Measurement Considerations of Peak Cadence Measures using NHANES Data

Background: A recent technological advancement in accelerometers has enabled the researchers to examine the cadence (i.e., steps/minute) in a free-living environment, which can be used to quantify the intensity of walking activity (i.e., peak cadence). However, little is known about the optimal measurement conditions (i.e., the number of days and the inclusions of weekdays or weekend days) for peak cadence measures in observational study.

Purpose: This study examined the measurement conditions to obtain accurate peak cadence measures in a free-living setting.

Methods: The NHANES 05-06 which includes the accelerometry-based step-count data among a nationally representative sample of US population was analyzed. A total of 1,282 adults (>17 years) who have a valid accelerometer data (≥10 hours of wear time) for 7 complete days were included. The peak 1-min and 30-min cadence measures were extracted for each measurement day. A total of 17 simulated datasets were created using 3 conditions (‘no weekend day’, ‘1 weekend day’, and ‘2 weekend days’) across 1 through 6 monitoring days (1 monitoring day has only 2 conditions). For each dataset, all possible combinations of weekdays and weekend days were considered (e.g., with ‘1 weekend day’ across 3 monitoring days possible combinations include: [1] Sat, Mon, Tue; [2] Sat, Mon, Wed; [3] Sun, Mon, Tue, etc.). The complete 7-day measurements of peak cadences were used as criteria, and the mean absolute percentage error (MAPEs) were obtained using the random intercept linear models.

Results: Overall, the significant differences in average peak cadences between weekdays and weekend days were found (97.11 vs. 101.38 for 1-min; 67.15 vs. 72.87 for 30-min; p’s <.001). The MAPEs tend to be lower in ‘1 weekend day’ condition compared to ‘no weekend day’ and ‘2 weekend days’ across all monitoring days conditions for both peak 1-min and 30-min cadences. In addition, 2+ monitoring days for peak 1-min cadence, and 3+ monitoring days for peak 30-min cadence measures resulted in lower MAPEs (<10%) regardless of the inclusion of weekend days.

Conclusions: This study showed the empirical evidences of optimal measurement conditions to minimize the biases in peak cadence measures. The results suggested in this study could be applied in future research as reference measurement conditions for peak cadences.