Examine the Fitbit-specific data from the ABCD
The Fitbit data and associated QA/QC features were obtained from the ABCD Study data portal. To assess the compliance and quality of data, we used the maximum common observation period of 21 days. We used an unsupervised hierarchical clustering approach to discover latent temporal patterns in each feature. The optimal number of clusters was selected using the Elbow method. Demographic differences between the Fitbit cohort and non-Fitbit cohort, as well as differences between those across clusters per QA/QC feature, were examined through Chi-Squared and Kruskal-Wallis tests.
Our research findings show the presence of bias in real-world wearable data collected from the ABCD Study cohort. The quality and quantity of the data is significantly different between White- and Black/African-American-identifying youth. These underlying significant differences in the quality of wearable data, if not addressed and adjusted for could negatively impact the personalized health trajectories learned wearable data using machine learning algorithms.
