Eurobarometer

We used data from the Special Eurobarometer 412 “Sport and physical activity”, which was carried out in the 28 EU Member States in November and December 2013 by the market research company TNS opinion & social[17]. Eurobarometers are biannual, cross-sectional surveys conducted on behalf of the European Commission that cover approximately 1000 respondents per EU Member State. Participants were sampled using a multistage random sampling design, with multiple sampling points based on country specific population size and -density. Households were selected by starting at a random address and following a pre-defined random route. In each household, one respondent was selected based on the closest birthday to the date of the interview. Data was obtained from people aged 15 years and over by the Eurobarometer, but in our research we only used data from people aged 18 years and over. Participants were interviewed face-to-face at home in their mother tongue using a standardized protocol. A total of 27,919 Europeans were surveyed, ranging from 500 in the Republic of Cyprus and Malta to 1600 in Germany. The average response rate was 46 percent, ranging from 11 percent in Luxembourg to 80 percent in the Netherlands. The response rates per country are presented in Fig 1. The European Commission approved the study protocols and informed consent was obtained from all participants. The information was anonymized and de-identified prior to analysis.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. The response rate of the Eurobarometer survey per country.

https://doi.org/10.1371/journal.pone.0149320.g001

Variables

Statistical analysis

No sampling weights were applied in the analyses. We excluded participants aged <18 years because our focus is on adults, as well as participants who answered ‘don’t know’ on the sitting question since this is our main outcome. For all other variables, the response options ‘refusal’ and ‘don’t know’ were indicated as missing values. Sample characteristics and sitting prevalence were analysed with descriptive statistics. Univariate odds ratios (ORs) of sitting >7.5 hours per day were assessed by country and socio-demographic variables using binary logistic regression analyses with dummy variables where appropriate. For the ordinal variables, we also assessed the overall trends.

We wanted to compare each country with all other countries, instead of choosing one country as a reference. Therefore, the multivariate ORs were assessed in 29 separate logistic regression analyses; 28 analyses to assess the ORs of the 28 countries, and one additional analysis to assess the ORs of all socio-demographic variables. A country-specific variable was computed for each country to distinguish people living in that country from people living in all other countries. To assess the OR of each country, the analysis included the appropriate country-specific variable and all socio-demographic variables. To assess the ORs of all socio-demographic characteristics, the final analysis included all country-specific variables as well as all socio-demographic variables.

Because of co-linearity between the education and occupation variables (i.e. the group ‘students’ is identified in both variables) we constructed and reported two multivariate models: one including ‘Age when stopped full-time education’ and excluding ‘Current occupation’, and vice versa.

In addition, we stratified the analyses by country to study the socio-demographic correlates of sitting within the different countries. Because these single-country analyses had reduced statistical power, we only included the variables ‘Age’, ‘Gender’, ‘Age when stopped education’ and ‘Current occupation’ in these analyses. Because of the aforementioned co-linearity, we constructed the same two multivariate models. In order to compare the stratified analyses with similar overall results, we also conducted a general analysis including all country-specific variables and the variables ‘Age’, ‘Gender’, ‘Age when stopped education’ and ‘Current occupation’.

All data analyses were conducted in SPSS, version 22. We present ORs with 95% confidence intervals (95% CI). In addition, statistical significance of p<0.05 and p<0.001 is indicated.

Prevalence of sitting time

Fig 2 shows the distribution of sitting time across the original response options. The highest percentage (15.4 percent) can be found at 3.5–4.5 hours. The shape of the curve is roughly the same across all countries (data not shown). Overall, the respondents reported a median of 300 sitting minutes per day (interquartile range: 180–420), ranging from a median of 180 minutes in Portugal to 360 in Denmark and the Netherlands. In addition, 18.5 percent reported sitting 7.5 hours or more per day, ranging from 8.9 percent in Spain to 32.1 percent in the Netherlands (Table 1). The distribution of these proportions across Europe is shown in Fig 3. The geographical pattern on this map shows generally lower sitting percentages in the southern countries and higher percentages of people reporting more than 7.5 hours per day in the north of Europe.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. The distribution (%) of sitting time across the original response options, for all countries combined.

https://doi.org/10.1371/journal.pone.0149320.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. The distribution of the proportion of European adults reporting sitting more than 7.5 hours per day across the 28 European Union Member States.

