Building Machine Learning Predictive Models for Adolescent Internet Addiction: Key Findings on Self-Esteem and Resilience Interaction

Building Machine Learning Predictive Models for Adolescent Internet Addiction: Key Findings on Self-Esteem and Resilience Interaction | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Building Machine Learning Predictive Models for Adolescent Internet Addiction: Key Findings on Self-Esteem and Resilience Interaction Rongmei Liu, Saiyi Wang, Clifford Silver Tarimo, Quanman Li, Yifei Feng, and 5 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5606509/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Objective: Internet addiction (IA) is a significant mental health concern among adolescents. This study aimed to develop machine learning (ML)-based predictive models to identify and explain key risk factors for IA. Method: A total of 8176 junior high school students from Henan Province were surveyed from April to May 2023. The dataset was randomly divided into training and test sets in an 8:2 ratio. Four ML algorithms were used to predict IA, and feature importance was determined using SHapley Additive exPlanations (SHAP). The XGBoost model, which achieved the highest area under the curve (AUC), was selected for detailed analysis and individualized prediction explanations. Results: The five most important predictors of IA were negative life events, self-esteem, school connectedness, parent-adolescent cohesion, and psychological resilience. Importantly, an interaction effect was found between self-esteem and psychological resilience: as self-esteem increased, the influence of low resilience transitioned from being a risk factor to a protective factor against IA. Conclusion: This study demonstrates the power of ML models combined with SHAP for predicting IA and identifying its psychosocial determinants. The findings highlight the critical interplay of self-esteem and psychological resilience, offering valuable insights for clinicians and educators in addressing IA among adolescents. Humanities/Medical humanities Social science/Psychology adolescents internet addiction machine learning SHAP self-esteem psychological resilience Figures Figure 1 Figure 2 Figure 3 Figure 4 Highlights ● This study developed machine learning (ML)-based predictive models to identify key risk factors for adolescent internet addiction (IA). ● Negative life events, self-esteem, school connectedness, parent-adolescent cohesion, and psychological resilience were identified as the top five predictors of IA. ● SHAP analysis highlighted a strong positive association between negative life events and IA risk. An interaction effect between self-esteem and psychological resilience showed that higher self-esteem transformed low resilience from a risk factor to a protective factor against IA. ● The integration of ML and SHAP provided clear and interpretable insights into IA risk factors, offering valuable guidance for prevention strategies. 1. Introduction Internet addiction(IA), is defined as an increased desire to use the internet and out-of-control behavior, which is characterized by an intensified desire to reuse the internet (Tams, Legoux ( 1 ).The internet into the daily lives of teenagers has not only brought convenience but also introduced risks. The prevalence estimates for IA exhibit significant variability, current reports in China have shown that there is persistent rise in the prevalence of IA among adolescents, indicating the incidence ranging from 2.2–21.5%), which surpasses that in the United States and several European countries( 2 ). In HK and Macau, about 1.9% of the college students were found to fulfil the criteria of IA( 3 ). More recent studies had found that about 12% of the young adults in South Korea are addicted to the internet( 4 ). Improper internet use is emerging as a public health issue that seriously influences the physical and mental health of adolescents( 5 ). While currently there are no evidence-based treatments for IA, evidence suggests that IA leads to severe social and psychological damage( 6 ). Hence, there is an emphasis on the importance of accurate projections and early intervention. Adolescence constitutes a pivotal phase in human life, marked by the formation of