Generational differences in patterns of physical activities over time in the Canadian population: an age-period-cohort analysis

Abstract

Background: Using longitudinal panel data, the aim of this study was to examine the contribution of age, period, and cohort effects on changes in physical activity over time in a population-based sample of Canadians. We focused on three domains of physical activities: leisure time, commuting (i.e. walking and cycling), and daily activities (i.e. sedentary behavior). We also examined whether changes in sedentary behavior related to changes in participation in leisure time and commuting activities.

Methods: We used data from the Longitudinal National Population Health Survey (1994-2011): 10050 participants born between 1935 and 1984 grouped in five 10-years birth cohorts. We examined three outcomes: moderate-to-vigorous leisure time physical activity, active commuting, and sedentary behavior. We also included education, income, and body mass index as covariates. We used hierarchical age-period-cohort analysis to examine the contribution of age, period, and cohort effects to changes over time for each outcome.

Results: We found that recent cohorts were more likely to report sedentary behavior and greater participation in leisure time physical activities and active commuting. We also found a significant trend of increasing participation in active leisure time physical activity and active commuting among Canadians from 1994/95 to 2010/11 and, at the same time, an increase in sedentary behavior. The greater participation in leisure time physical activities and active commuting in each succeeding recent cohort was partially related to the secular trend of increasing participation in physical activities over time in the population. Furthermore, those with sedentary behavior were less likely to report participation in physical activities. Overall, obese individuals were less likely to be physically active and more likely to be sedentary, while the effect of socio-economic status varied by outcome.

Conclusions: The greater participation in physical activities (leisure time and commuting) in recent cohorts is encouraging and was substantially explained by period effects, which reflect broad social and environmental factors affecting the whole population. The large cohort effect of increasing sedentary behavior and the inverse relationship between sedentary behavior and physical activity is concerning, and identifies a target group for future interventions.

Keywords: Active commuting; Birth cohort; Leisure time physical activity; Period effect; Sedentary behavior.

Fig. 1

Age, Period, and Cohort effects on active leisure time physical activity and active commuting. Canadian National Population Health Survey, 1994–2011, Birth cohorts: 1980s = Born 1975–84; 1970s = Born 1965–74; 1960s = Born 1955–64; 1950s = Born 1945–55; 1940s = Born 1935–44. LTPA, leisure time physical activity. Values in A are predictions from model 1in Additional file 1 and values in B and C are predictions from model 2 in Additional file 1. Values in D are predictions from model 1in Additional file 2 and values in E and F are predictions from model 2 in Additional file 2. Each panel shows p-values for age, period, and/or cohort effects from the corresponding models

Fig. 2

Age, period, and cohort effects on sedentary behavior. Canadian National Population Health Survey, 1994–2011 Birth cohorts: 1980s = Born 1975–84; 1970s = Born 1965–74; 1960s = Born 1955–64; 1950s = Born 1945–55; 1940s = Born 1935–44. Values in A are predictions from model 1 and values in B and C are predictions from model 2 in Additional file 3. Each panel shows p-values for age, period, and/or cohort effects from the corresponding models

Fig. 3

Leisure time physical activity (active) and active commuting by sedentary behavior and birth cohort. Canadian National Population Health Survey, 1994–2011 Birth cohorts: 1980s = Born 1975–84; 1970s = Born 1965–74; 1960s = Born 1955–64; 1950s = Born 1945–55; 1940s = Born 1935–44. Predictions calculated with variables at their average from models in Additional file 4. Each panel shows p-values for the interaction between birth cohort and sedentary behavior