Agreement Between Consumer and Research-Grade Physical Activity Monitors in a Public Health Intervention for Adolescent Latinas

Abstract

Consumer wearables are increasingly used in physical activity (PA) interventions, but their validity as a measurement tool among low PA groups, like adolescent girls, is unclear. We assessed the minute- and day-level agreement between PA measures among adolescent Latinas from an intervention. Participants wore a Fitbit Inspire HR and an ActiGraph GT3X+ for overlapping epochs. ActiGraph data were classified using two different cut points and aligned with Fitbit data to produce 1,149,169 matched minutes of wear across 137 adolescent girls (M = 15.73 yrs). Confusion matrices were calculated for pairwise comparisons to determine minute-level Moderate-Vigorous PA (MVPA) classification. Data were aggregated to 1007 days for Bland–Altman analyses. ActiGraph cut points showed moderate agreement for minute-level MVPA classification (Balanced Accuracy = 0.71, AC1 = 0.98), while Fitbit showed fair agreement (Balanced Accuracy = 0.50, AC1 = 0.95–0.97) largely driven by non-MVPA observations. The Freedson cut point overestimated daily MVPA relative to Treuth by 14.7 min/day and Fitbit by 14.2 min/day in Bland–Altman space. The daily Treuth and Fitbit comparison did not significantly differ. Findings suggest systematic differences between cut points that warrant further consideration. Fitbit showed moderate agreement with ActiGraph, but heteroscedasticity and the epoch of aggregation significantly impacted agreement. Understanding device differences has implications for promoting/researching public health among adolescents.