habits, behaviors, and social confidence, during which individual undergoes various physical and mental transformation( 7 , 8 ). Adolescents experiencing IA are commonly engaged in diverse online activities including gaming, social media usage, and online pornography( 9 ). IA in adolescents has been linked to diverse psychological factors( 10 ), encompassing personality traits, negative emotions, self-esteem and impulsivity( 11 ). Prior studies have demonstrated a correlation between elevated IA levels and increased aggressive behavior, including thoughts of suicide and self-harm among adolescents( 12 ). Despite the noted negative consequences of IA among the adolescent population in China, there is a lack of exploration into the potential burden, prevalence and predictors of IA in research studies. Accurate factors predictions for IA in adolescents and a comprehensive understanding of the underlying factors are crucial for designing timely and targeted interventions. Previous studies have employed statistical models, which are known for their stringent assumptions of a linear relationship between the outcome and explanatory variables, in modeling the predictors of IA across diverse populations. Machine Learning (ML) algorithms offer researchers powerful tools. Emphasizing algorithmic prediction in studies within this field is considered promising because ML algorithms are known for their reliability. ML algorithms provide researchers with robust tools utilized across various medical domains, including diagnosis, outcome prediction, treatment, and interpretation of medical images( 13 ). Nevertheless, there remains a deficiency in research concerning ML for the risk prediction of IA. This gap is due to limited evidence supporting real-world clinical applications and the development of interpretable risk prediction models. To address these limitations, the current study used ML algorithms (Logistic Regression (LR), Random Forest (RF), Extreme Gradient Boosting Machine (XGBoost), Support Vector Machines (SVM)), combined with SHapley Additive exPlanations (SHAP), to investigate the predictions of IA in adolescents. These methodologies can assess the interpretability of the model in guiding decisions related to interventions. 2. Methods 2.1 Participants Participants were selected between April and May 2023 using a stratified random cluster sampling approach from six junior high schools in Henan Province, China. Within each school, 8 to 10 classes were randomly chosen from each grade. A total of 8,176 valid questionnaires were collected. The protocol received approval from Zhengzhou University, and all participants provided informed consent. 2.2 Measures 2.2.1 Internet addiction A 20-item Internet Addiction Test was used to assess Internet addiction ( 14 ). It was evaluated by a Likert 5-point scale (ranging from 1 = rarely to 5 = always). Each question was summed up to calculate the total IA score. A total score above 50 was considered indicative of internet addiction ( 15 ) (Cronbach's α coefficient = 0.91). 2.2.2 Negative life events The Adolescent Self-rating Life Events Checklist was adopted to assess the frequency of negative life events ( 刘贤臣 ( 16 ), where in a higher score denotes a greater prevalence of negative life events. The Cronbach's α coefficient in the study was 0.91. 2.2.3 Parent-adolescent cohesion The 10-item Parent-Adolescent Cohesion Questionnaire was employed to measure parent-child and mother-child cohesion levels( 17 ). For questions 3, 4, 8, and 9, reverse coding was applied. The Cronbach's α coefficient in this questionnaire was 0.87. 2.2.4 School connectedness Based on prior studies ( 18 ), school connectedness was evaluated using 10 questions, encompassing three dimensions: teacher support (items 1, 5, and 8), school belonging (items 3, 6, and 9) and classmate support (items 2, 4, 7, and 10) ( 19 ). Response options ranged from 1 to 5 (1 = strongly disagree, 2 = strongly disagree, 3 = uncertain, 4 = strongly agree, 5 = strongly agree). Items 1 and 10 were reverse coded. The higher scores indicated higher levels of student connection (Cronbach's α coefficient = 0.85). 2.2.5 Psychological resilience Psychological resilience was measured by the Connor-Davidson Resilience Scale( 20 ) ( 21 ). 