Study sample & recruitment
In this study, data from the baseline measurement of the Be Active & Relax “Vitality in Practice” (VIP) project were used . This randomized controlled trial was set up to evaluate a physical activity and relaxation intervention aimed at improving NFR in office workers. In September 2011, all 1,182 office employees (> 18 years) of a single financial service provider received an invitation to participate in the project. Those on sick leave for more than four weeks were not eligible to participate. An off-line questionnaire was administered at baseline, including measures of NFR, daily physical activity, general health, mental health, sleep quality, stress and vitality. All data were self-reported. A total of 414 employees (response rate: 35%) from 19 departments signed the informed consent form, after which 412 employees completed the baseline questionnaire (providing data on NFR) and were included in the Be Active & Relax project. This study was approved by the Medical Ethics Committee of the VU University Medical Center Amsterdam. The development and design of the Be Active & Relax project has been described in full detail elsewhere .
In this study, need for recovery (NFR) was assessed using the Need for Recovery after Work scale of the Dutch version of the Questionnaire of the Experience and Evaluation of Work (Dutch abbreviation: VBBA). The used subscale of the VBBA, a questionnaire on the perception and judgement of work , consists of 11 dichotomous items (yes/no). These items represent short-term effects of a day at work, with questions like “I find it hard to relax at the end of a working day” and “When I get home, people should leave me alone for some time”, which were coded ‘0’ or ‘1’ in such a way that higher scores are related to more complaints. The NFR scale was computed by summing up the scores of the 11 items, of those providing data for at least 8 of the 11 items. The NFR total score was standardised to a score ranging from 0–100, based on the number of items with valid data. When three items are missing, the total score is expressed as percentage of the eight items, instead of the eleven items (the number of scored points was divided by the number of answered items and multiplied by 100). Higher scores indicate a higher NFR after work. Internal consistency of the scale was tested in this study and was found to be of good quality (α = 0.85), which is comparable to previous studies on NFR [28, 29]. Employees with percentage scores of 54.5 or higher (a cut-off point of six or more positive responses, as recommended by Broersen et al.) [30, 31] were considered to have a high NFR, as previous research showed that they have a higher risk for developing psychological complaints than people with a percentage score below 54.5 [28, 30, 32].
Overweight and obesity were investigated by calculating BMI as the body weight in Health Organization recommendations . Body weight and body height were assessed by self-report. BMI was categorized into three categories: normal body weight (cut-off point: <25 kg/m2), overweight (cut-off point: 25- < 30 kg/m2), and obesity (cut-off point: ≥30 kg/m2) . Those participants (1.9%) categorized as underweight (<18.5 kg/m2) were included in the normal body weight category because the number of participants did not allow a separate category (n = 6). For the same reason, participants categorized as extremely obese (≥ 35 kg/m2, n = 9), were included in the overweight category. Normal body weight was chosen as reference category in the analyses.
The following health measures were investigated: general health, mental health, stress, sleep quality and vitality.
General health and mental health were measured by items of the Dutch validated version of the Rand-36 measure of health-related quality of life . General health perceptions were measured by asking employees to give an indication on how they perceived their health on a 5-pointscale (bad, moderate, good, very good or excellent) and to indicate on four propositions (e.g. “I more easily fall ill than others”) to which extent on a 5-pointscale they “totally agreed” or “totally disagreed”. To assess mental health, employees were asked to indicate how often they had felt nervous, down, calm/relaxed, depressed/gloomy and tired, during the past four weeks. In this study, the Rand-36 measure of health-related quality of life has shown satisfactory internal consistency (α = 0.85) for the assessment of mental health and reasonable internal consistency (α = 0.73) for the assessment of general health. The internal consistency found in this study for mental health is comparable to other studies and for general health is slightly lower than previous research . Both the general and mental health scale were computed by summing up the scores of the 5 items (in those providing data for at least 3 of the 5 items). The general and mental health total scores are transformed to a 0–100 range (as percentage of maximum total score) and all items are averaged in the same scale together. When items are missing, the scale average is filled, assuming that the respondent would have answered this item in a similar way as the others. Certain items were transformed such that higher scores indicate a better health status.
