Purpose: To assess test-retest reliability and validity of the Youth Risk Behavior Survey (YRBS) items for moderate and vigorous physical activity in middle school students.
Methods: Students (N = 125; 12.7 +/- 0.6 yr) wore Actigraph accelerometers for 6.1 +/- 1.0 d and twice completed surveys that included YRBS moderate and vigorous physical activity items. Accelerometer counts were transformed into minutes of moderate (3-6 METs) and vigorous (> 6 METs) physical activity. Days per week meeting moderate and vigorous physical activity recommendations were estimated using four summary methods. Reliability was assessed using intraclass correlation coefficients (ICC) from the two surveys. Validity was assessed as percent concordance, kappa coefficients, and sensitivity and specificity using binary YRBS and Actigraph outcomes.
Results: Test-retest ICC for the moderate and vigorous physical activity items were 0.51 and 0.46, respectively. Twenty-two percent of students met the recommended level of moderate physical activity (>or= 30 min.d(-1), >or= 5 d.wk(-1)) according to self-reports, whereas 90.4 and 66.4% met the recommendation according to accumulated accelerometer minutes and 5-min-bout criteria, respectively. Concordance between YRBS and Actigraph moderate physical activity measures was highest using accumulated accelerometer minutes. Sensitivity of the moderate YRBS item ranged from 0.19 to 0.23 for four comparisons, and specificity was 0.74-0.92. More than two thirds of students reported vigorous physical activity at recommended levels (>or= 20 min.d(-1), >or= 3 d.wk(-1)), whereas the highest prevalence according to Actigraph monitoring was 22.4%. Sensitivity of the YRBS vigorous item was high (0.75-0.92) compared with the four Actigraph measures; specificity was low (0.23-0.26).
Conclusion: YRBS questions underestimate the proportion of students attaining recommended levels of moderate physical activity and overestimate the proportion meeting vigorous recommendations. Use of accelerometry for physical activity surveillance seems to be indicated. At the minimum, new questions demonstrating greater validity are needed.