Between January 2014 and February 2015, the cross-sectional Swiss National Nutrition Survey, menuCH, was conducted in ten centers across Switzerland [30, 31]. Two 24-h dietary recalls (24HDR) were conducted, along with one self-administered questionnaire that included sociodemographic, dietary, and lifestyle variables as well as details about PA [31]. The target sample of the menuCH study covered the seven major regions of Switzerland (Lake Geneva, Midlands, Northwest, Zurich, Eastern, Central, and Southern Switzerland), the three main language regions (CH-German, CH-French, CH-Italian), five age categories (18–29, 30–39, 40–49, 50–64, and 65–75 years old), and both sexes (male, female) and consisted of 4,627,878 residents. From the source study sample of 13,606 individuals 5,496 eligible residents were successfully contacted by mail and phone. A total of 2086 individuals agreed to schedule an interview. Among them, 2057 completed two 24HDR, while the remaining 29 completed only one. The 2057 individuals with two complete 24HDR were included in the current study [33].

Data on PA was collected using the short-form International Physical Activity Questionnaire (IPAQ). The IPAQ was developed by an International Consensus Group in the late 1990s. It is a standardized tool to measure PA levels through a self-reported questionnaire with 7 questions to measure the amount and intensity of PA as well as the sitting time over the last 7 days. The IPAQ assesses PA during leisure, occupational, domestic, and commuting time, but does not distinguish between the different domains. PA is measured using the Metabolic Equivalent of Task (MET), which is a unit to measure the amount of energy expenditure of PA (walking = 3.3 METs, moderate PA = 4.0 METs, vigorous PA = 8.0 METs) [34, 35]. In our study, the participants were classified into three levels (high, moderate, and low). Participants in the high and moderate levels meet the recommended level of PA according to WHO [20]. The category high included participants who met any of the following criteria: vigorous-intensity activity on at least 3 days achieving a minimum of at least 1500 MET-minutes/ week or 7 or more days of a combination of walking, moderate-intensity, or vigorous-intensity activities reaching a minimum of at least 3000 MET-minutes/week. The category moderate included participants meeting any of the following criteria: 3 or more days with at least 20 min of vigorous-intensity activity, 5 or more days of moderate-intensity or walking of at least 30 min per day, 5 or more days of any combination of walking, moderate-intensity, or vigorous-intensity activities accumulating a minimum of at least 600 MET-minutes/week. The category low included participant who did not meet the criteria for moderate or high PA levels [34].

The menuCH study included data on the participants’ lifestyle and sociodemographic factors, which were collected with a self-administered questionnaire [33]. The following lifestyle factors were included in our study: Body mass index [(BMI); calculated using the participants’ body weight and height and classified according to WHO definitions: underweight (BMI <18.5 kg/m²), normal weight (18.5 kg/m² ≤ BMI <25.0 kg/m²), overweight (25.0 kg/m² ≤ BMI <30.0 kg/m²), obese (BMI ≥30 kg/m²)], smoking habits (never, former, current), currently following a weight-loss diet (yes, no), and self-reported health status (good, medium-bad). The following sociodemographic variables were used in our study: sex (male, female), age group (divided in 4 categories: 18–29, 30–44, 45–59, and 60–75 years old), Swiss language region (CH-German, CH-French, CH-Italian), education level (primary, secondary, tertiary), and civil status (married, divorced, single, other).

Two 24HDR interviews were performed. The first interview was conducted face-to-face and 2 weeks later the second interview was conducted, which was done over the phone. Both interviews were performed by trained dietitians [36]. The food items reported in the interview were grouped in 19 categories using the trilingual Swiss version (0.2014.02.27) of the automated software GloboDiet^® (GD, formerly EPIC-Soft^®, IARC; Lyon, France [37, 38], adapted by the Federal Food Safety and Veterinary Office, Berne, Switzerland). The quantification of the consumed amounts was done with a book including 119 series of six graduated portion-size pictures and 60 actuals household measures, which were presented to the participants [39].