10 items were assessed using a 5-point Likert scale (0 = never, 1 = rarely, 2 = sometimes, 3 = often, 4 = always). The composite score ranged from 0 to 40, with higher scores indicating better psychological resilience. The Cronbach's α coefficient of the scale was 0.90. 2.2.6 Self-esteem Self-esteem was assessed using the Self-esteem Scale developed by Rosenberg( 22 ). 10 items were assessed on a 5-point Likert scale. Items 1, 2, 4, 6, and 7 were reverse coded, and higher scores indicate higher self-esteem (Cronbach's α coefficient = 0.90). 2.2.7 Control variables Control variables for this study included gender, residence (rural, urban), grade (7th, 8th, 9th), family structure (intact family, others), maternal educational level (primary school and below, junior high school, senior high school, university and above), economic level (poor, moderate, good), study burden (light, moderate, heavy) and academic performance (poor, moderate, good). 2.3 Statistical analysis Statistical analysis and data visualization were conducted using Python version 3.7. Categorical variables were presented as numbers and proportions, while continuous variables were reported in Mean ± SD. Binary logistic regression analysis was employed to explore factors related to IA. Additionally, the data were split into training and test sets with an 8:2 ratio, where 80% of the data were used for training the models and 20% for testing the models employing four algorithms, including Logistic LR, RF, XGBoost, and SVM. The LR model forecasts the probability of the binary dependent variable through the application of maximum likelihood estimation for ascertaining the regression coefficient. Both RF and XGBoost are algorithms grounded in tree-based learning. SVM serves as a generalized linear classifier within the framework of supervised learning. SHAP values were utilized to assess the contribution of each feature within each prediction model( 23 ). P <0.05 were considered statistically significant. 3. Results 3.1 Sample characteristics In total, 8176 adolescents (average age 14.42 years; 53.1% men) were included in this study. The prevalence of IA was observed in 1584 (19.4%) participants. Significant statistical differences were observed in gender, age, grade, family structure, maternal educational levels, family economic status, residence, study burden, academic performance, negative life events, psychological resilience, school connectedness, parent-adolescent cohesion, and self-esteem to IA. (See Table 1 ). Table 1 Descriptive statistics of the sample (n = 8176) Variables Internet addiction No (%)/Mean ± SD Yes (%)/Mean ± SD Total P Gender < 0.001 Men 3576(82.3) 769(17.7) 4345(53.1) Women 3016(78.7) 815(21.3) 3831(46.9) Grade 0.017 7th 2528(82.9) 523(17.1) 3051(37.3) 8th 2290(78.3) 635(21.7) 2925(35.8) 9th 1774(80.6) 426(19.4) 2200(26.9) Residence 0.014 Rural 1964(79.0) 522(21.0) 2586(30.4) Urban 4628(81.3) 1062(18.7) 5690(69.6) Only-child 0.422 No 6031(80.7) 1439(19.3) 7470(91.4) Yes 561(79.5) 145(20.5) 706(8.6) Family structure < 0.001 Intact family 6116(81.3) 1410(18.7) 7526(92.0) Others 476(73.2) 174(26.8) 650(8.0) Maternal educational levels < 0.001 Primary school and below 517(74.8) 174(25.2) 691(8.5) Junior high school 2991(81.1) 695(18.9) 3686(45.1) Senior high school 1440(80.3) 353(19.7) 1793(21.9) Un University and above 1644(82.0) 362(18.0) 2006(24.5) Family economic status < 0.001 Poor 363(72.5) 138(27.5) 501(6.1) Moderate 5062(81.4) 1160(18.6) 6222(76.1) Good 1167(80.3) 286(19.7) 1453(17.8) Study burden < 0.001 Light 399(83.1) 81(16.9) 480(5.9) Moderate 3792(85.6) 636(14.4) 4428(54.1) Heavy 2401(73.5) 867(26.5) 3268(40.0) Academic performance < 0.001 Poor 1528(72.8) 572(27.2) 2100(25.7) Moderate 3600(83.4) 716(16.6) 4316(52.8) Good 1464(83.2) 296(16.8) 1760(21.5) Age(years) 14.40 ± 0.94 14.46 ± 0.91 14.42 ± 0.94 0.020 Negative life events 39.25 ± 11.76 52.56 ± 16.39 41.83 ± 13.83 < 0.001 Psychological resilience 24.23 ± 8.15 19.50 ± 8.03 23.32 ± 8.34 < 0.001 School connectedness 38.00 ± 6.77 32.86 ± 7.42 37.01 ± 7.19 < 0.001 Parent-adolescent cohesion 36.29 ± 8.32 31.47 ± 8.40 35.36 ± 8.55 < 0.001 Self-esteem 29.95 ± 5.38 25.98 ± 6.15 29.18 ± 5.75 < 0.001 3.2 Model evaluation Four ML models, LR, RF, XGBoost, and SVM were utilized to predict the