National data indicate that most Americans do not engage in the recommended amount of regular physical activity (PA) [1, 2]. Low levels of PA have been shown to be associated with chronic disease, such as diabetes, cardiovascular disease, and certain cancers [3, 4]. Most of the conclusions in these studies were based on self-reported PA. However, with the advent of personalized motion detecting devices (e.g. pedometers, accelerometers), research has shifted towards re-examining PA levels using objective measures. Despite the fact that research consistently shows that African American adults engage in low levels of self-reported exercise [5, 6] and have low levels of physical fitness [7, 8], very little objective data exist on PA behaviors in African American adults [9, 10]. This lack of objectively measured habitual PA limits our ability to confirm findings [11–13] or illuminate the true relationship between PA and chronic disease in African American adults [14]. In this population, health outcomes such as diabetes, hypertension, cardiovascular diseases (e.g. coronary heart disease, strokes) and certain cancers are of paramount importance because they are considered principal sources of health disparities [15–19].

Several studies have objectively assessed PA in African American adults via pedometry [20–25]. The average daily step count in these studies has ranged from 4,355 [23] to 7,654 steps/day [24]. According to Tudor-Locke and Bassett [26], African American adults would be categorized as sedentary on average (i.e., taking < 5,000 steps/day) in two [23, 20] of the five referenced studies, and in none of the five would they be classified as active (i.e., taking ≥ 10,000 steps/day). Data in all but one [25] of these studies in African American adults were derived from a single assessment occasion (i.e., multiple days of monitoring over a single brief interval of time) spanning three days [21] to two weeks [23]. However, there are no data in the current literature to indicate that a single assessment occasion is sufficient to provide a reliable estimate of habitual PA in African American adults.

Assessment of PA by pedometry is reliable and accurate [27, 28]. However, Kang et al. [29] have suggested that numerous (and randomly selected) assessment occasions are necessary to establish a precise estimate of habitual PA, although these conclusions were based on modeling the number and patterns of single day values against a year-long average criterion. Specifically, these researchers concluded that a minimum of 30 consecutive days or 14 randomly selected days throughout the year were necessary to ensure valid (i.e., relative to the year-long average) and stable estimates of habitual PA [29]. However, longer and more complex monitoring periods are burdensome to both study participants and researchers, and therefore, shorter assessment periods are desirable in pragmatic research. Tracking of PA across multiple, short, assessment occasions may be a better approach to evaluating habitual PA compared to a single, extended assessment period. In two clinical studies [30, 31], a 3-day monitoring period (any combination of days in a week) provided stable estimates (Intraclass Correlation Coefficient (ICC) ≥ .80) of PA obtained over the course of the full week. However, the longer-term stability of a 3-day monitoring period is not known.

To date, no study has attempted to characterize the stability of habitual PA in an African American adult population. The Jackson Heart Study (JHS) [32], an epidemiological study of cardiovascular disease in African American adults, provides an ideal opportunity to accomplish this goal. The JHS enrolled 5,301 adults, of which a subset of participants wore pedometers for a 3-day monitoring period (consecutive days) on two to three distinct occasions, each separated by approximately one month [33]. The current analyses used data from the JHS to assess the stability of pedometer-determined estimates of PA. Conclusions will inform the appropriateness of using a single assessment occasion in future studies that examine relationships between PA and health outcomes in this population.

The median of the overall percent change in absolute values for those with two valid assessment occasions was 35.4 % (Q1: 20.9 %; Q4: 63.4 %) and ranged from 30.2 % to 44.5 % for participants differing in sex, age, and BMI status (Table 2). The median of the overall percent change in absolute values was 29.9 % (Q1: 20.6 %; Q4: 44.6 %) for participants with three valid assessment occasions, and ranged from 26.0 % to 31.6 % for participants differing in sex, age, and BMI status. All of these values were significantly different from 0 (p values < .001).

Figure 1 illustrates the quartile analysis. Approximately 85 % and 76 % of the participants with two and three valid assessment occasions, respectively, were at least “moderately stable” (remained in the same steps/day quartile or changed one quartile across time) in their activity. The proportion of participants who became more active was equal to the proportion who became less active for both two (p = .636) and three (p = .710) valid assessment occasions.