The menuCH participants were categorized into three alcohol consumption categories (none, moderate, heavy) based on recommended intake thresholds given by the Swiss Federal Office of Public Health [40]. A male participant was classified as a heavy drinker if his daily alcohol consumption exceeded 24 g of pure alcohol, and a female participant was classified as a heavy drinker if her daily alcohol consumption exceeded 12 g of pure alcohol. A male participant consuming 24 g or less of pure alcohol per day and a female participant consuming 12 g or less of pure alcohol per day were classified as moderate drinkers [41]. Non-alcohol-consumer were those participants who declared in the self-administered questionnaire to avoid drinking alcohol and did not report any alcohol consumption during either 24HDR.

The Alternative Healthy Eating Index (AHEI) score was calculated based on ten components (0–10 points per component), namely vegetables, fruits, whole grains, sugar sweetened beverages, nuts, meat, trans fat, long chain omega-3 fatty acids, polyunsaturated fatty acids, sodium, and alcohol. Higher AHEI scores indicate a higher diet quality [42]. The AHEI score is calculated for each 24HDR and the average of both AHEI scores was determined. The alcohol component was excluded from the AHEI score used in our analysis, as alcohol was considered as a separate exposure variable.

The menuCH study sample was weighted based on sex, age, marital status, major living region, nationality, household size, weekday and season of the 24HDR day to ensure having a representative sample of the population of Switzerland. The weighting strategy is described in detail in the open survey data repository [43].

Descriptive statistics [absolute numbers, percentages, median, and interquartile range (IQR)] were used to characterize the study population.

Multivariate imputation by chained equations (MICE) was used to impute missing values for PA level (n = 524), education level (n = 3), sitting time (n = 585), currently being on a diet (n = 4), self-reported health status (n = 4), civil status (n = 3), and smoking category (n = 4) [44]. Rubin’s rule was used to pool the results of the 25 imputed data sets [45].

Ordinal regression models were fitted to investigate the association between PA level and sociodemographic and lifestyle factors. The dependent variable, PA level, was defined as an ordinal outcome because it has a natural order (low < moderate < high). Therefore, an ordinal logistic regression approach was chosen, adjusted for sex, age group, Swiss-language region, education level, civil status, nationality, smoking, alcohol intake, BMI, health status, and currently on a diet. All confounders were chosen a priory.

Separate models were fitted for males and females to explore potential sex-based differences in the associations between PA levels and the sociodemographic and lifestyle factors.

To assess the goodness-of-fit of the model, the Hosmer-Lemeshow test and the Pulkstenis-Robinson test were conducted with each 25 imputed data sets. Subsequently, the median estimates were determined. The Hosmer-Lemeshow test groups the observations based on deciles of predicted probabilities for the outcome variables and compares the observed and expected frequencies within each group using a Chi-squared statistic. The Pulkstenis-Robinson test groups observations into homogenous groups based on predicted probabilities for the outcome variables and compares observed and expected frequencies within each group. Chi-squared test and deviations tests are used to provide p-values to assess model fit [46].

For the sensitivity analysis, the original ordinal regression model was additionally adjusted for income and/or sitting time as explanatory variables. The first model included income as additional explanatory variable, the second model sitting time, and the third model both, income and sitting time, to explore the effect on PA levels.

The R programming language (version 4.1.2) was used to conduct all analyses. The following R packages were used for weighting (stats), ordinal logistic regression (MASS), multivariate imputation by chained equations (mice) and goodness-of-fit tests (generalhoslem) [47]. The statistical significance level was set to 0.05 for all analyses.

Baseline characteristics of the menuCH participants stratified by PA level are shown in Table 1. The largest group was the high PA group with 827 (40.2%) individuals, followed by the moderate PA group with 487 (23.7%) individuals, and the smallest group was the low PA group with 219 (10.7%) individuals. 524 (25.5%) individuals did not provide complete information on PA and therefore the PA level could not be determined. 63.9% of all participants (participants with a moderate and high PA level) met the PA recommendations, which is a minimum of 150 min of moderate PA. Participants aged 30 to 59 were more likely to be in the low PA group, compared to the younger (aged 18–29) and older (aged 60–76) participants. German- and French-speaking participants were more likely to be in the low PA group compared to the Italian-speaking participants, as well as obese and overweight participants compared to the underweight and normal weight participants. No trend in alcohol intake, AHEI score, smoking status, civil status, or nationality among the PA levels was observed. Individuals with a low PA level had higher median daily sitting time (8 h) than individuals in the moderate (7 h) and high (5 h) PA group.