occurrence of IA in adolescents (See Table 2 ). XGBoost had the highest AUC (Area under the curve,0.790) and precision (0.795), and its accuracy (0.822), recall (0.608), and F1(0.792) were second high among the four ML models. Figure 1 shows the ROC (Receiver operating characteristic) curves for all models in test sets, while the ROC curves in the training set are supplied as Supplemental Fig. 1. Therefore, we selected XGBoost for further analysis. Table 2 Evaluation of the machine learning model performance. Algorithm AUC Accuracy Precision Recall F1 P a LR 0.782 0.828 0.749 0.618 0.799 < 0.001 RF 0.773 0.811 0.765 0.525 0.739 < 0.001 XGBoost 0.790 0.822 0.795 0.608 0.792 < 0.001 SVM 0.724 0.816 0.760 0.549 0.757 < 0.001 Note: AUC: Area under the curve. a: P value is the result of one-way analysis of variance for the AUC of the five models. LR: logistic. RF: Random Forest. XGBoost: Extreme Gradient Boosting. SVM: Support Vector Machine 3.3 SHAP model interpretation The SHAP value with less than 0 indicates a negative contribution, equal to 0 indicates no contribution, and greater than 0 indicates a positive contribution. The top five features are negative life events, self-esteem, school connectedness, parent-adolescent cohesion, and psychological resilience. The higher the SHAP value of a feature, the higher the probability of developing IA. Adolescents with elevated levels of negative life events (depicted as red dots) were more prone to developing IA compared to those with lower levels (depicted as blue dots). Conversely, adolescents with low levels of self-esteem, school connectedness, parent-adolescent cohesion, and psychological resilience were more likely to develop IA (see Fig. 2 ). Supplemental Fig. 2 depicts the top five variables of feature importance on the model output, revealing a nearly monotonic increase in local SHAP values for negative life events. The SHAP interaction plot (see Fig. 3 ) demonstrates the interaction effects between self-esteem and psychological resilience. A low value for psychological resilience (depicted as blue dots) poses a risk factor. However, as self-esteem improves, a low value for psychological resilience undergoes a transition from being a risk to exhibiting a somewhat protective effect. 3.4 SHAP values of individual prediction for interpretation The force plot illustrates predictions for two randomly selected adolescents No. 5 and adolescents No. 6692, explaining the individual predictions in this study. The function f(x) represents the model output, indicating the predicted probability for each adolescent, while E[f(X)] means the average of the model predictions. Adolescents No. 5, a boy from rural, demonstrates a low risk of IA (-0.023) attributed to protective factors, including self-esteem ( 10 ), psychological resilience ( 21 ), grade ( 9 ), school connectedness ( 12 ), rural residence, and moderate family economic status (see Fig. 4 (a)). In contrast, Adolescent No. 6692, a girl diagnosed with IA in the study, exhibits a high probability of IA (0.442) due to risk factors such as negative life events (55), parent-adolescent cohesion ( 33 ), and good academic performance. (see Fig. 4 (b)). 4. Discussion In this study, we proposed a prediction model developed based on ML algorithms to accurately identify IA in adolescents. Our study provides a meaningful explanation based on the SHAP model. As previously described, the incidence of IA varied significantly under the influence of different social, cultural, and economic backgrounds( 24 ). After the COVID-19 pandemic, IA may become a common problem for society, particularly affecting adolescents( 25 ). Hence, reducing prevalence has become a crucial goal in the management of IA. In the present investigation, comprising 8716 adolescents, and ML prediction models were constructed utilizing 13 distinct features. The significance of a prediction is contingent upon its precision, thereby providing substantial contributions to clinical application. Our study suggests that compared with other features, psychological features play a more significant role in the ML prediction of IA in adolescents. Specifically, the top five features in the prediction models are negative life events, self-esteem, school connectedness, parent-adolescent