The participants were 1,310 students from a junior high school. After excluding incomplete or invalid questionnaires (e.g., those with blank responses) (44), 1,205 valid questionnaires were obtained, yielding a validity rate of 91.98%. In this study, left-behind children were identified using the following questions: (1) “Do any family members migrate (leave the hometown) for work?” (2) “How often does your father return home?” (3) “How often does your mother return home?” Participants were classified as left-behind children if they were under 18 years of age and had at least one parent who had migrated to an urban area for work for six months or longer, while the child remained in their hometown. Based on parental migration status, left-behind children were further categorized as: (a) migrant father only (mother remained at home), (b) migrant mother only (father remained at home), and (c) both parents migrated. In our sample, 242 students (20.1%) were identified as LBC, including 164 with migrant fathers, 17 with migrant mothers, and 61 with both parents migrating. The remaining 963 students (79.9%) were non-left-behind children. The average age of the participants was 14.60 years (SD = 1.17). The sample included 597 boys (49.5%) and 608 girls (50.5%), with 397 students (32.9%) in the seventh grade, 497 (41.2%) in the eighth grade, and 311 (25.8%) in the ninth grade (as Table 1).

Data from junior high school students were collected in September 2024. A junior high school in Shandong Province was randomly selected, and twenty classes were subsequently randomly chosen from within this school. Data were collected during the class using a paper/pencil version survey administered to all students in this school classes. Research staff were trained before they administered the survey. The study was conducted in accordance with the Declaration of Helsinki as revised in 1989 and was approved by the Ethics Committee of Liaoning Normal University, the approval number is No. LL2024166. Student assent was obtained from both schools and parents.

Descriptive statistics and correlations for all study variables were calculated. Multicollinearity among the variables was considered by evaluating variance inflation factors (VIF). Latent profile analysis (LPA) was conducted using Mplus 8.3 to identify distinct profiles of adolescents based on their experiences of childhood maltreatment and peer victimization. Latent profile analysis is a person-centered approach designed to identify mutually exclusive, discrete latent profiles derived from a set of observed continuous indicators. In order to provide a much more intuitive interpretation, all study variables were converted into Z scores before conducting latent profile analysis. Models specifying one- through six-profile solutions were estimated and compared. After the optimal latent profile solution was identified, associations between demographic characteristics and profile membership were examined using the R3STEP approach (49). Specifically, a multinomial logistic regression was performed within the LPA framework, in which the latent profile variable was regressed simultaneously on gender and left-behind status. Finally, differences in pathological Internet use across the identified latent profiles were examined using the DU3STEP approach implemented in Mplus (50). Pathological Internet use and its specific dimensions were treated as continuous distal outcomes, while gender and left-behind status were included as covariates in the third step of the analysis. Overall profile differences were evaluated using omnibus chi-square tests, followed by pairwise comparisons conducted via MplusAutomation when significant effects were observed. This approach accounts for classification uncertainty in latent profile membership.

Table 2 shows the descriptive and correlation of study variables. The variance inflation factor indicated that VIFs < 5 in the study, suggesting no significant multicollinearity issues (51). Childhood maltreatment is significantly positively correlated with peer victimization. All dimensions of childhood maltreatment and peer victimization are positively correlated with pathological Internet use.

Table 3 presents the estimation of six latent profiles, ranging from a one-profile solution to a six-profile solution. The statistical analysis showed that the values of the Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), and adjusted Bayesian Information Criterion (aBIC) decreased as the number of specified profiles increased. For all profiles, entropy values above 0.8 indicated good profile separation, and the Bootstrap Likelihood Ratio Tests (BLRTs) were significant. Additionally, the AIC, BIC, and aBIC values consistently decreased with an increasing number of profiles (52).

Although the four-profile model yielded slightly lower AIC and BIC values, the Lo–Mendell–Rubin test was no longer significant (p = 0.29), and entropy decreased from 0.92 to 0.88, indicating that the additional profile did not substantially improve model fit or interpretability. The three-profile model demonstrated a more parsimonious structure with clearer class separation and theoretical meaningfulness. Therefore, the three-profile model was retained as the optimal representation of the data.

