Measuring sleep duration in Brazilian adolescents: A validation of self-reports against accelerometry

Sleep plays an essential role in physical, cognitive, and emotional development. Although susceptible to memory and perceptual biases, self-reports are often used. We aimed to assess the validity of self-reported sleep duration in comparison to accelerometry (as the gold standard) in adolescents from a population-based prospective birth cohort. We used data from the 2004 Pelotas Birth Cohort study in Brazil. Sleep duration was assessed at the 11-year follow-up using objective (triaxial accelerometers) and subjective (self-reports based on the Pittsburgh Sleep Quality Index) methods. Covariates included the adolescent’s sex, skin color, sleep quality, screen time, and emotional problems, and maternal age, schooling, and depression. We compared the mean sleep duration by accelerometry and self-report across covariates using t-tests and ANOVA. To assess the validity of self-reported sleep duration, we estimated sensitivity, specificity, accuracy, positive predictive value (PPV), and negative predictive value (NPV), as compared to accelerometry. The average sleep duration was higher in self-reported data (9.28 hours; 95% CI 9.22; 9.35) compared to measurements obtained via accelerometry (8.02 hours; 95% CI 8.00; 8.06). Sensitivity was 48.9% (95% CI: 43.8; 54.1), and specificity was 58.4% (95% CI: 56.5; 60.3). The PPV and NPV were 14.2% and 89.0%, respectively. The overall accuracy of self-reported sleep duration was 57.2%. The validity attributes of selfreported sleep duration remained weak across all covariates, with accuracy ranging between 49.2% and 60.7%. Our findings emphasize the importance of using objective sleep measures, particularly in studies involving vulnerable populations, where selfreports may be less reliable.