cohesion and psychological resilience. The ranking of variable importance based on SHAP values revealed that negative life events were the most significant factor influencing IA, a finding that has been rarely reported in previous literature on IA. More specifically, compared to adolescents with lower negative life events, adolescents with higher negative life events were more likely to experience social media addiction. These results indicated that negative life events may play an important role in IA( 26 ). However, our results indicate that lower self-esteem, school connectedness, and parent-adolescent cohesion are associated with the likelihood of IA occurrence. Studies have shown an association between IA and lower self-esteem( 27 ), aligning with our findings. Additionally, lower school connectedness is related to increased IA has been reported( 28 ). It seems plausible that the phenomenon is linked to the promotion of school connectedness, which enhances children's sense of belonging within the school environment. Improving positive interactions between teachers and students may reduce the risks of IA among adolescents. Furthermore, a study has reported parent–adolescent cohesion association with IA, which is consistent with our research. Worthy to pay attention, these points toward different psychological factors are closely related to IA in adolescents( 10 ). The research found that adolescents with psychosocial problems are more prone to IA. A study from Turkey showed that the risk of psychosocial problems in adolescents was 19.8% and 18.8%, respectively( 29 ). It was worth noting that the current findings were aligned with existing evidence suggesting that participants with poor mental health were more likely to cause IA. More importantly, the SHAP model can be used to investigate the individual effects of risk factors and their interaction effects. The SHAP dependence plot depicts the top five variables of feature importance on the model output is presented in Supplementary Fig. 2, revealing a discernible positive correlation between negative life events and IA. Moreover, our study proves the observed SHAP interaction between self-esteem and psychological resilience in adolescents. As self-esteem improves, a lower value for psychological resilience undergoes a shift from being a risk factor to displaying a slightly protective effect. To provide a detailed explanation and interpretation of IA prediction, we presented an individualized explanation of the model predictions through a force plot, illustrating predictions for two randomly selected adolescents. Remarkably, our ranking of variable importance closely aligns with observed differences in variables between subjects with and without IA. For example, adolescents without IA tend to exhibit higher levels of self-esteem, psychological resilience, and school connectedness. Conversely, Adolescent No. 6692 is associated with a high probability of IA risk (f(x) = 0.442) due to factors that elevate the prediction, including higher negative life events, lower parent-adolescent cohesion, good academic performance, and a tendency to be in grade 9 and reside in a rural area. The study contains the following advantages. To date, no predictive models for IA in adolescents have been developed using ML algorithms and SHAP models. Unlike previous research in China, which predominantly relies on traditional regression models, our study represents an innovative approach to addressing the challenges associated with predicting IA in adolescents ( 30 ). The utilization of advanced ML models holds the potential to significantly enhance prediction accuracy. The application of ML algorithms to the medical field is a new trend, demonstrated in studies exploring connections between posttraumatic stress disorder (PTSD) and emotion regulation( 31 ), predicting alcohol use( 32 ) in adolescents, and other clinical applications( 33 ). Additionally, the interaction analysis between self-esteem and psychological resilience is a novel approach to examining the interplay between variables for IA. However, it is imperative to acknowledge the limitations of this study, when interpreting its findings. Firstly, Firstly, adolescents answered self-report questionnaires have recall