Figure 1 presents the levels of childhood maltreatment and peer victimization for the three profiles. The first profile reported low scores on both childhood maltreatment and peer victimization, and was therefore labeled as the “low-risk” profile (n = 974, 80.83%). The second profile showed relatively high scores on peer victimization and relatively low scores on childhood maltreatment, and was labeled the “severe peer victimization” profile (n = 143, 11.86%). The third profile demonstrated high scores on emotional and physical abuse, as well as neglect, with relatively high scores on peer victimization, and was labeled the “severe childhood maltreatment with elevated peer victimization” profile (n = 88, 7.31%).

Table 4 presents the distribution of gender and left-behind status across the three latent profiles of childhood maltreatment and peer victimization. Adolescents in the low-risk profile constituted the largest proportion of the sample. Notably, a substantial proportion of left-behind children were classified into this low-risk profile, whereas the severe peer victimization and the severe childhood maltreatment with elevated peer victimization profiles comprised comparatively smaller subgroups.

To examine whether gender and left-behind status were associated with latent profile membership while accounting for classification uncertainty, a multinomial logistic regression was conducted using the R3STEP approach in Mplus. In this analysis, the retained three-profile latent variable was regressed simultaneously on gender (1 = male, 0 = female) and left-behind status (0 = non–left-behind children, 1 = left-behind children with migrant fathers, 2 = left-behind children with migrant mothers, 3 = left-behind children with both migrant parents). Table 5 presents the regression estimates for gender and left-behind status predicting membership in the childhood maltreatment and peer victimization profiles. Significant differences were observed across profiles. Specifically, compared with adolescents in the low-risk and severe peer victimization profiles, boys were more likely to belong to the severe childhood maltreatment with elevated peer victimization profile. In addition, adolescents who were left behind by migrant mothers showed a higher likelihood of membership in the severe childhood maltreatment with elevated peer victimization profile than those who were not left behind.

To examine whether pathological Internet use differed across latent profiles of childhood maltreatment and peer victimization, distal outcome analyses were conducted using the DU3STEP approach in Mplus. This procedure estimates differences in distal outcomes across latent profiles while accounting for classification uncertainty. All dimensions of pathological Internet use were treated as continuous distal outcomes, and gender and left-behind status were included as covariates in the third step of the model. After controlling for gender and left-behind status, a Wald χ² omnibus test indicated a significant overall difference among the three profiles in the negative consequences dimension of pathological Internet use, χ² (2) = 6.05, p = 0.04 (see Table 6). Subsequent pairwise comparisons showed that adolescents in the severe childhood maltreatment with elevated peer victimization profile reported significantly higher levels of negative consequences than those in the low-risk profile. Differences between the severe peer victimization profile and the other two profiles were not statistically significant. For the remaining dimensions of pathological Internet use—including salience, tolerance, compulsive Internet use, mood alteration, social comfort, and overall pathological Internet use—the omnibus Wald χ² tests were not significant. Consistent with these results, pairwise comparisons revealed no significant differences between any of the latent profiles on these dimensions after controlling for gender and left-behind status.