Designed by Showeet.com.

https://doi.org/10.1371/journal.pone.0149320.g003

Correlates of sitting time

The results of the uni- and multivariate analyses are shown in Table 2. The highest ORs of sitting more than 7.5 hours per day were found regarding the occupation variable. White collar workers (OR: 5.00 (95% CI: 4.31–5.80)), students (3.84 (3.03–4.86)), managers (3.65 (3.13–4.25)), people who are self-employed (2.01 (1.67–2.42)), retirees (1.51 (1.25–1.82)) and people who are unemployed (1.31 (1.06–1.62)) all showed significantly higher ORs compared to manual workers, while house persons showed a lower OR of 0.72 (0.54–0.97). Students were also identified as a high-sitting group in the education variable, with an OR of 2.35 (1.81–3.04) of sitting more than 7.5 hours per day compared to people who were 15 years or younger when they stopped full-time education, as well as people who were 20 years or older when they stopped education (1.51 (1.28–1.79)). The multivariate models additionally show consistently significantly higher ORs of sitting more than 7.5 hours per day for (in descending order of effect size): people who use the internet every day (ref: never/no access) and people who are not at all satisfied with their life (ref: very satisfied. Consistently significantly lower ORs were found for (also in descending order of effect size): people in the most active quartile (ref: least active quartile), people in the third activity quartile, people with three or more children (ref: no children), people living in a rural area or village (ref: large town), people in the second activity quartile, people who experience difficulties paying their bills from time to time (ref: never), people who experience difficulties paying their bills most of the time, women (ref: men) and inhabitants of a small/medium sized town.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Univariate and multivariate odds ratio (OR) of sitting more than 7.5 hours per day, by country and socio-demographic characteristics.

https://doi.org/10.1371/journal.pone.0149320.t002

Single-country analyses

The analyses stratified by country are shown in S1 Table and Tables A-BB in S1 File. In general, these analyses show that the above-mentioned associations are fairly consistent across the EU countries. This is most visible for the occupation variable. White collar workers show substantial as well as statistical significant higher ORs in all 28 countries and managers, students, self-employed, retired and unemployed show higher ORs in the majority of countries, although not all statistically significant. House persons were shown to have a higher OR in half of the countries while the other half showed lower ORs, with no obvious geographic pattern.

In comparison to people who were up to 15 years when they stopped full-time education, people who were 20 years or older when they stopped full-time education and people whose current occupation was “still studying” showed higher ORs in almost all countries, with approximately one-third of the countries showing a significant association. Most countries indicated a lower OR for people who were between 16 and 19 years old when they stopped education; the countries indicating a higher OR for this group were mainly countries in the south-east of Europe.

Approximately two thirds of the countries showed a lower OR for people aged 18–24 years as compared to people who were 25–34 years old. The countries indicating a higher OR for this group were mainly situated in the south-eastern part of Europe. In addition, two-thirds of the countries also showed lower OR for people aged 55–64 years, with no obvious geographic pattern. The countries showed a 50/50 distribution for the other age groups. The above-mentioned lower ORs for women were also visible in approximately two-thirds of the countries. The countries indicating the opposite were mainly southern European countries.

Prevalence of sitting time

Respondents to the 2013 Eurobarometer indicated a median of 300 sitting minutes per day, which is the same as the 2005 Eurobarometer sample[20] and a worldwide 20-country comparison conducted by Bauman and colleagues.[23] However, when Milton and colleagues[18] recently studied the trends in self-reported sitting time based on the 2002, 2005 and 2013 Eurobarometer data, they concluded that the prevalence of sitting more than 7.5 hours per day had decreased between 2002 and 2013. This unexpected result highlights the importance of regular monitoring of SB.

When comparing the self-reported results of this Eurobarometer with objectively measured population levels of sitting time, the Eurobarometer seems to underestimate actual sitting time, even when taking possible temporal changes into account. For example, the United Kingdom respondents of the Eurobarometer reported a median of 300 sitting minutes, while the accelerometer data of the 2008 Health Survey of England showed an average of 584 daily sitting minutes for women and 595 for men [24]. A similar, but smaller, difference can be seen in Sweden, where accelerometer data showed a mean of 459 sitting minutes per day, while Swedish Eurobarometer respondents reported a median of 300 minutes in 2001.[25] The difference is even more remarkable in Portugal, with 180 median minutes based on the Eurobarometer, against a range of 529 to 612 average sitting minutes based on accelerometer data from 2006–2008.[26] Moreover, a Norwegian validation study in a national sample concluded that people on average report two hours less sitting time with the IPAQ-short compared to the ActiGraph accelerometer.[27] These results suggest that not only the absolute numbers, but also the relative order of countries might be dependent on the instrument used, and that underreporting may be more prevalent in some countries than in others. This observation shows that more research in this field is warranted.

