In our initial examination of pedometer-determined physical activity among a predominately racial or ethnic minority sample of lower-income housing residents, we found that, on average, participants took just over 5300 steps each day, suggesting that much of our sample is currently sedentary. Tudor-Locke et al. (29
) reviewed 32 studies evaluating pedometer-determined physical activity, which led them to conclude that daily steps ranged from (a) 7,000–13,000 for healthy adults; (b) 6000–8500 for healthy older adults; and (c) 3500–5000 for sedentary individuals and those with disabilities or chronic illness. In our sample, 56% of participants fell below their sedentary index of 5000 steps per day. Further highlighting the high rates of sedentary behavior, only 24% of all participants took between 7,000 and 13,000 steps, and only 8% of participants over age 50 yr took between 6000 and 8500 steps. Consistent with previous findings, we also found significant sociodemographic variation by age, gender, BMI, and occupational status. Our findings extend previous research in a number of ways. First, they demonstrate the utility of pedometers to assess physical activity in population-based samples of individuals in lower-income settings, given appropriate methodological considerations. Particularly in population health research, where use of more objective measurement strategies may be unfeasible, our findings demonstrate that pedometers can be used effectively in traditionally understudied populations to assess physical activity in a manner that is largely devoid of the recall biases that can have an impact on self-report measures. Next, these data are useful to ensure sociodemographic representation in the ongoing efforts to compile expected reference categories for daily steps (29
). Finally, our findings would appear to support national self-reported survey data indicating a higher prevalence of physical inactivity among individuals in lower social class groups. This highlights the importance of directing additional research attention to the design of interventions to promote physical activity among this segment of the U.S. population.
Our findings are comparable to those of several recent studies that investigated pedometer-determined physical activity (13
). Tudor-Locke et al. (27
) found, among a sample of 209 residents of Sumter County, SC, that participants took a mean of 5931 steps per day; significant variation in steps was found for race (white vs nonwhite), age, education, income and BMI. Tudor-Locke’s sample was more heterogeneous than our own with respect to socioeconomic makeup, which may account for their finding of slightly higher daily steps. Our findings are slightly higher than those studies conducted among more racially or ethnically diverse samples, such as those of Whitt et al. (20
) and Henderson et al. (12
), who found daily steps of 4770 ± 5171 among a sample of 200 African American and American Indian women.
Consistent with previous self-report (6
) and pedometer data (20
), we found that women recorded significantly fewer steps per day than did men. Studies utilizing self-report measures have also often found racial or ethnic variation in physical activity patterns, particularly during leisure time (5
); we found no such racial or ethnic differences in our sample. Given the restricted socioeconomic range in low-income housing, it was not surprising that daily steps did not vary by either education or income. Occupational status may arguably be a more sensitive socioeconomic measure among those in this setting. At a minimum, employed participants would be expected to accumulate higher additional daily steps owing to their occupational activities and transportation to and from work. Indeed, we found that, compared with those who reported not working, individuals working in part-time jobs had more than 1450 additional daily steps, whereas those in full-time positions attained almost 3000 additional daily steps. The average number of daily steps among those who reported not working was well within the sedentary range (4526 ± 3435). These findings would appear to highlight the centrality of work to regular physical activity among individuals in lower-income settings. Rates of current employment in our sample, however, were low; only 33% reported current employment, and only 19% were employed in full-time positions. However, we cannot discount the possibility that unmeasured factors ormisclassification of income (given the narrowrange in this sample) may have had a more profound impact on daily steps among those healthy enough to work.
As might be expected, we found an inverse association between age and daily steps, such that participants over age 70 yr accumulated almost 3000 fewer daily steps compared with those under age 25 yr. In our formative research in this population, we found that many older-aged participants reported being largely socially isolated. With the exception of regular housing site programming, many reported having few opportunities for regular participation in physical activity. Given the extant data indicating health benefits of physical activity among older-aged individuals (2
), these data indicate the need (and tremendous potential opportunity) to implement physical activity promotion interventions for this group. Lower-income housing residents aged 50 yr and older will be a primary target for intervention efforts in the ODH study.