This experimental investigation identified three distinct co-occurrence patterns of childhood maltreatment and peer victimization through latent profile analysis: low-risk profile (80.83%), severe peer victimization profile (11.86%), and severe childhood maltreatment with elevated peer victimization profile (7.31%). Although the number of profiles differs from previous studies—such as the longitudinal UK–US investigation identifying four groups (15) and the six-profile solution reported by Fu et al. (23) among Chinese adolescents—these discrepancies likely reflect regional, developmental, and methodological variations. Importantly, our findings converge with prior research in demonstrating that childhood maltreatment and peer victimization frequently co-occur (23, 53). It should be noted that the profile characterized by severe childhood maltreatment and elevated peer victimization represents a relatively small proportion of the overall sample (7.31%). Rather than indicating population-level prevalence, this subgroup reflects a high-risk, high-need minority of children who are simultaneously exposed to adversity within both family and peer contexts. The identification of this subgroup provides empirical support for spillover theory, which posits that dysfunction in one social system (e.g., the family) may heighten vulnerability in other contexts (e.g., peer relationships). Children experiencing severe maltreatment may develop emotional dysregulation and maladaptive interpersonal behaviors that increase their susceptibility to peer victimization, thereby accumulating risks across contexts (54, 55). Consistent with prior literature, the majority of children in the present sample were classified into the low-risk profile, experiencing little to no maltreatment or peer victimization (23). This pattern aligns with evidence suggesting that most children are not chronically exposed to victimization, while a relatively small subset experiences multiple, overlapping forms of adversity. From a practical perspective, these findings underscore the importance of moving beyond population-level averages to identify and support children within this particularly vulnerable subgroup. Targeted screening and intervention efforts focusing on children exposed to both family maltreatment and peer victimization may be especially critical for preventing long-term maladjustment.

Gender and the type of left-behind status can predict profiles of childhood maltreatment and peer victimization. Boys were more likely to belong to the severe childhood maltreatment with elevated peer victimization profile compared to girls, suggesting a heightened exposure to multiple forms of adversity. This finding is consistent with prior Chinese studies showing that boys tend to experience more frequent and diverse forms of victimization (56). This gender difference may be explained by traditional childcare patterns in China, where families often have higher expectations for boys. Parents expect boys to succeed and carry on the family name, becoming primary providers for their families and aging parents in adulthood (55). Moreover, previous research has consistently shown that boys are more vulnerable to peer victimization than girls (57). This gender disparity may be attributable to differences in peer interaction patterns. Compared with girls, boys typically engage in peer groups that are larger, more loosely structured, and more competitive, which increases both the frequency and intensity of aggressive interactions among peers (58). In contrast, prevailing social norms and moral expectations often position girls as a protected group, affording them greater adult supervision and social buffering, which may reduce their risk of peer victimization (59).

The present study further revealed that approximately 70% of left-behind children were classified into the low-risk group. However, compared with non-left-behind children, adolescents whose mothers had migrated for work were significantly more likely to be assigned to the dual-risk profile characterized by severe childhood maltreatment and elevated peer victimization. This finding underscores the particularly salient role of maternal absence in shaping children’s exposure to cumulative adversity across family and peer contexts. In the Chinese cultural context, mothers are typically the primary caregivers and are more directly involved in children’s daily supervision, emotional support, and socialization (60). Maternal migration may therefore disrupt consistent caregiving, reduce emotional availability, and weaken monitoring of children’s peer interactions, thereby increasing vulnerability to both maltreatment within the family and victimization in peer settings (27). Compared with mothers who migrate to distant urban labor markets and thus are absent from daily caregiving, fathers who remain behind often have fewer caregiving resources, weaker emotional engagement, and lower motivation or capacity to fulfill the traditional breadwinner-caretaker role, which may exacerbate developmental disadvantages among left-behind children (61). From a family systems perspective, maternal absence may also exacerbate stress within the caregiving environment, particularly when alternative caregivers such as fathers or grandparents face limited resources, caregiving capacity, or emotional closeness with the child. These strains may heighten the risk of harsh, neglectful, or inconsistent parenting practices, increasing the likelihood of childhood maltreatment. Concurrently, insufficient parental guidance and emotional support may impair children’s emotion regulation and social competence, which are critical protective factors in peer relationships, thereby elevating the risk of peer victimization. The co-occurrence of family maltreatment and peer victimization among children with migrant mothers thus provides further support for a spillover process, whereby adversity in the family context increases susceptibility to harm in peer environments (62).