We observed a north-south division across Europe, with southern countries reporting less sitting and northern countries sitting more, even when adjusted for socio-demographic variables. Notable exceptions to this observation are Ireland (with 10.4 percent of the respondents reporting sitting more than 7.5 hours per day), Croatia and the Czech Republic (22.7 and 26.5 percent, respectively). The previous Eurobarometer studies reported a similar geographical pattern.[19,20] These findings could reflect real differences between countries, possibly caused by differences in wealth, culture or climate. In this light, it should be noted that all three Eurobarometer surveys were conducted from October to December, a time in which people in northern Europe might be more inclined to stay in, use motorised transportation and replace PA with SB. Alternatively, the observed differences could also be caused by differences in recall- and/or social desirability bias between countries. For example, people living in countries with more public health campaigns against sitting might be more inclined to underreport their sitting time. As there is no literature to support these hypotheses, more research is needed to investigate the differences in sitting time across Europe.

Correlates of sitting time

Our multivariate analyses showed that women have a lower OR of sitting more than 7.5 hours per day than men. These findings are in line with the previous Eurobarometer studies.[19,20] However, the International Prevalence Study (IPS)[23] found no gender differences in the adjusted models and a review by Rhodes et al.[13] reported no association for gender and general sitting behaviour in the majority of studies, although men were shown to sit more in two studies.

With regard to age, one of our multivariate analyses showed that people aged 55–64 years and people aged 65+ years have a lower OR of reporting sitting more than 7.5 hours per day than people aged 25–34 years. This is in line with previous Eurobarometer studies[19,20] and IPS[23], who all reported that younger people sat more. However, the same model also shows that people aged 18–24 have a lower OR than people aged 25–34. Since our univariate analysis shows a (non-significant) higher OR for people aged 18–24 years, a possible explanation for our multivariate results might be that a large proportion of this group are studying/students, which could be more strongly related to sitting than age itself. Notably, Rhodes et al.[13] reported ‘extremely mixed’ results for age and sitting time, suggesting no clear cut association.

One of the more novel socio-demographic variables in our analysis was marital status. The other studies[19,20,23] did not include this variable in their analysis, and Rhodes et al.[13] concluded more evidence is needed. However, we did not find any associations in our multivariate models. People who are married/with a partner showed a significant lower OR in the univariate model, but not in the multivariate models.

People with higher education levels are consistently identified as more likely to be high-sitters.[19,20,23] This is possibly related to a higher prevalence of sedentary occupations in higher educated people. Occupation was found to be the strongest correlate of self-reported sitting time. All occupations have higher ORs than manual workers, except for house persons. The review of Rhodes et al.[13] confirms that unemployed and retired people usually sit more, but did not report on the other groups. Our findings suggests that sedentary occupations are important predictors of high overall sitting time. This highlights the importance to target SB at the workplace. A recent systematic literature review and meta-analysis by Neuhaus et al. concluded that so-called activity permissive workstations (e.g. standing desks, treadmill desks, etc) can be effective in decreasing occupational SB, without affecting work performance.[28]

Our findings regarding physical activity, type of community and number of children are in line with previous research.[13,19,20,23] With regard to the children, Rhodes et al.[13] concluded that “the presence of children is associated with less SB.” However, our results seem to suggest this association is only present in families with three or more children, which might suggest having more children is associated with more daily responsibilities and less opportunities to sit down.

No associations were found between television and computer ownership and sitting time, and car owners show a significantly lower OR of sitting more than 7.5 hours per day compared to people without a car in one of the models. These findings are counter-intuitive, since time driving in a car, watching television and using a computer is usually spent sitting down. However, this might have been a measurement issue, as respondents were asked whether or not they owned a television, computer and car, and not how much time they spent using it. This hypotheses is strengthened by the observation that habitual internet users did show higher ORs, when compared to people who never use the internet or have no access.

