To our knowledge, this is the first study to identify distinct classes of long term trajectories of LTPA in a diverse sample of adults using latent class growth analysis. Conceptualizing physical activity involvement as a dynamic behavioural process rather than a static risk category, we used growth modeling to identify groups of individuals with vastly differing behavioural trajectories; this allowed us to explore meaningful contrasts and, ultimately, to identify predictors of trajectory class membership. Indeed, observational studies show that tracking of physical activity over the life span is moderate at best, underscoring the need for repeated assessments. [18–23] While cross-sectional designs that incorporate past experience can examine reported change in behaviours, recall of past physical activity involvement is generally poor. [24–26] Even prospective studies have generally employed simple methods to categorize individuals, for example by averaging repeated assessments of physical activity, or through modeling change in physical activity relative to some baseline value. Some prospective studies incorporate absolute values of change (or meaningful categories thereof), typically grouping individuals into only two [6, 28–30] or three [31, 32] categories. Trajectory analysis improves on these approaches by fully exploiting the longitudinal nature of the data, incorporating both the time and sequence of assessments to produce estimates. Although only three measures of LTPA were obtained in the current study, and fluctuations in yearly LTPA between assessment periods are unknown, each measure of LTPA has a reference period spanning one year, minimizing seasonal and other short term variations. Furthermore, the 22 year follow-up period provides a unique opportunity for examining patterns of activity over a substantial portion of the life span.
It has previously been reported that the PALS cohort is largely representative of the 1981 population, albeit PALS participants were more educated, and less likely to have very low income, to be underweight or to smoke, compared with those who did not participate in the PALS. The cohort retained for the current analysis was younger, healthier, and enjoyed more favorable social circumstances; thus, estimated trajectories describe the experience of a healthier, more advantaged sub-set of the population that was first surveyed in 1981. Nevertheless, even in this healthy sub-sample of the Canadian population, the majority of subjects were consistently inactive. At the last follow-up, 62 percent of adults were below recommended levels of LTPA (i.e. consistently inactive and decreasers), a proportion similar to findings from the British Birth Cohort, in which 60 percent of adults failed to meet recommended levels of LTPA at age 42 years. Clearly, increasing population levels of physical activity remain an important target for health promotion. In particular, because so few adults reverse behaviours that are acquired earlier in adulthood, programs should aim to help active youth remain active until early adulthood.
The impact of socio-economic factors on physical activity behaviours has been investigated in longitudinal studies,[18, 21, 33–36] but the extent to which social factors can influence long term trajectories of LTPA remains unclear. In the CARDIA study, higher average attained education was not associated with declines in physical activity, except in white women, while additional years of education during the seven years of follow-up were associated with declines in physical activity only among men. In the British Birth Cohort study, neither social class at birth nor educational attainment predicted change in physical activity over time. The British household panel survey observed that decreases in physical activity over time were predicted by lower grade occupations but not by income. In our study, a decreasing trajectory of LTPA was strongly predicted by both low family income, and lower education. These findings, albeit based on a limited number of studies, suggest that socio-economic disadvantage may be a particularly important risk factor for declines in physical activity among North American populations.