bias. Secondly, analysis of cross-sectional data sets cannot be used to draw arbitrary conclusions. Finally, other factors affecting IA in adolescents, such as genetic factors( 34 ), were not taken into account in the current study. Future research endeavors may incorporate genetic factors and other potential molecular level predictors, providing insights into the role of genetic mechanisms in IA. 5. Conclusion In summary, our study combined the ML models and the explanation model to reliably predict the risk of IA in adolescents. This approach could assist physicians in intuitively understanding the influence of key features and detecting IA risks early by observing signs of psychological health in individuals. The findings have the potential to promote the identification of IA factors and the development of subsequent strategies for the follow-up care of adolescents. Declarations Conflicts of Interest The authors declare that there are no conflicts of interest. Human Ethics and Consent to Participate declarations Not applicable. Ethical approval The protocol received approval from Zhengzhou University. Informed consent All participants provided informed consent. Data availability statement Due to the confidentiality of the data, our data will not be publicly released. Our data will be provided by the corresponding author if required. Funding Statement Collaborative Innovation System Research on Drug Intervention & Non-drug Intervention in Proactive Health Context (20220518A); Research on Cardiovascular Disease Screening and Healthy Lifestyle Intervention in Children and Adolescents(20230014B) and Platform for Dynamic Monitoring and Comprehensive Evaluation of Healthy Central Plains Action (20220134B). Acknowledgements The authors wish to express their gratitude to all investigators and all participants and thank all their colleagues for their valuable inputs to the study design and data collection. References Tams S, Legoux R, Leger PM. Smartphone withdrawal creates stress: A moderated mediation model of nomophobia, social threat, and phone withdrawal context. Computers in Human Behavior 2018;81:1-9. Mihara S, Higuchi S. Cross-sectional and longitudinal epidemiological studies of Internet gaming disorder: A systematic review of the literature. Psychiatry Clin Neurosci 2017;71:425-444. Ding YJ, Lau CH, Sou KL, et al. Association between internet addiction and high-risk sexual attitudes in Chinese university students from Hong Kong and Macau. Public health 2016;132:60-63. Na E, Lee H, Choi I, et al. Comorbidity of Internet gaming disorder and alcohol use disorder: A focus on clinical characteristics and gaming patterns. The American journal on addictions 2017;26:326-334. Christakis DA. Internet addiction: a 21st century epidemic? BMC Medicine 2010;8. Boer M, Stevens G, Finkenauer C, et al. Attention Deficit Hyperactivity Disorder-Symptoms, Social Media Use Intensity, and Social Media Use Problems in Adolescents: Investigating Directionality. Child development 2020;91:e853-e865. Karaer Y, Akdemir D. Parenting styles, perceived social support and emotion regulation in adolescents with internet addiction. Comprehensive psychiatry 2019;92:22-27. Seider S, Jayawickreme E, Lerner RM. Theoretical and Empirical Bases of Character Development in Adolescence: A View of the Issues. Journal of youth and adolescence 2017;46:1149-1152. Griffiths MD, van Rooij AJ, Kardefelt-Winther D, et al. Working towards an international consensus on criteria for assessing internet gaming disorder: a critical commentary on Petry et al. (2014). Addiction (Abingdon, England) 2016;111:167-175. Zhang W, Pu J, He R, et al. Demographic characteristics, family environment and psychosocial factors affecting internet addiction in Chinese adolescents. Journal of affective disorders 2022;315:130-138. Gao YX, Wang JY, Dong GH. The prevalence and possible risk factors of internet gaming disorder among adolescents and young adults: Systematic reviews and meta-analyses. Journal of psychiatric research 2022;154:35-43. Kuang L, Wang W, Huang Y, et al. Relationship between Internet addiction, susceptible personality traits, and suicidal and self-harm ideation in Chinese adolescent students. Journal of behavioral addictions 2020;9:676-685. Rajkomar A, Dean J, Kohane