Elsevier

Preventive Medicine

Volume 40, Issue 3, March 2005, Pages 293-298
Preventive Medicine

How many days of pedometer monitoring predict weekly physical activity in adults?

https://doi.org/10.1016/j.ypmed.2004.06.003Get rights and content

Abstract

Background. The study purpose was to establish the number (and type) of days needed to estimate mean pedometer-determined steps/day in a field setting.

Methods. Seven days of data were collected from 90 participants (33 males, age = 49.1 ± 16.2 years, BMI = 27.2 ± 4.1 kg/m2; 57 females, age = 44.8 ± 16.9 years, BMI = 27.0 ± 5.9 kg/m2). Mean steps/day were computed for all 7 days (the criterion), each single day, and combinations of days. Analyses included repeated measures ANOVA, intra-class correlations (ICC), and regression.

Results. There was a significant difference (P < 0.001) between days. The difference was limited to Sunday and accounted for 5% of the variance. ICC analyses indicated a minimum of 3 days is necessary to achieve a reliability of 0.80. The adjusted R2 was 0.79 for a single day (specifically Wednesday), 0.89 for 2 days (Wednesday, Thursday), and 0.94 for 3 days (Wednesday, Thursday, Friday). Sunday was the last day to enter the model.

Conclusions. Although there is a statistical difference between days, there is little practical difference, and the primary distinction appears limited to Sunday. Although a single day of collection is not acceptable, any 3 days can provide a sufficient estimate.

Introduction

Although pedometers are increasingly being used for both measurement and motivation in physical activity programs and studies, objective monitoring using these simple and inexpensive instruments is still in its infancy. The validity of pedometers for physical activity measurement (specifically ambulatory activity) has been confirmed in several studies previously summarized [1]. However, researchers are just beginning to consider and assemble protocols for standardized assessment [2]. For instance, researchers are still striving to understand the sources of wide intra- and inter-individual variability (e.g., presented as a high standard deviation in relation to the mean steps/day) in pedometer-determined physical activity data and their impacts on both measurement and its interpretation. For example, a study of 365 days of pedometer monitoring in 23 individuals produced 8197 person days of data [3]. Intra-individual variability can be gleaned from the proportion of person-days ≤5000 steps/day (15.9%), between 5000–9999 (55.6%) and ≥10,000 steps (28.5%). Day-to-day (i.e., intra-individual) variability of pedometer-determined physical activity does not appear to be random and can be explained in part by day of the week, attendance at work, and participation in structured sport/exercise, all of which represent patterns of real-life fluctuations in behavior [3]. An example of wide inter-individual variability comes from a recent cross-sectional bi-ethnic study of 109 individuals. Participants took 7370 ± 3080 steps/day, equivalent to 52% inter-individual variability [4].

Researchers can learn from previous similar work undertaken in the study of nutrition assessment. The number of days of nutrition monitoring is related to an individual's day-to-day variability (i.e., intra-individual variability) and the degree to which individuals differ from one another (i.e., inter-individual variability). When intra-individual variability is large in relation to inter-individual variability, more days are required to estimate expected or usual mean values [5]. For example, the intra-individual variability for cholesterol intake (mg/100 g fat) in males is 49.8% vs. 9.4% for inter-individual variability [6]. This translates to 13 days required to determine group annual intake and 85–139 days for individual annual intake [5].

Relatively large intra-individual variability is a feature of pedometer-determined physical activity, important enough to deserve rigorous study and further examination. However, determining an appropriate monitoring frame (i.e., the number of days necessary to obtain a stable, and therefore, reliable measure of physical activity) is a pressing issue for both researchers and front-line practitioners who need to consider surveillance and screening costs in addition to response burden [2]. To date, monitoring frames used to quantify pedometer-determined physical activity in free-living populations range from 1 day [7] to 365 days [3]. Further, since pedometer-determined physical activity collected on weekdays and weekend days differs significantly [3], we also need to consider the appropriateness of type of day within the monitoring frame.

