Work

Hidden Markov Trajectories of Early-Adolescent Media Overdependence and Machine Learning Prediction of High-Risk Maintenance from Early Childhood and Lower Elementary Predictors

Year: 2025

Type: article

Abstract: = 1354), we examined predictors measured from early childhood to Grades 1-2 (2014-2016) and modeled digital media overdependence from Grades 3 to 6 (2017-2020). Hidden Markov Models (HMMs) were used to identify developmental trajectories, and machine learning models characterized risk signals using SHAP-informed feature importance. Five trajectories emerged, including one subgroup that maintained persistently high risk. The predictive model showed good discriminative performance (strong predictive performance [Receiver Operating Characteristic Area Under the Curve (ROC AUC) = 0.84]). Executive function difficulties in Grade 1 and their worsening through Grade 2 predicted an elevated risk, whereas longer or increasing sleep duration, stronger family interactions, and appropriate parental control were protective. In contrast, higher maternal parenting stress, greater overall media use time, and a larger proportion of game-centered media use functioned as risk factors. These findings identify modifiable early childhood and early elementary predictors of high-risk maintenance trajectories of media overdependence and may inform early screening and preventive interventions in families, schools, and communities.

Source: Behavioral Sciences

DOI: 10.3390/bs15121725

Authors: Eun-Kyoung Goh, Juyoun Kyun

Institutions: Dong-A University, Kaya University

Language: English

Topic: Child Development and Digital Technology

Subfield: Education

Field: Social Sciences

Domain: Social Sciences

SDG: Good health and well-being

Open Access status: gold

Estimated APC paid: $2,654