I. Machine Learning in Medicine. The New England journal of medicine 2019;380:1347-1358. Young KS. Internet Addiction: The Emergence of a New Clinical Disorder. Mary Ann Liebert, Inc 1998. Zhang J H, Zhang X Q, Lu X Y, et al. The mediating role of depression in the relationship between childhood abuse and Internet addiction in adolescents. Chinese Journal of Health Psychology 2022;30:6. Liu Xianchen. Development and reliability and validity test of adolescent life Events Scale. Shandong Psychiatry 1997;10: 5. Zhang W X, Wang M P, Andrew, et al. Adolescents' expectation of autonomy, attitude towards parental authority and parent-child conflict and affinity. Journal of Psychology 2006. Mcneely CA, Nonnemaker JM, Blum RW. Promoting School Connectedness: Evidence from the National Longitudinal Study of Adolescent Health. (Research Papers). Journal of School Health 2002; 72. Yu C F, Zhang W, Zeng Y Y, et al. The relationship between gratitude and problem behavior in adolescents: the mediating role of school bonding. Psychological Development and Education 2011;27;9. Campbell-Sills L, Stein MB. Psychometric analysis and refinement of the Connor-davidson Resilience Scale (CD-RISC): Validation of a 10-item measure of resilience. Journal of Traumatic Stress 2010;20:1019-1028. Connor KM, Davidson JRT. Development of a new resilience scale: The Connor‐Davidson Resilience Scale (CD‐RISC). Depression and Anxiety 2003;18. Rosenberg M. Rosenberg Self-Esteem Scale (RSES). . APA PsycTests 1965. Lundberg S, Lee SI. A Unified Approach to Interpreting Model Predictions. Nips; 2017; 2017. Twh C, Smy S, Mwl C. Adolescent Internet Addiction in Hong Kong: Prevalence, Psychosocial Correlates, and Prevention. The Journal of adolescent health : official publication of the Society for Adolescent Medicine 2019;64:S34. Li YY, Sun Y, Meng SQ, et al. Internet Addiction Increases in the General Population During COVID-19: Evidence From China. The American journal on addictions 2021;30:389-397. Wang X, Ding T, Lai X, et al. Negative Life Events, Negative Copying Style, and Internet Addiction in Middle School Students: A Large Two-year Follow-up Study. International journal of mental health and addiction 2023:1-11. Sevelko K, Bischof G, Bischof A, et al. The role of self-esteem in Internet addiction within the context of comorbid mental disorders: Findings from a general population-based sample. Journal of behavioral addictions 2018;7:976-984. Liu S, Yu C, Conner BT, et al. Autistic traits and internet gaming addiction in Chinese children: The mediating effect of emotion regulation and school connectedness. Research in developmental disabilities 2017;68:122-130. Ozturk F, Ayaz-Alkaya S. Internet addiction and psychosocial problems among adolescents during the COVID-19 pandemic: A cross-sectional study. Archives of psychiatric nursing 2021;35:595-601. Zhang X, Zhang J, Zhang K, et al. Effects of different interventions on internet addiction: A meta-analysis of random controlled trials. Journal of affective disorders 2022;313:56-71. Christ NM, Elhai JD, Forbes CN, et al. A machine learning approach to modeling PTSD and difficulties in emotion regulation. Psychiatry research 2021;297:113712. Afzali MH, Sunderland M, Stewart S, et al. Machine-learning prediction of adolescent alcohol use: a cross-study, cross-cultural validation. Addiction (Abingdon, England) 2019;114:662-671. Li W, Wang J, Liu W, et al. Machine Learning Applications for the Prediction of Bone Cement Leakage in Percutaneous Vertebroplasty. Frontiers in public health 2021;9:812023. Tereshchenko S, Kasparov E. Neurobiological Risk Factors for the Development of Internet Addiction in Adolescents. Behavioral sciences (Basel, Switzerland) 2019;9. Additional Declarations No competing interests reported. Supplementary Files supplementary.