Trost et al. [8] determined that 4–5 days of monitoring by uniaxial accelerometer was necessary in children and 8–9 days in adolescents to achieve an intra-class correlation (ICC) reliability level of 0.80. The measurement output studied was daily time spent in moderate–vigorous physical activity, which is not directly comparable to steps/day [2]. Vincent and Pangrazi [9] reported that 3–4 days of pedometer monitoring are required of elementary school children to confidentially estimate their habitual activity (defined as steps/day over 8 days of data) at an ICC of 0.70, and 5 days are required to obtain an ICC of 0.80. Little previous research exists to inform the selection of such monitoring parameters in adults. Gretebeck and Montoye [10] are frequently cited to support a monitoring frame of 5–6 days (including weekend days) of adult pedometer data collection with less than 5% error. That study was conducted with a small (n = 30), select sample of young adult males recruited for their varied physical activity lifestyles. An uncommon pedometer brand was used and no validity data were offered. Empirical studies indicate that pedometer accuracy varies significantly between brands [11], [12]. Finally, Gretebeck and Montoye [10] also relied primarily on the Spearman–Brown Prophesy statistical technique to establish their minimal monitoring frame. This approach is not considered appropriate for multiple days of data collection [13]. The Spearman–Brown uses a split half method and is only appropriate if an intra-class correlation ICC cannot be computed (i.e., if there is only 1 day of collection) [13]. Data sets containing multiple days of pedometer data are appropriate for analyses using ICC.

Although this is a complex area of research, the present study shall focus on a single version of the research question: how many days (and what type) are enough to reliably estimate adult physical activity performed in a single free-living week?

Section snippets

Methods

Another article [14] has been written using this data set. Study details (summarized below) were reported in the initial manuscript that focused on the process of collecting self-monitored pedometer data by mail [14]. The present study examines the number of days required to obtain adequate estimates of mean weekly pedometer steps/day and is therefore novel. The Institutional Review Board for Research at the University of South Carolina approved the study.

Results

The study participants averaged 6838 ± 3643 steps/day over the 7-day monitoring frame, ranging from a low of 5766 steps on Sunday to a high of 7234 steps on Tuesday. Table 1 presents the means ± SD steps/day and 95% confidence interval (CI) for each day of the week by sex and for the whole sample. There were no gender differences in steps/day for any day of the week or for the weekly average of steps/day. The CoVw was 32.7% and the CoVb was 53.3%.

The assumption of sphericity was violated so the

Discussion

Using an assortment of statistical procedures to describe intra-individual variability and address the identified research question, we can conclude that a minimum of 3 days of pedometer data is sufficient to estimate free-living adult pedometer-determined physical activity in a week. The minimal length of monitoring frame has important study design ramifications, obviously to guide assessment protocols, reduce surveillance/screening costs and respondent burden, and enhance participant

Acknowledgements

These data were collected at the University of South Carolina Prevention Research Center, Arnold School of Public Health, University of South Carolina. This project was supported under a cooperative agreement from the Centers for Disease Control and Prevention through the Association of Schools of Public Health, Grant number U36/CCU300430-20. The pedometer ancillary study was attached to an environmental correlates of physical activity study that was funded by the U.S. Centers for Disease

References (30)

  • S.D. Vincent et al.

    Does reactivity exist in children when measuring activity levels with pedometers?

    Pediatr. Exerc. Sci

    (2002)
  • R.J. Gretebeck et al.

    Variability of some objective measures of physical activity

    Med. Sci. Sports Exerc

    (1992)
  • S.C. Crouter et al.

    Validity of 10 electronic pedometers for measuring steps, distance, and energy cost

    Med. Sci. Sports Exerc

    (2003)
  • P.L. Schneider et al.

    Accuracy and reliability of 10 pedometers for measuring steps over a 400-m walk

    Med. Sci. Sports Exerc

    (2003)
  • R.A. Charter

    It is time to bury the Spearman–Brown “Prophecy” formula for some common applications

    Educ. Psychol. Meas

    (2001)
  • Cited by (363)

    View all citing articles on Scopus
    View full text