How Many Days of Commercial Wearable Monitor Data Reflect Habitual Steps per Day? Insights From 9 Weeks Continuous Data in People Living With Diabetes

Understanding how many days of wearable device monitoring are needed to accurately reflect habitual behavior is important when aiming to assess population activity levels, both in research and public health contexts. However, few studies have focused on data collected from commercial wearables and on more interpretable metrics such as daily step count. Therefore, we investigated the number of days required to achieve a good level of reliability (intraclass correlation coefficient [ICC] ≥ .8) in group-level habitual daily step count using data collected in a sample of 239 people living with diabetes who wore a Fitbit Charge 4 consecutively for at least 9 weeks. Subgroup analyses also included assessments of reliability by average step count, variability in step count, diabetes type, sex, and age. Results of the Spearman–Brown analyses revealed that 7 days of monitoring was found to obtain an ICC value ≥ .9 in the overall sample, with 4 days of monitoring providing a reliability of ≥.8. All subgroups required 7 days for an ICC of ≥ .8 and 16 days for an ICC of ≥ .9. Additional analyses revealed that having low intraindividual variability in step count is an important factor in reducing the number of days required to meet acceptable reliability. Therefore, measures of intraindividual or within-person variability should be included alongside calculations of average daily steps to determine the potential reliability of group level estimates.<p></p>