The association between PA and several sociodemographic and lifestyle factors based on an ordinal logistic regression model are shown in Table 2. The findings suggested an influence of the sociodemographic factors, as well as indicating an association with the lifestyle factors BMI and self-reported health status and PA level. Sociodemographic variables, such as age and education level emerged as notable correlates of PA level. Specifically, individuals aged 30–44 and 45–59 in comparison to those aged 18–29 had lower odds [OR = 0.53, 95% CI (0.37, 0.77) and 0.60 (0.41, 0.89)] of being in a higher PA level (compared to the lower PA levels). Similarly, individuals with tertiary education in comparison to individuals with a primary education had lower odds [0.50 (0.29, 0.86)] of being in a higher PA level (compared to the lower PA levels).

The other sociodemographic factors, namely sex, language region, nationality, and civil status, did not reveal evidence for an association with PA levels.

The two lifestyle factors that were significantly associated with PA levels, were BMI and self-reported health status. Obese individuals in comparison with normal weight individuals had lower odds [0.64 (0.45, 0.90)] of being in a higher PA level (compared to lower PA levels). Individuals reporting a medium to bad health status had lower odds [0.68 (0.50, 0.93)] of being in a higher PA level (compared to the lower PA levels) compared to individuals reporting a good health status. For the other lifestyle factors (smoking status, being on a diet, alcohol consumption, and AHEI score), no association with PA levels was observed.

To explore the sex-based differences in the association between PA and sociodemographic and lifestyle factors, the analysis was stratified by sex. The findings suggest that there is a difference between males and females. In the model including only females, we did not observe any statistically significant associations between sociodemographic and lifestyle factors and PA level (Supplementary Table SI 1). In the male-only model, the following associations are consistent with those in the main model: age 30–44 [0.38 (0.11, 0.69)] and 45–59 [0.43 (0.23, 0.79)] years, education [tertiary 0.45 (0.42, 0.95)], and self-reported health status [bad to medium 0.63 (0.42, 0.95)]. In contrast, BMI was no longer significantly associated with PA, but secondary education was associated [0.46 (0.21, 0.99)] (Supplementary Table SI 1).

The Hosmer-Lemeshow and the Pulkstenis-Robinsons tests were performed to assess the goodness-of-fit of the ordinal logistic regression model. The Hosmer-Lemeshow test produced a Chi-squared statistic of 15.285 (df = 7, p = 0.033), indicating a potential lack of fit. The Pulkstenis-Robinson test yielded a median Chi-squared test statistic of 3,598.6 (df = 3,528, p = 0.200) and the deviance test resulted in a median value of 3,485.8 (df = 3,528, p = 0.690). Both, the Chi-squared test and the deviance test, indicated no significant lack of fit.

We conducted a sensitivity analysis exploring the variables income and sitting time as additional explanatory variables in the regression analysis on the association of sociodemographic and lifestyle factors with PA levels (Supplementary Table SI 2–4). The regression model additionally adjusted for income (Supplementary Table SI 2) indicated that income was not statistically significantly associated with PA. The regression model additionally adjusted for sitting time (Supplementary Table SI 3) revealed that sitting time was negatively associated with PA level, while education was no longer significantly associated with PA level. Including both, income and sitting time, as additional explanatory variables in the regression model (Supplementary Table SI 4) confirmed that sitting time remained negatively associated with PA level, while income still was not.

The majority of the population of Switzerland met the WHO recommendations on PA. Sociodemographic and lifestyle factors, such as education level, age group, BMI group, and self-reported health status are related to an individual’s PA level. We observed that sedentary behavior is an important risk factor for lower PA levels. This is particularly concerning given the increasing sitting time observed in the population of Switzerland. Additionally, our findings highlight differences between males and females. Future studies should focus on sedentary behavior, with a clear distinction between occupational and leisure-time PA, as well as more detailed analysis to investigate the difference between sexes. The findings of our study might shape decisions of policymaker and thereby help to plan future preventive strategies against NCD.