Taken together, the findings of the present study indicate that both gender and differentiated forms of left-behind status are associated with distinct patterns of exposure to childhood maltreatment and peer victimization. Rather than exerting uniform effects, these sociodemographic factors appear to shape children’s vulnerability in nuanced ways, with certain subgroups—such as boys and children experiencing maternal absence—showing elevated risk for compounded adversity across family and peer contexts. Prevention and intervention efforts may benefit from tailored strategies that account for heterogeneity in children’s developmental contexts, enabling more precise identification of high-need subgroups and more effective allocation of support and resources.

Consistent with prior research, adolescents exposed to childhood maltreatment and peer victimization are at elevated risk for problematic patterns of Internet use (42, 63). However, the present findings extend this literature by demonstrating that, after controlling for gender and left-behind status, differences across latent profiles emerged only in the negative consequences dimension of pathological Internet use. Specifically, adolescents characterized by co-occurring high levels of childhood maltreatment and peer victimization exhibited significantly greater functional impairments related to Internet use than those in the low-risk profile. This pattern suggests that cumulative interpersonal adversity may not necessarily translate into higher levels of compulsive or emotionally driven Internet use, but rather manifests in more pronounced disruptions to daily functioning, such as academic difficulties, interpersonal conflicts, and reduced engagement in offline activities (64). Importantly, this pattern suggests that adolescents exposed to cumulative family and peer adversity may not primarily engage in Internet use for heightened pleasure or social comfort, but instead experience greater difficulty containing Internet use within adaptive boundaries. As a result, negative outcomes may represent a more proximal and sensitive manifestation of risk than motivational or affective PIU components in this population. Adolescents who have experienced peer victimization and childhood maltreatment are more likely to develop intense negative emotions and may initially use the Internet as an accessible coping resource to regulate these feelings. However, prolonged reliance on this strategy can undermine real-life emotional regulation and daily functioning, leading to observable negative consequences (65). In addition, the co-occurrence of family- and peer-level victimization may compromise adolescents’ self-esteem and social competence, limiting their capacity to regulate Internet use effectively and increasing their vulnerability to maladaptive outcomes (66). While prior studies have focused on global Internet addiction outcomes, our finding emphasizes the need to consider specific functional impairments as a sensitive indicator of risk in adolescents exposed to complex adversity profiles.

This study employed latent profile analysis to concurrently examine childhood maltreatment and peer victimization, identifying distinct risk profiles based on these adverse experiences. From a practical perspective, intervention strategies can be tailored to each latent profile. For instance, adolescents in the severe peer victimization profile may benefit from peer support programs and social skills training, whereas those in the severe childhood maltreatment with elevated peer victimization profile may require combined family interventions, emotion regulation training, and targeted psychosocial support to mitigate the compounded risks. For left-behind children, particularly those experiencing maternal absence, interventions may need to further emphasize enhancing caregiving quality, emotional support, and monitoring within the home to buffer both family- and peer-related adversities. However, this study still has some limitations. First, childhood maltreatment, peer victimization, and pathological Internet use are dynamic experiences that may change across developmental periods. The cross-sectional design precludes the examination of temporal relationships and causal inferences, and may also be subject to quantitative biases such as classification uncertainty inherent in single-wave latent profile analyses. Future research could address these limitations by implementing multi-wave longitudinal designs using trajectory-based or transition-based mixture models, which allow for examination of changes in risk profiles over time, the stability of latent classes, and the effects of time-varying covariates while reducing potential bias. Second, the reliance on self-report measures may introduce common method bias and social desirability effects. Future research could incorporate multiple informants or behavioral indicators to enhance measurement validity. Third, the sample was drawn from middle school students in a specific region of China, which may limit the generalizability of the findings; replication across diverse regions and cultural contexts is recommended. Lastly, this study focused solely on two types of adverse experiences and did not consider other important factors—such as family functioning, emotion regulation, or school context—that may interact with adversity to influence Internet use. Future studies should integrate broader ecological variables to develop a more comprehensive risk framework.