Estimating physical activity from incomplete accelerometer data in field studies.

BACKGROUND

The purpose of this study was to develop a data-driven approach for analyzing incomplete accelerometer data from field-base studies.

METHODS

Multiple days of accelerometer data from the Stanford Girls health Enrichment Multi-site Studies (N = 294 African American girls) were summed across each minute of each day to produce a composite weekday and weekend day. Composite method estimates of physical activity were compared with those derived from methods typically described in the literature (comparison methods).

RESULTS

The composite method retained 99.7% and 100% of participants in weekday and weekend-day analysis, respectively, versus 84.7% to 94.2% and 28.6% to 99.0% for the comparison methods. Average wearing times for the composite method for weekday and weekend day were 99.6% and 98.6%, respectively, 91.7% to 93.9% and 82.3% to 95.4% for the comparison methods. Composite-method physical activity estimates were similar to comparison-methods estimates.

CONCLUSION

The composite method used more available accelerometer data than standard approaches, reducing the need to exclude periods within a day, entire days, and participants from analysis.