Number of daily measurements needed to estimate habitual step count levels using wrist-worn trackers and smartphones in 212,048 adults
Abstract Daily step count is a readily accessible physical activity measure inversely related to many important health outcomes. However, its day-to-day variability is not clear, especially when measured by recent mobile devices. This study investigates number of measurement days required to reliabl...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | article |
| Language: | EN |
| Published: |
Nature Portfolio
2021
|
| Subjects: | |
| Online Access: | https://doaj.org/article/036df5d67b42453ca666bd85d1d9f8e7 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract Daily step count is a readily accessible physical activity measure inversely related to many important health outcomes. However, its day-to-day variability is not clear, especially when measured by recent mobile devices. This study investigates number of measurement days required to reliably estimate the weekly and monthly levels of daily step count in adults using wrist-worn fitness trackers and smartphones. Data were from a 5-month physical activity program in Singapore. The 5-month period was divided into 22 weekly and 5 monthly time windows. For each time window, we leveraged data sampling procedures and estimated the minimum number of measurement days needed to achieve reliable mean daily step count with intraclass correlation coefficients (ICC) above 80%. The ICCs were derived using linear mixed effect models. We examined both simple random and random consecutive measurement days and conducted subgroup analysis by participant characteristics and tracking devices. Analysis of weekly and monthly step count included 212,048 and 112,865 adults, respectively. Fewer simple random measurement days are needed than random consecutive days for weekly time windows (mean 2.5, SD 0.5 vs mean 2.7, SD 0.5; p-value = 0.025). Similarly, monthly time windows require fewer measurements of simple random days than random consecutive days (mean 3.4, SD 0.5 vs mean 4.4, SD 0.5; p-value = 0.025). Younger participants and those tracking steps via smartphones consistently required more days. Being obese was associated with more measurement days for weekly time windows. In sum, to obtain reliable daily step count level, we recommend at least 3 measurement days for weekly and 5 days for monthly time window in adults. Fewer days could be considered for adults age 60+ years, while more days are required when tracking daily step via smartphones. |
|---|