To measure stress, the Dutch short form of the Perceived Stress Scale (PSS) was used . Participants were asked to indicate on a 5-pointscale (“never” to “very often”) how often they had had certain feelings during the last month (e.g. “In the last month, how often have you felt that you were unable to control the important things in your life?”). Items were coded in such a way that higher scores indicate a higher level of stress. PSS-4 is considered to be sound, but previous studies have found a rather low internal reliability (α = 0.60) . In this study, however, a satisfactory internal consistency was found (α = 0.70).
Sleep quality was assessed by the Dutch Jenkins Sleep Problems Scale . This scale contains four items, i.e., trouble falling asleep, trouble to continue sleeping, waking up feeling tired and worn out, and trouble staying awake during the day. Participants were asked to indicate how often they had experienced the four criteria mentioned above in the past month (‘0 days’ to’22-31 days’), coded in such a way that higher scores indicate a better sleep quality. In this study, a marginal satisfactory Cronbach’s alpha value of 0.61 was found, which is comparable to previously found internal consistency for the same construct .
Vitality was assessed using a part of the Utrecht Work Engagement Scale (UWES) , which contains 6 items (e.g. “At my work, I feel myself bursting with energy”) that had to be answered on a 7-pointscale (“never” to “always”). Answers were coded in a way that higher scores designate better vitality. In this study, vitality indicates a satisfactory internal consistency (α = 0.82). Previous studies reported comparable or even lower internal consistency (α = 0.68-0.80) [37, 38].
Sociodemographic variables, such as gender, marital status (married/cohabitating, in a relation, no cohabitating, single, divorced, widowed) and educational level were self-reported. Educational level was divided into lower education (no education, primary school, lower vocational education or lower secondary school), middle education (intermediate vocational education or intermediate/higher secondary education) and higher education (higher vocational education and university). Although all from one single service provider, a great diversity in job types was found among respondents. As the job types within the financial service provider are related to educational level, educational level was examined as a potential confounder as a proxy for job type/skill level. Age was calculated by extracting the self-reported date of birth from the date of completion of the questionnaire by participant.
Job demands were taken into account, as previous studies have identified job demands to be associated with NFR . This work-related variable was assessed using the items “work fast”, “work hard”, “no excessive work”, “enough time” and “conflicting demands”  (4-point scale from “strongly agree” to “strongly disagree”) which are part of the validated Dutch version of the Job Content Questionnaire (JCQ). An acceptable Cronbach’s alpha value was found in this study for job demands (α = 0.79). This is consistent with previous studies in different countries, including the Netherlands .
General health, described previously, was included as potential confounder in the relationship between overweight and obesity and NFR. Literature shows that obesity and general health are related [41–43]. Obese persons, as compared to their normal weight peers, seem to be more likely to have adverse health outcomes, such as poor general health [41, 43]. As general health and overweight/obesity seem to be associated, we aimed at investigating the independent association between overweight and obesity and NFR, by correcting for associations between general health and NFR.
Descriptive analyses were performed to summarize the characteristics of the population using means and standard deviations or percentages. For the outcome variable (NFR), a square root transformation was formally applied as its distribution was positively skewed, due to the high number of respondents scoring the minimum score of NFR 0–20 (37.7%). This percentage is consistent with other studies on NFR . However, the square root transformation did not meaningfully improve the distribution and therefore no transformation of original values was applied, which is in line with research of de Croon et al.. Univariate linear regression analyses were used to determine the associations with NFR, which was treated as a continuous outcome variable, and each independent variable.
Potential confounders were included in the adjusted analyses as confounders when the Beta coefficient of the independent variable changed at least 10% following addition of the potential confounder to the model. Furthermore, potential effect modification by age, gender and job demands was tested in the adjusted models. For each effect modifier, a linear regression model was fitted by crossing a predictor (overweight, obesity and health measures) and a modifier and adding this interaction term to the regression model. Thereby, the interaction between overweight/obesity and general health was examined by performing an analysis of variance (two-way ANOVA F-test). The level of significance was set at p < 0.05. Data were analysed using SPSS Version 20.0 (SPSS Inc., Chicago, IL, USA).