These data represent the first known longitudinal pedometer data in adults. Previously, Raustorp and colleagues  reported pedometer tracking data for 97 Swedish adolescents assessed three times over five years, essentially capturing their development between 12 and 17 years of age. Pearson's correlations indicated low to moderate tracking in these adolescents, with patterns of higher tracking in boys than girls (i.e., correlations ranged from 0.55 in boys to 0.22 in girls). Most of the earlier tracking studies (children, adolescents, and adult) have relied on self-reported behaviors . Spearman correlations for adult men and women were 0.31 and 0.23, respectively, in a large British cohort assessed on self-reported frequency of leisure physical activity at 33 and 42 years of age . Spearman correlations ranged from -0.10 to 0.33 for self-reported time in physical activity assessed 7 years apart in a Canadian adult population . In the present study, we found that pedometer-determined physical activity behavior tracks to a relatively higher extent over one year in Australian adults disrupted by relocation, than both pedometer-assessed physical activity in adolescents and self-reports of physical activity in adults. Specifically, most of the correlations evaluated herein (both Pearson's and Spearman's) fit within Malina's  suggested ranges for moderate to moderately high tracking of physical activity behavior. The primary exception was in older females. This group was comparatively less stable in behavior over time, that is, characterized by a less predictable shifting in rank order between time points. Although this group was relatively smaller in sample size, it was also somewhat lower in mean steps/day in addition to representing a more vulnerable age group.
A higher correlation is anticipated with a briefer time span between assessments . Accordingly, the relatively higher correlations observed herein compared to the more prolonged duration between pedometer-based assessments of the Swedish adolescents  is somewhat anticipated. In contrast, however, Jackson et al.  reported correlations of 0.40 (Spearman) to 0.49 (Pearson) (again, generally lower than those we found) in 3–4 years old children assessed one year apart with accelerometers. Together, these findings suggest that physical activity behavior tracking may be more stable from year to year in adulthood, at least until older age groups (this last especially for women). Developing children and transitioning adolescents are exposed to continually changing personal, social, and physical environments, so some degree of instability in behavioral tracking is to be expected. The relative stability of these adults in their behavior, however, is especially interesting, given that the entire sample studied relocated between assessments.
The reasons that people relocate often differ as a function of their place in the lifecycle: younger and middle aged adults relocate for professional and personal opportunities while older adults often relocate for improved access to amenities and to be closer to family . The process of relocation in generally considered a stressful life event , especially if it requires long distance moves accompanied with inevitable adjustments to new environments, services, and supports . However, all of the participants in this study were already residents of Perth, Western Australia prior to relocating again within this community to different neighborhoods and locales; disruptions associated with long distance moves should therefore be considered minimal. Regardless the magnitude and scope of this now common life interruption, however, it is remarkable to note the overall stability of behavior (whether active or inactive) between time points in these adults. Overall, 25.9% of participants were stably inactive and 46.4% were stably active. From a health perspective, the higher proportion of those who are stably active is desirable, of course .
Exploring relative stability/change using a simple stratified analysis and Chi-square testing, we observed significant differential tracking patterns by sex, age, and BMI-defined weight categories. For example, when steps/day were categorized dichotomously and then evaluated for stability of such categorization over time, an age-related pattern was apparent for females, but not for males. However, it is also interesting to scrutinize the groups that changed their physical activity behavior over just one year. For example, of the ten specific sex-and-age strata studied, eight demonstrated a greater proportion of individuals who decreased their behavior compared to those who increased their behavior. This is an interesting finding worth pursuing in terms of confirmation and further exploration within other data sets; promoting maintenance of higher physical activity levels may be a separate and important approach to population health as opposed to focusing primarily on interventions directed at increasing behavior of sedentary individuals.
Overall, however, a tracking pattern of steps/day (expressed as a continuous variable) appeared to be most influenced by BMI-defined weight status. Further, relative stability of behavior (expressed as a categorical variable) appeared to be moderated by BMI-defined weight status for both sexes. To emphasize, normal weight individuals were more stable in their behavior (i.e., they maintained their position within the group) over time in contrast to those classified as obese. Scrutinizing Figure 2 further reveals that the proportion of obese individuals who increased their physical activity over the previous year was higher than those who decreased their behavior. Whether this phenomenon is merely regression to the mean (since those with higher BMI tend to take fewer steps/day) or a result of a wider spread personal decision to change behavior (perhaps driven by a motivation to influence weight) is a question worthy of future research.
A number of study limitations must be noted. Although this study of objectively monitored physical activity represents an improvement over self-report estimates, pedometers are not designed to detect intensity of movement. We are therefore unable to make firm conclusions about individual's participation in health-related quantities of at least moderate intensity physical activity, an outcome of public health interest . However, using the cut point of 7,500 steps/day is a defensible proxy threshold value for a healthful level of physical activity [16–18]. Further, pedometers "miss" or underestimate non-ambulatory activities like weight training and cycling, and because they are not worn during water activities, swimming is not detected at all . However, such activities account for only a small proportion of physical activity on a population level and the estimated average underestimation is approximately 300–700 steps/day . Another limitation of this study is that weight and height were self-reported; estimates of overweight and obesity may therefore be underestimated . Regardless of this potential bias, however, we were still able to observe clear moderating effects of BMI-defined weight categories on tracking of pedometer-determined physical activity. As is often the case for longitudinal studies, attrition is a threat to validity. Although study recidivism is evident, the data herein is reasonably representative of the originally recruited sample. These data, are of course, limited in their generalizability to populations most similar to Perth, Western Australia. Using similar pedometer brands, this sample (≅ 8,600 steps/day) was more similarly active compared to a previous independent Perth sample (≅ 9,500 steps/day)  than at least two U.S. samples: Colorado  (≅ 6,800 steps/day) and South Carolina (≅ 5,900 steps/day) .
In summary, in terms of pedometer-assessed physical activity, Western Australian adults held their rank position within groups to a moderate to moderately high extent (with few exceptions) between assessments separated by one year and disrupted by relocation. The observed pattern of change in steps/day was not influenced by sex or age, but was influenced by BMI-defined weight categories. Categorized and expressed as relative stability/change over time (i.e., stably inactive, decreased activity, increased activity, stably active), sex, age, and BMI patterns were evident. Of concern was the observation that individuals more frequently decreased (than increased) their physical activity over one year. Although not within the scope of this current study, it is possible to examine the additional demographic, personal, and environmental correlates of those categorized according to their relative stability or change between assessments points in future work.