutely not; 10 = absolutely sure). The average sum score of the three variables was used in the analyses (Cronbach’s α = 0.938).
Other variables were also assessed, but are not described here as they are not related to the current focus of this study.
Statistical analyses
Baseline and dropout characteristics
One-way analyses of variance (ANOVAs) and Chi-square tests were conducted to test for baseline differences in participant characteristics (i.e. age, gender, BMI, educational level, intention to be sufficiently physically active and the presence of chronic physical limitation) among the five research conditions. These characteristics are considered as covariates in the remaining analyses. Hierarchical logistic regression analyses were performed to study whether these baseline characteristics were predictors of dropout at the 12 month measurement, and whether these predictors differed between the research conditions. Analyses were performed using SPSS for Windows (Version 18).
Effect on physical activity behaviour
Participants were nested within neighbourhoods resulting in the probability of interdependence between them. To account for this interdependence, multilevel linear regression analyses were performed with a random intercept for two levels (1: individual; 2: neighbourhood). Each 12 month outcome (i.e. total weekly days of sufficient PA and weekly minutes of moderate to vigorous PA) was first regressed onto the Active Plus intervention in general (i.e. the four intervention conditions together, in contrast to the control group), and secondly regressed onto the dummies for the different research conditions independently (with the control group as a reference case), its baseline values and the covariates (gender, age, education, intervention type, BMI, intention, having a chronic limitation). The analyses were repeated with different intervention conditions as a reference case, to study the difference between the intervention conditions. Furthermore, it was studied whether the intervention effect was moderated by the aforementioned participant characteristics. This moderation effect was studied by adding an interaction term to the model between these participant characteristics and the dummies for the intervention conditions.
By regressing the outcome values against their baseline values, the amount of change in the variables is represented independently of their baseline value. This method is preferred to using absolute change scores because groups with lower levels are more likely to increase their levels by chance than groups with higher levels [16]. Analyses were applied to the total dataset, including missing data. Applying multilevel analyses to an incomplete dataset has been shown to give more accurate estimations than applying imputation methods [17]. Respondents were excluded from the analyses when they reported being physically active for more than 6,720 minutes per week since being physically active for 7 days per week, over 16 hours per day was considered to be impossible [14].
To obtain a better interpretation of the intervention effects and a better comparison with other tailored interventions, Cohen’s d effect sizes (ESs) were calculated for each intervention condition compared to the control group. ESs were defined as the mean differences in effect between the intervention conditions and the control group (corrected for baseline PA) divided by the pooled standard deviations for those means, in which d = 0.15, 0.20, and 0.25 are considered for small, medium and large effects [18].
Additionally, all analyses were repeated with the inclusion of all participants who filled in the baseline measurement, using the multiple imputation method (based on the participants’ age, gender, baseline intention, and PA behaviour at T0, T1, T2 and T3; 5 times imputed) for participants who did not fill in the 12 months questionnaire. Several studies have shown that the multiple imputation method is preferable over single-imputation methods, such as last observation carried forward [19].