The China Internet Information Center reported that the internet addiction rate among Chinese adolescents has reached as high as 10%, indicating that this problem has become a major social health concern among adolescents in China. Previous studies have identified one or more subgroups of adolescents whose trajectory of internet use behaviors puts them at a high risk of addiction, but further research is needed to determine and understand these high-risk groups and fill research gaps. Furthermore, most previous studies have approached the problem from the perspective of the variables of internet addiction, but its symptomatology remains poorly understood. The current study combines a growth mixture model (GMM) with network analysis to identify heterogeneous groups of adolescents at a high risk of internet addiction and to explore the changes in symptomatology in these groups.

A three-year longitudinal study followed students from the time they entered junior high school. Three assessments were conducted yearly at identical intervals (T1: October 2016 to November 2016, T2: October 2017 to November 2017, and T3: October to November 2018). Ultimately, 1, 279 adolescents (662 boys and 617 girls) completed the assessments at each time point. Internet addiction was assessed using the 10-item Internet Addiction Test. Mplus 8.0 was used for the descriptive statistics, correlation analysis, and the GMM to estimate the development trajectories of various heterogeneous groups. An R package was used to estimate the network structure and core symptoms of internet addiction of each high-risk group at each time point.

The GMM showed two class model showed better goodness of fit. The intercept means of the two potential categories were C1: 2.36 (*SE* = 0.25, *t* = 9.47, *p* < 0.001) and C2: 1.48 (*SE* = 0.05, *t* = 27.32, *p* < 0.001) respectively. There was a significant difference between the intercept means of the two potential categories, with the C1 group having a higher initial value of internet addiction score and the C2 group having a relatively lower initial value score. In addition, the mean growth rate of each category was examined by the mean of the slopes. The mean slope values for the two potential categories were C1: 1.62 (*SE* = 0.14, *t* = 11.45, *p* < 0.001); C2: −0.27 (*SE* = 0.03, *t* = −8.36, *p* < 0.001). The level of internet addiction scores changed significantly over time in both groups, with a significant increase over time in group C1 and a decrease in group C2. The analysis of the intercept and slope means showed that the initial levels were higher in group C1 and increased significantly over time. In contrast, the C2 group had a lower initial level and decreased significantly over time. Based on this, the two potential categories were named: C1 as high-risk group, with 11.65% of the sample (*n* = 149); and C2 as normal group, with 88.35% of the sample (*n* = 1130).

Network analysis revealed that the mean network density at the three time points was 0.25, 0.30 and 0.15, respectively, indicating that the strongest association between symptoms was at T2, while the weakest association between symptoms was at T3. The core symptoms of internet addiction among the adolescents in the high-risk group differed at each time point (See

This study enhances the understanding of the symptomatology of internet addiction among high-risk adolescents, indicating that targeted interventions must be developed based on the various stages of adolescence. From the first year of junior high school, strategies should be implemented to prevent the development of internet addiction in high-risk groups. In the second year, adolescents in the high-risk group should be identified by focusing on their satisfaction deficits. In the last year of junior high school, interventions should target adolescents’ withdrawal symptoms of internet addiction.