Heterogeneous Sleep Duration Trajectories from Late Childhood to Mid-Adolescence: A 7-Year Longitudinal Study of Korean Youth Using Growth Mixture Modeling

Background and Purpose: Adequate sleep is essential for healthy adolescent development, yet sleep duration declines substantially across childhood and adolescence. Although prior research has documented average-level declines, little is known about whether distinct subgroups of youth follow qualitatively different sleep trajectories and what biopsychosocial factors predict membership in those subgroups particularly across a period that spans the COVID-19 pandemic. Using a nationally representative Korean panel, the present study applied growth mixture modeling (GMM) to identify heterogeneous sleep duration trajectory classes across seven annual waves and examined baseline predictors of class membership via multinomial logistic regression.

Methods: Data were drawn from the Korean Children and Youth Panel Survey (KCYPS) 2018 4th cohort (N = 2,462; 48.5% female), spanning Waves 1 through 7 (2018–2024; Grades 4–10). Sleep duration (hours/day) was computed as a weighted weekly average of self-reported weekday and weekend sleep times. GMM with quadratic growth functions and a linear link function (lcmm package in R) was estimated for one- to four-class solutions; model selection was guided by AIC, BIC, SSABIC, entropy, and class interpretability. Multinomial logistic regression then examined Wave 1 baseline predictors of trajectory class membership, including demographics (sex, BMI category), psychosocial factors (self-esteem, depression, social anxiety, life satisfaction, happiness, interpersonal competence), behavioral factors (academic engagement, grit, smartphone use and addiction), and family factors (parental warmth, rejection, autonomy support, control, and inconsistency). BMI was categorized using sex-specific Wave 1 sample percentile cutoffs (underweight < 5th; normal ≥85th–<95th; overweight ≥85th–<95th; obese ≥95th), with normal weight as the reference.

Results: GMM identified three distinct sleep duration trajectory classes: High/Slow Decline (15.1%; n = 372), characterized by persistently higher sleep with a gradual decrease; Steep Decline (14.7%; n = 361), showing the most pronounced drop across waves; and Moderate Decline (70.2%; n = 1,729), the normative trajectory with a moderate decline. All three classes showed continuous sleep reduction across late childhood and adolescence, with declines persisting through the COVID-19 disruption period (Waves 3–5; 2020–2022) and the post-pandemic period (Waves 6–7; 2023–2024). Multinomial regression revealed that female sex significantly predicted membership in the Steep Decline class relative to Moderate Decline (OR = 1.43, p =.006), and boys were more likely than girls to be in the High/Slow Decline versus Steep Decline class (OR = 0.60 for female, p=.002), suggesting sex differences operate through differential class selection rather than within-class trajectory shape. Relative to the Moderate Decline class, High/Slow Decline membership was predicted by underweight BMI status (OR=1.84, p=.010), higher self-esteem (OR=1.80, p=.022), greater depressive symptoms (OR=1.59, p=.005), lower social anxiety (OR=0.75, p=.005), higher parental warmth (OR=1.47, p=.034), and lower parental inconsistency (OR=.77, p=.049). No psychosocial or family predictor significantly differentiated the Steep Decline from Moderate Decline class beyond sex.

Conclusions and Implications: Findings demonstrate that Korean youth do not follow a single uniform sleep trajectory; rather, meaningful subgroups with distinct developmental courses emerge as early as Grade 4. The persistence of sleep decline through the COVID-19 pandemic and beyond further highlights the need for sustained, longitudinal monitoring of adolescent sleep in Korea.