Physical activity recall (PAR) reliability was estimated in a three-site sample of African American and white adults. The sample was sedentary at baseline and more varied in physical activity 24 months later. Intraclass correlation coefficients (ICCs) were used to estimate the number of PAR assessments necessary to obtain a reliability of 0.70 at both timepoints.
The PAR was administered ≤ 30 d apart at baseline (n = 547) and 24 months (n = 648). Energy expenditure ICC was calculated by race, gender, and age.
Baseline reliability was low for all groups with 4–16 PARs estimated to attain reliable data. ICCs at 24 months were similar (ICC = 0.54–0.55) for race and age group, with 2–3 PARs estimated to reach acceptable reliability. At 24 months, women were more reliable reporters than men.
Low sample variability in activity reduced reliability, highlighting the importance of evaluating diverse groups. Despite evaluating a sample with greater physical activity variability, an estimated 2–3 PARs were necessary to obtain acceptable reliability.