Sedentary behavior (Outcome variable)

Participants were asked to state the total time they usually spent (expressed in minutes per day) sitting or reclining including at work, at home, getting to and from places, or with friends (e.g., sitting at a desk, sitting with friends, travelling in car, bus, train, reading, playing cards or watching television). This did not include time spent sleeping. SB was assessed as a categorical variable [<8 or ≥8 hours per day (high SB)] [22].

Socio-demographic variables

These included age (18–44, 45–64, ≥65 years), sex, highest level of education achieved (completed secondary or less), wealth, marital status (married/cohabiting or else), living arrangement (alone: Y/N), setting (urban or rural), and employment status (engaged in paid work ≥2 days in last 7 days: Y/N), Wealth quintiles were created based on country-specific income.

Health behavior

These comprised of smoking (never, quit, current), alcohol consumption (never, non-heavy, heavy [23]), and physical activity. The validated Global Physical Activity Questionnaire was used to assess levels of physical activity [24]. Physical inactivity (i.e., not meeting the recommended guidelines) was defined as <150 min of moderate-to-vigorous physical activity in a typical week [25].

Physical health

A stadiometer was used to measure height, while weight was measured with a routinely calibrated electronic weighting scale. Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared, and was categorized as <18.5 (underweight), 18.5–24.9 (normal), 25.0–29.9 (overweight), and ≥30 (obese) kg/m². The total number of chronic physical conditions (angina, arthritis, asthma, stroke, diabetes, edentulism, cataract, chronic obstructive lung disease, hypertension, hearing problems) was calculated. The participant was considered to have the condition in the presence of self-reported diagnosis and/or symptom-based diagnosis using algorithms (blood pressure in case of hypertension), observation by interviewer (hearing problems), or self-reported conditions (edentulism: loss of all natural teeth). Detailed information on the variables on chronic conditions are provided in S1 and S2 Tables. Details on mobility and pain/discomfort are provided in the section on health status below.

Mental health

Variables pertaining to this domain included affect, sleep/energy, and cognition. Details on these variables are provided on the section on health status below.

General health

‘In general, how would you rate your health today?’ was the question used to assess self-rated health. Those who answered ‘bad’ or ‘very bad’ were considered to have poor self-rated health [26]. The 12-item validated version of the World Health Organization Disability Assessment Schedule 2.0 (WHODAS 2.0) was used to assess disability [27], and a scale ranging from 0 (no disability) to 10 (maximum disability) was created [28].

Health status (mobility, pain/discomfort, affect, sleep/energy, cognition)

Ten health-related questions pertaining to five health domains were used to assess health status: (a) mobility; (b) pain/discomfort; (c) affect; (d) sleep/energy; (e) cognition. Details of these variables are provided in previous publications [29, 30]. Briefly, a scale ranging from 0 to 10 was created with factor analysis with polychoric correlations for each domain based on two questions assessing past 30-day health function (actual questions are provided in S3 Table). Higher scores represented greater levels of impairment in health. The overall correlation coefficients between the two questions in each domain were: mobility (0.57); pain/discomfort (0.89); affect (0.82); sleep/energy (0.79); cognition (0.76).

Statistical analysis

The statistical analysis was done with Stata 14.1 (Stata Corp LP, College station, Texas). The selection of the 22 correlates of SB was based on past literature [13–15]. We assessed the association between the correlates (exposure) and SB (outcome) with multivariable logistic regression. First, we assessed the sociodemographic correlates of SB by constructing a model which includes all the sociodemographic variables (age, sex, education, wealth, marital status, living arrangement, setting, employment status). Next, we assessed the association between each of the other correlates with SB while adjusting for all the sociodemographic variables mentioned above. Analyses using the overall sample including all countries and country-wise analyses were done. Country adjustment was done in the analysis using the overall sample by including dummy variables for each country. All variables were included in the models as categorical variables with the exception of number of chronic conditions, mobility, pain/discomfort, affect, sleep/energy, cognition, and disability (continuous variables). The sample weighting and the complex study design were taken into account in all analyses. Results from the regression analyses are presented as odds ratios (ORs) with 95% confidence intervals (CIs). The level of statistical significance was set at P<0.05.

Sociodemographic SB correlates

Among the sociodemographic factors, being unemployed was an important sociodemographic correlate of high SB levels overall. One hypothesis that might partially explain this association is that those who are unemployed may also have several chronic physical and mental health conditions, which in our study were associated with higher levels of SB. It might be speculated as well that being employed increases levels of social connectedness, which may lead to less SB and more opportunities for leisure time physical activity [31]. Another consideration is that typical jobs in LMICs often involves manual labor, thus higher physical activity and less SB.

Apart from being unemployed, our study suggests that people living in urban environments were more likely to engage in high SB than those living in rural areas. It might be hypothesized that urban employment (e.g., service-based jobs) is more sedentary than rural employment (e.g., farming). Cities in LMICs are also often more likely to have higher traffic and crime which reduces activity and may promote more sedentary lifestyle. These factors are linked to more motorized transport use and higher risk for depression [32], which in turn, are also linked to SB. Evidence from prospective studies suggests that in cities, the built environment, which can be defined as the totality of places built or designed, including buildings, grounds around buildings, layout of communities, transportation infrastructure and parks and trails [33], is an important determinant of TV viewing time and other screen behaviors [34]. A 4-year follow-up study in Australia [35] identified that adults in low-walkable neighborhoods increased their TV viewing time compared with those who lived in high-walkable neighborhoods. Given that much of the evidence on the relationship of environmental attributes with SB comes from studies in Australia, Belgium, and the USA [34], the full range of these environmental exposures and their impacts on SB remain to be examined in LMICs [36]. Also, differences between countries in LMICs need to be explored in more detail in terms of its association with urbanicity as the association was particularly strong in South Africa and Mexico.

