Related Experiment Video
Updated: Sep 17, 2025
Visualization of Intensity Levels to Reduce the Gap Between Self-Reported and Directly Measured Physical Activity
Published on: March 7, 2019
Characterization and clustering of intra-day physical activity patterns using accelerometry among sexual and gender
David Lopez-Veneros1, Billy A Caceres1,2, Kasey Jackman1
1Columbia University School of Nursing, New York, NY, 10032, USA.
Background:
Sexual and gender minority (SGM) adults experience significant health disparities linked to chronic exposure to minority stressors (e.g., discrimination), and could be reciprocally associated with physical activity (PA) behavior. While PA is a health-protective factor, research on PA patterns in SGM adults is limited. Identifying potential latent PA profiles can inform tailored behavior change approaches.
Objective:
To investigate latent profiles (i.e., clusters) of daily PA trajectories among sexual and gender minority (SGM; lesbian, gay, bisexual, transgender, queer) adults using functional latent block models (FLBMs), a co-clustering technique that simultaneously accounts for variation at the individual- and day-level.
Study Sample:
The study included 42 Black and Latinx SGM adults who wore Fitbit trackers for up to 30 days of PA data collection as a part of a sleep health study, yielding 1,209 person-level days of step count data.
Methods:
Each 24-h period of step counts was smoothed using Fourier-transform to create the functional data matrix and fit the FLBMs. The optimal number of clusters was determined using the integrated completed likelihood (ICL) criterion.
Results:
The best-fitting model identified 3 individual-level clusters (K) based on the daily step count patterns (ICL = -88,495.88). Low activity cluster (n = 11) was characterized with the lowest overall PA, slightly later bedtimes, and the least intra-day and hourly variability. Steady moderate activity cluster (n = 23) was characterized by a gradual increase in step counts that spread over the course of the day, with a small peak in the afternoon. Fluctuating high activity cluster was characterized by a peak in activity earlier in the day, compared to other clusters. Cluster 3 membership was also associated with the highest volume of PA overall, along with hourly and daily variability in step counts and higher intensities of PA. The model secondarily identified 2 day-level clusters (L), representing weekday and weekend PA patterns.
Conclusions:
We identified distinct habitual daily PA trajectories among SGM adults based on daily volume and variability. Analyzing individual PA variances can help identify inactive periods and individuals at higher risk, which can inform the design of tailored interventions and self-management strategies to promote PA.
