(PS5-69) Do Subjective and Objective Sleep Measures Predict Adolescent Depression? Examining Sleep Assessed via Youth-report and Fitbit in the Adolescent Brain Cognitive Development Study

Background: Adolescent-onset depression is associated with adverse life outcomes, and psychotherapy for youth depression is only modestly efficacious, underscoring the importance of identifying risk factors of adolescent depression for prevention and early intervention. Prior research has shown that sleep may be an important risk factor that predicts the development of depression in adolescents. However, most studies rely on self-reports of sleep, which are susceptible to reporting bias; use a cross-sectional design, which precludes inference of causal and temporal relationships; and have small participant sample sizes, which limits power to detect associations. Therefore, more longitudinal studies that incorporate both subjective and objective measures of sleep are needed.

OBJECTIVES: We will examine whether (a) objective and subjective measures of sleep converge at both Year 2 and Year 4, and (b) sleep problems assessed via subjective and objective methods (e.g., shorter sleep duration, longer sleep onset latency, more frequent awakenings, and longer awakening time) at Year 2 predict subsequent depression and internalizing symptoms at Year 4 using data from the Adolescent Brain Cognitive Development (ABCD) Study.

Methods: The ABCD study is a longitudinal cohort study of 11,878 youth recruited at age 9-10 years from 21 sites across the U.S. from 2016-2018, and followed up for 10 years. Youth participants and their caregivers complete an in-person visit each year. The present study utilizes data from Years 2 and 4 collected via the parent-reported Child Behavior Checklist (CBCL), youth-reported Munich ChronoType Questionnaire (MCTQ),and FitBit Charge HR, a consumer-level wearable tracker for sleep and physical activity worn by youth.

Results:  Sleep duration derived from MCTQ was positively correlated with sleep duration derived from Fitbit at Year 2 (r(5037) = .19, p < .001) and Year 4 (r(2096) = .16, p < .001), as expected. However, the number of awakenings obtained from MCTQ was negatively correlated with the number of awakenings obtained from Fitbit at Year 2 (r(5037) = -.06, p < .001). Sleep variables and CBCL depression symptoms at Year 2 significantly predicted depression symptoms at Year 4, F(9,2865) = 130.5, p < .001, R2 = .289. Specifically, sleep duration from MCTQ (b = -.13, t(2865) = -1.966, p < .05) and number of awakenings from Fitbit (b = -.07, t(2865) = -2.67, p < .01) at Year 2 were significant predictors of depression symptoms at Year 4. Furthermore, the variance in CBCL depression scores at Year 4 uniquely accounted for by sleep variables at Year 2, controlling for CBCL depression scores at Year 2, was small but significant (Delta R2  = 0.25%, F = 2.27, p < .05). 
DISCUSSION: Consistent with prior literature, the results revealed discrepancies between subjective and objective sleep measurements. Subjective and objective measures of sleep problems at Year 2 significantly predicted depression at Year 4, but effect sizes were small. Future research needs to better understand the incongruence between subjective and objective sleep measurements because they may represent different elements of sleep problems. Sleep variables appear to be risk factors for the development of adolescent depression.