The analyses of the JHS DPASS data show that a single assessment occasion comprising a 3-day monitoring period of pedometer-determined steps/day provides estimates of PA that are reasonably stable across 4 – 6 months. This was illustrated by the moderate to moderately high ICCs, the fact that between 76 % and 85 % of participants were ‘moderately stable’ or ‘stable’ in their behavior, and that there were no significant differences in mean steps/day across assessment periods. These findings are consistent with those reported in a study of pedometer tracking of steps/day in Australian adults, which utilized a 7-day monitoring period and assessed stability in mean steps/day over one year [37]. In addition, the current study revealed that mean steps/day was reasonably stable within different sex, age, and BMI categories. Although confirmation is needed with larger and more diverse populations, our results suggest that, in the absence of interventions to increase PA, a single shorter (i.e., three days) and less burdensome assessment period can be confidently utilized to provide a reasonably stable estimate of habitual PA over a period of several months.

Our results extend previous reports demonstrating that a 3-day monitoring period accurately reflected activity over a week [30, 31] by showing that it provides a reasonably stable estimate of habitual PA over a period of 4 – 6 months. However, there were some differences between those who completed two versus three valid assessment occasions. First, steps/day was lower in participants with only two valid assessment occasions. It has been shown that individuals who are inactive tend to wear activity devices for a fewer number of days and for fewer hours per day compared to those who are active [38]. Second, the ICCs were generally lower in participants with two (overall ICC = 0.57) compared to three (overall ICC = 0.76) valid assessment occasions. Thus, between 57 % and 76 % of the total variability in steps/day is attributed to differences among occasions, with the remainder (about 25 % – 40 %) being attributed to natural shifts in PA patterns within the participants. This is consistent with the fact that the individual variability was within this range (overall percent change in absolute values ranged from 29.9 % to 35.4 %). The lower ICC values indicate that PA may not track as well in those completing two versus three valid assessment occasions. However, in those completing two valid assessment occasions, there were no significant differences in mean steps/day across time and a high proportion (>75 %) were at least ‘moderately stable’. These results suggest that mean steps/day provide an accurate depiction of participant’s personal habitual PA behaviors.

As noted above, previous studies of pedometer-determined steps/day have demonstrated that African Americans take an average of between 4,700 – 7,600 steps/day [20–25], whereas the current study provides slightly higher estimates (5,300 – 6,800 steps/day). Together, these estimates reveal that the average steps/day for African Americans are indicative of either a sedentary (< 5,000 steps/day) or low active lifestyle (5,000 – 7,499 steps/day) [39]. In fact, only 31.1 % of the participants in the current study averaged ≥ 7,000 steps/day, a threshold that has recently been shown to be indicative of accumulating 150 minutes/week of moderate to vigorous physical activity (MVPA) [40]. These JHS data are also consistent with other objective PA data [9, 10] (which show prevalence of regular MVPA ranging from 1 % – 38 %) and self-reported PA data [5, 6] (which show prevalence of regular MVPA ranging from 24 % – 40 %). The findings indicate that a limited number of African Americans are engaging in the recommended amounts of regular MVPA, and these inadequate levels of PA are placing them at increased risk for developing chronic disease. These multiple sources of PA data should serve as an impetus for developing theoretically sound interventions to promote regular PA in this population.

There are some limitations to the current study which should be taken into account when interpreting the findings. Only 270 of 480 (56 %) participants were included in the final analyses, and only 28 % (133 of 480) of the total DPASS sample provided data at all required DPASS assessment visits. Therefore, only a third of the participants fully complied with the study requirements. The compliant participants were more active and their activity was more stable than the rest of the participants, and therefore, the number of steps/day data may not be generalizable to the entire JHS sample. However, the data show that steps/day are reasonably stable across time regardless of whether participants had two or three assessment occasions. Furthermore, the mean steps/day for participants who completed one and two valid assessments were not significantly different from one another, thus the stability is likely similar between them. Another limitation is that although the use of pedometry provided an objective measure of activity, pedometers are not designed to detect intensity nor type of activity. In addition, the thresholds distinguishing stability vs. instability used herein were based on dividing the participants into steps/day quartiles. Ideally, the threshold would have been set at the level of PA consistent with the current recommendations of 150 minutes per week (>7,000 steps/day) [40], however, too few individuals met this criteria (31.1 %).

We used pedometers to track PA in African Americans enrolled in the JHS DPASS substudy. The findings are important because they suggest that a single 3-day monitoring period provides sufficient data for reliable estimation of PA over the course of 4 – 6 months in African Americans. In other words, researchers can choose any 3-day assessment occasion and can be confident that it represents habitual PA for this sample. These estimates appear to be robust to differences in sex, age, and BMI. The findings from the current study suggest that the JHS steps/day data are reasonably stable across time and accurately reflect habitual PA in this population. This is significant because it provides the basis for future studies to confidently assess the relationship between objectively measured habitual PA and health outcomes in the JHS, one of the most comprehensive epidemiological studies of cardiovascular disease in African American adults to date.