Older age was associated with higher levels of SB in Ghana, India, and Mexico which is consistent with previous data from the general population in Western countries [13]. Given the unfavorable biomarker profile associated with SB in older adults [37], our data add to the need to develop and test interventions to reduce SB in older age in LMICs.

Other socio-demographic factors showed mixed results depending on the country. For example, in contrast with Western studies [13], Indian and South African men were more likely to be highly sedentary. It might be that in these countries, particularly in more rural areas, women are more responsible for the livelihood of the family, thus, spending less time sedentary. Clearly, future research is required to clarify gender differences in SB in the LMIC context. Nonetheless, the present findings suggest that interrupting sedentary time may be especially important for men in India and South Africa.

Our data was in agreement with research conducted in Western countries [38], demonstrating that richer people were more likely to be highly sedentary in China and Ghana. In urban centers of LMICs, a more Western lifestyle may be evident such as the use of more motorized transport, less labor-demanding jobs, and physically undemanding, mostly screen-based leisure, which may account for the higher sedentary levels in richer individuals in these settings.

More research to clarify the between-country differences (e.g., age difference) is needed but it can be hypothesized that other country-specific factors, which were not accounted for in the current analyses, are important for explaining SB. For example, job characteristics, access to sports facilities, socio-cultural beliefs about being physically active, and the level of awareness about the benefits of avoiding SB following public health campaigns should be explored in more detail.

Health behavior correlates

In agreement with data from Western studies [13, 14], smoking and physical inactivity but not alcohol consumption was associated with being more sedentary. SB, physical inactivity and smoking may be reflecting a clustering of unhealthy lifestyles and behavior which is increasing in LMICs [39]. The observation that high sedentary levels are also associated with physical inactivity may especially be a relevant target for interventions, as recent findings suggest that the hazardous effects of SB may be reduced if people are highly active [22].

Health-related correlates

The current study suggests that SB is associated with a higher number of chronic conditions, which is in line with the wider high-income literature [13–15]. Given the inability to ascertain temporality with the SAGE cross-sectional design, it is plausible to suggest that multi-morbid people engage in more SB because of the associated mobility restrictions, pain/discomfort, or mental health burden (affect, sleep/energy problems) associated with chronic conditions. It is however equally plausible as well to suggest that prolonged SB may precipitate the development of chronic conditions, pain/discomfort, or mental health problems. Our data do suggest that interventions focusing on reducing SB should consider chronic conditions, pain/discomfort, and mental health problems. In LMICs, chronic pain conditions are leading causes of years lived with disability and a recent meta-analysis demonstrated that 35% of working adults had chronic pain [40]. In addition, in Western populations, it has been established that chronic pain is associated with higher SB levels, possibly due to psychological concerns about falling and lower balance confidence [41]. Previous research has also demonstrated that people with mental health problems may be more likely to have chronic pain [42, 43], which impacts upon mobility [44], and this may predispose the individual to sedentary behavior [41]. Previous research in Western populations has also demonstrated that SB may increase the risk of developing anxiety [6] and depression [5], possibly through increasing inflammatory markers [45]. Another hypothesis is that being sedentary and not engaging in activities may lead to social solitude and withdrawal from interpersonal relationships, both of which have been linked to increased feelings of social anxiety and depression [46]. Cognitive problems were another important mental health correlate of high SB. Cognitive problems are associated with impairments in executive functioning and this can result in an increased risk of falls, mainly in older people. Falls are associated with a fear of falling again and avoidance of activities [47].

There were also some country-specific differences. For example, underweight was associated with high SB only in South Africa. Low body weight may be an indicator of malnourishment or other serious health problems such as HIV, which is highly prevalent in South Africa, and is associated with being more sedentary [48]. Differences in associations between SB and mental and physical health parameters may also be explained by variations in the use of health care services. It is possible for example that differences in out-of-pocket payments may influence our findings to some extent. When out-of-pocket payments are high, people often delay or defer accessing or using services even if they believe to be in need. Out-of-pocket payments account for a large share of total health expenditure in LMICs [49]. In 2010, out-of-pocket expenditure as a percentage of total health expenditure was 35% in China, 28% in Ghana, 62% in India, 47% in Mexico, 36% in the Russian Federation, and 7% in South Africa [49]. Deferring treatment or receiving suboptimal treatment may mean that the underlying conditions leading to SB may be more severe in some countries.

Limitations and future research

Whilst our data has a number of strengths, it should be considered in the light of some limitations. First, the study is cross-sectional, therefore directionality cannot be deduced. Thus, future prospective research is required to explore the directionality and mediators of the relationships observed in our study. Longitudinal studies will also allow identification of true determinants of SB and clearly separate them from correlates. Second, SB was measured with a self-report questionnaire, which is known to be prone to bias and less accurate than objective assessments [50, 51]. It is well known that self-reported measures can underestimate sedentary levels [52]. For instance, previous self-reported data on SB among older adults in the general population found that high-income adults engaged in 5.3 hours of SB per day, while this figure increased to 9.4 hours per day when relying on objective measures [3]. In addition, since the survey was conducted between 2007 and 2010, it is possible that our results may not reflect the current situation in the countries included in the study. Finally, Given the large number of comparisons made, there is the possibility for Type I errors.

Conclusions

Our data provides some guidance that in order to reduce the burden of SB, health policy makers in LMICs should focus on unemployed people and those living in urban centers. Our data add additional, albeit cross-sectional evidence to previous concerns that urbanization, which may offer many opportunities in LMICs, including potentially better access to mental and physical health care, can also introduce new hazards such as a sedentary lifestyle and consequently a higher risk for NCD. Finally, national health policies focusing on reducing SB should consider a wide range of physical and mental health barriers.