Similar to several previous investigations, we found a significant inverse association between BMI category and daily steps that became marginally associated in multivariable analyses. Obese participants took almost 1100 fewer steps than those of normal weight. Tudor-Locke et al. (24
) have proposed preliminary ranges of daily steps that are associated with BMI categories to identify individuals who might benefit from physical activity intervention. Based on these findings, we would call for some modification to those ranges. We concur with the previously established range for obese individuals (4600–6000 daily steps); however, integration of the current data suggests that the step ranges for normal-weight and overweight individuals should be adjusted to accommodate the lower daily steps taken among lower-income individuals. As others have suggested (27
), it is notable that relatively few individuals (17% in our sample) attained more than 9000 steps per day, a threshold associated with reduced obesity in this and previous investigations. Given the cross-sectional nature of our data, however, we are unable to establish causality of the association between BMI and daily steps; a bidirectional association almost certainly exists.
Those who reported not exercising took significantly fewer steps than those reporting exercise on 5 d or more of the week; however, the difference (1186 daily steps) was not as large as might have been expected and became nonsignificant in multivariable analysis. This finding suggests the limitations of relying exclusively on leisure-time estimates for measuring physical activity in lower-income populations. Future studies should investigate this finding using more comprehensive measures of exercise behaviors. Tudor-Locke et al. (23
) have proposed that daily steps in excess of 8000 may be roughly equivalent to the accumulation of 30 min of moderate-intensity activity on a single day. By that measure, only about 20% of our sample would meet current Centers for Disease Control and American College of Sports Medicine recommendations for daily physical activity (18
). Although this estimate is inherently biased (because pedometers are incapable of measuring activity intensity, a primary component of the national guidelines), nevertheless it may constitute a useful target for population-based physical activity promotion efforts.
Several considerations may limit interpretation of these findings. We chose to blind pedometers to study participants to prevent reactivity and to minimize the potential threats to sampling log validity presented by low literacy levels in the sample. Participants were instructed to wear the pedometer for 5 d, and we staggered participant start day; thus we have only one blinded 5-d iteration without a weekend day. Results from a recent validation study, however (26
), suggest that any 3 d (weekday or weekend) are sufficient for the reliable estimation of physical activity performed in a free-living week. We found sizeable site-specific variation in daily steps, and thus we estimated random effects models to control for these differences. We adopted a conservative analytic approach that resulted in the exclusion of 72 participants whose pedometer data were in some way compromised; however, sociodemographic characteristics of this group did not differ those participants included in it. A higher response rate would have been desirable, although it is clear that recruitment challenges experienced in the initial two housing sites (during the startup phase of the study) had a most negative impact on the estimate. Excluding the first two sites, the response rate improved dramatically (66.8% for the latter five sites combined). Allowing room for such improvement is an important consideration in the conduct of community-based research of this type. Finally, despite the some-what nonnormal distribution (skewness = 0.98, kurtosis = 0.56) of daily steps, we chose to report analyses in which the variable was modeled as normal for consistency with previous investigations and ease of interpretation. However, in bivariate analyses using the transformed outcome (using the square root daily steps), gender was not significant (P
= 0.06), and income was of significance to merit consideration for multivariable analysis (P
= 0.07). The statistical significance of all other variables was un-changed. In multivariable analysis with the transformed outcome, no change was seen in variable significance, and income became nonsignificant (P
Despite their promise, we recognize that pedometers have limitations as measures of total accumulated physical activity. Compared with accelerometers, pedometers are not designed to measure the intensity or duration of physical activity, and thus they are unable to determine concordance with national guidelines for regular physical activity. Nevertheless, given the challenges with self-report measures (particularly for lower-income groups) and the high costs and logistical considerations associated with accelerometers, pedometers may be useful in population-based studies of physical activity among lower-income and multiethnic audiences where the primary form of physical activity is walking.
This report constitutes the largest collection of pedometer-determined physical activity conducted in the United States to date. Our work extends the results of previous studies by suggesting that pedometers might be useful in the measurement of physical activity among a lower-income, multiethnic population.