We found that people who reported having difficulties paying bills “from time to time” have lower ORs than people who “almost never/never” experience these difficulties. In addition, people who experience these difficulties “most of the time” show even lower ORs. Since this variable is related to the socio-economic status of the respondents, it is not surprising these results are in line with the education and occupation variables.

Finally, people with lower life satisfaction have higher ORs of sitting more than 7.5 hours per day, showing a dose-response relationship. This association could go both ways; people with high levels of sitting might be less satisfied with their life; and people with low life satisfaction might be more prone to sitting. This association could also be explained by the health status of the respondent, e.g. people with illness/depression are often more sedentary and less satisfied with their life. This is supported by a review by Teychenne et al.[29], who concluded that the available evidence suggests that SB is associated with a higher risk of depression.

Country specific results

The analyses stratified by country showed that the associations between type of occupation, and to a lesser extent, educational level, and sitting time are similar across European countries, while this was less clear for the age and gender variables. This finding suggests the need to target individuals with sedentary occupations in interventions aiming to reduce sitting time throughout Europe.

In addition, for some socio-demographic groups there seemed to be different associations within south-eastern or eastern European countries in the single-country analyses. Although this is only a first impression, this might indicate that different groups engage in high sitting time in these countries, which could have implications for policy and need to be investigated further.

Strengths and limitations

The response rate and representativeness of the Eurobarometer are our first point of concern. The average response rate was 46 percent, ranging from 11 to 80 percent. Several procedures were used to maximize response rates, such as using experienced interviewers, reviewing the training materials and methods, and multiple call-backs on different days of the week and times of the day. The origin of the variation in response rates across countries is unclear. This response rate, together with the high mean age (52 years) of the sample, provide an indication that the study sample is not representative for the adult population at large. This might be especially true for the countries with the lowest response rates.

The main limitation of this Eurobarometer survey, however, lies with the assessment of sitting time and PA. The survey included an adapted version of the IPAQ-short,[21] a questionnaire that is widely used and has reasonable validity and reliability.[22,30] The developers of the Eurobarometer survey adapted the IPAQ-short so that respondents were asked to choose between several categorical response options, with a range of 60 (sitting time) or 30 (PA) minutes, as opposed to the original continuous reporting in hours and minutes. Unfortunately, the response categories for PA were adapted from the original IPAQ in such a way that it was impossible to accurately distinguish people that met PA recommendations from people that did not. In addition, it prevented us from accurately estimating the time respondents spend on PA. For this reason, we used the midpoint of each category as a rough estimate of the time spend in PA, and constructed quartiles to compare these relative rather than absolute values.

Even though the IPAQ-short sitting question response categories were acceptable, the self-report nature of the sitting time assessment remains a limitation of this study. Self-reported sitting time is subject to recall and social desirability bias, which most likely results in underestimations of sitting time.[31] Furthermore, possible cultural and language differences in these biases and in the interpretation of the questionnaire might complicate comparison between countries. In addition, the question only asks about a ‘usual day’, which means it was not possible to assess differences in sitting time between week- and weekend days.

Even though these limitations should be taken into account in the interpretation of the findings, the Eurobarometer is currently the only source of comparable sitting time data across all 28 EU Member States, providing the opportunity to study and compare current levels of SB and its correlates.

Conclusions

There is substantial variation in the prevalence of sitting among European adults across countries and socio-demographic groups. Across Europe, a geographical pattern can be observed, with northern countries reporting more sitting and southern countries reporting less. Non-blue collar occupations and higher educational levels were found to be the strongest correlates of higher levels of sitting time. In addition, women, physically active people, people living in a rural area or in a small/medium sized town, people with three or more children, infrequent internet users, people with difficulties paying their bills, and people who are satisfied with their life report lower sitting levels. The observed associations between occupation, and to a lesser extent education, and sitting were reasonably consistent across Europe.

The Eurobarometer is currently the only source of comparable sitting data across Europe, but its measures of sitting time have been irregular across time and inconsistent between consecutive surveys. In addition, the self-report nature of the measurement of sitting used in the Eurobarometer surveys may introduce bias. To enable monitoring of population levels, identification of populations at risk and targeted actions for these populations, regular surveillance of population levels of SB is needed, using identical methodologies over time, and preferably including objective measures of SB. In addition, more research should be focussed on the correlates and determinants of SB since they are mainly unknown. The results of this study provide entry points for targeted interventions aimed at high-sitting populations, such as people with sedentary occupations.