Globally, many adults and children do insufficient physical activity to maintain good health.1 Furthermore, the population burden of inactivity is unacceptably high.2 Although strategies to increase physical activity are being developed,3, 4 effect sizes are usually small to moderate, and effective interventions are not widely applied. The prevalence of physical activity is slow to improve and is worsening in some countries.5 As the global burden of non-communicable diseases increases, risk factors such as physical inactivity become relevant in low-income and middle-income countries, not just in the most developed nations.6 Understanding the causes of physical activity behaviour is essential for development and improvement of public health interventions,7 much as aetiological studies of disease provide information about treatments. Of particular interest is how aetiological factors differ between physical activity domains—ie, areas of life in which activity is done (at home, at work, in transport, and in leisure time)—and with country, age, sex, ethnic origin, and socioeconomic status.
One challenge in the interpretation of evidence is that most studies have used cross-sectional designs. This so-called correlates research assesses only statistical association, rather than providing evidence of a causal relationship between factors and physical activity.8, 9 Longitudinal observational studies and experimental data could identify factors that have strong causal associations with physical activity.9 When such factors are identified in studies of aetiological design, they are described as determinants.8
Because physical activity is affected by diverse factors, behavioural theories and models are used to guide the selection of variables for study.8 Integration of ideas from several theories into an ecological model (including inter-relations between individuals and their social and physical environments) is now common.10 This approach uses a comprehensive framework to explain physical activity, proposing that determinants at all levels—individual, social, environmental, and policy—are contributors. A key principle is that knowledge about all types of influence can inform development of multilevel interventions to offer the best chance of success.10 Figure 1 shows a multilevel model of physical activity influences, which guided our classification of variables in this report. The model is ecological because inter-relations between individuals and their social and physical environments are included. A key principle is that understanding of all levels of influence can inform development of multilevel interventions that offer the best chance for success.10 Variables within individuals, such as psychological and biological factors, are widely studied, as are interpersonal variables. Environmental, policy, and global variables are less studied, but are thought to have widespread effects. The combination and interaction of factors and at these levels are expected to influence physical activity.
Key messages
- •
Population levels of physical activity participation are low, and improved understanding of why some people are active and others are not is needed
- •
Some consistent correlates of physical activity are individual-level factors such as age, sex, health status, self-efficacy, and previous physical activity
- •
Ecological models posit that the physical and social environments—ie, economic conditions, societal norms, urbanisation, industrialisation—are important determinants of physical activity
- •
Correlates have been less studied in low-income and middle-income countries than in other nations, and although broadly similar to those in high-income countries, they are more focused on the prevalent domains of physical activity in developing countries—ie, correlates of transport and occupational activity
- •
New research has identified genetics, evolutionary biology, and variation in physical activity behaviour throughout life as important determinants
- •
Improvement of the research base, with a stronger focus on determinants research (with improved causal inference rather than repetition of cross-sectional correlates studies) will further an understanding of physical activity in populations and interventions designed to increase activity levels
Physical activity is done for various reasons, and the SLOTH (sleep, leisure-time, occupation, transportation, and home-based activities) model11 delineates the domains of physical activity. Ecological models of physical activity have been developed that suggest correlates are specific to domains.12 All domains are important for understanding of worldwide physical activity, because frequency of activity in each domain varies greatly between countries.13, 14 For example, occupational, household, and transport domains are the most common types of physical activity in low-income and middle-income countries, whereas leisure-time activities contribute more to total physical activity in high-income countries than elsewhere.14
We have three objectives. First, we aim to summarise present knowledge about correlates and determinants of physical activity in adults and children, on the basis of evidence from systematic reviews of physical activity correlates.15, 16 We provide an outline of new research into physical activity domains, particularly exploring correlates of active leisure and recreation, and active transportation. Additionally, we describe the rapidly evolving field of environmental correlates of physical activity. Second, we examine correlates and determinants research in countries of low and middle income, where physical inactivity is rapidly becoming a major risk factor for non-communicable disease.17 Third, we analyse correlates and determinants of physical activity that are least studied, such as genetic factors, lifecourse trajectories, evolutionary and societal factors, and obesity (figure 1).