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-5606509","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":398801027,"identity":"d2099034-4ae5-4f8b-8bf7-c95bf96ed0d8","order_by":0,"name":"Rongmei Liu","email":"","orcid":"","institution":"Central China Fuwai Hospital, Central China Fuwai Hospital of Zhengzhou University","correspondingAuthor":false,"prefix":"","firstName":"Rongmei","middleName":"","lastName":"Liu","suffix":""},{"id":398801028,"identity":"566eaea6-7788-4854-88ed-468fad1fd816","order_by":1,"name":"Saiyi Wang","email":"","orcid":"","institution":"Zhengzhou University","correspondingAuthor":false,"prefix":"","firstName":"Saiyi","middleName":"","lastName":"Wang","suffix":""},{"id":398801029,"identity":"8a22ae6f-08ca-4296-b031-04ded0072b71","order_by":2,"name":"Clifford Silver Tarimo","email":"","orcid":"","institution":"Zhengzhou University","correspondingAuthor":false,"prefix":"","firstName":"Clifford","middleName":"Silver","lastName":"Tarimo","suffix":""},{"id":398801030,"identity":"50c68a6e-6fb2-48a5-aa0e-954c48e5c171","order_by":3,"name":"Quanman Li","email":"","orcid":"","institution":"Zhengzhou University","correspondingAuthor":false,"prefix":"","firstName":"Quanman","middleName":"","lastName":"Li","suffix":""},{"id":398801031,"identity":"14a9aae9-7602-415f-a6c4-bb3a95211323","order_by":4,"name":"Yifei Feng","email":"","orcid":"","institution":"Zhengzhou University","correspondingAuthor":false,"prefix":"","firstName":"Yifei","middleName":"","lastName":"Feng","suffix":""},{"id":398801032,"identity":"7f4b9ee9-dde1-4354-8c90-99eba914e450","order_by":5,"name":"Lipei Zhao","email":"","orcid":"","institution":"Zhengzhou University","correspondingAuthor":false,"prefix":"","firstName":"Lipei","middleName":"","lastName":"Zhao","suffix":""},{"id":398801033,"identity":"d51e22a5-29fd-4f49-9c9c-e4f40d29ef74","order_by":6,"name":"Shuaibin Liu","email":"","orcid":"","institution":"Zhengzhou Economic Development District","correspondingAuthor":false,"prefix":"","firstName":"Shuaibin","middleName":"","lastName":"Liu","suffix":""},{"id":398801034,"identity":"f6bba27d-b6b8-4611-9fc4-bd91d3062627","order_by":7,"name":"Xinghan Chen","email":"","orcid":"","institution":"Henan University of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Xinghan","middleName":"","lastName":"Chen","suffix":""},{"id":398801035,"identity":"46895f53-46eb-4e84-b98d-2389569fb2d9","order_by":8,"name":"Jian Wu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAv0lEQVRIiWNgGAWjYHCChAMMDDYQJg8JWtJI0wICh0nQYnAj4eGBnzvO28vPSGB88LaNQd6cCC0JB3vP3GZmnJHAbDi3jcFwZwMBLWZALQd4226zMUsksEnztjEkGBwgQsvBv23neNgkEth/E63lMG/bAQkeoC3MRGmxP/Mg4bBsW7KBBM/DZsk55yQMNxDSItmek/zxbZudvXx78sEPb8ps5AnaAoyLBCiDsQFISBBUDwTshE0dBaNgFIyCEQ4AJqo/lEN1fesAAAAASUVORK5CYII=","orcid":"","institution":"Zhengzhou University","correspondingAuthor":true,"prefix":"","firstName":"Jian","middleName":"","lastName":"Wu","suffix":""},{"id":398801036,"identity":"766c70f9-0f15-4c7b-b464-be77e8eeef47","order_by":9,"name":"Qiuping Zhao","email":"","orcid":"","institution":"Central China Fuwai Hospital, Central China Fuwai Hospital of Zhengzhou University","correspondingAuthor":false,"prefix":"","firstName":"Qiuping","middleName":"","lastName":"Zhao","suffix":""}],"badges":[],"createdAt":"2024-12-09 07:08:06","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5606509/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5606509/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":73357437,"identity":"504eaf2c-4dba-44b2-9c95-5a4d9510c707","added_by":"auto","created_at":"2025-01-09 08:20:13","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":51991,"visible":true,"origin":"","legend":"\u003cp\u003eROC curves for all models in test sets.\u003c/p\u003e\n\u003cp\u003eNotes: ROC: Receiver operating characteristic.\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-5606509/v1/48564b014302b9eb543c832e.png"},{"id":73357439,"identity":"fc3152c0-9665-47c9-b186-9f5ca338fa78","added_by":"auto","created_at":"2025-01-09 08:20:13","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":208998,"visible":true,"origin":"","legend":"\u003cp\u003eFeature importance ranking based on SHapley Additive exPlanations (SHAP) values in XGBoost.\u003c/p\u003e\n\u003cp\u003eThe left figure: importance ranking. The right figure: SHAP summary diagram. Each row is a variable, each dot represents a sample, and the color represents the value of the variable in order of highest absolute importance. The point to the left of the X-axis is the protection point, and the point to the right of the X-axis is the danger point in the right graph.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-5606509/v1/2e55c3e83c647ce26c58311c.png"},{"id":73358865,"identity":"c054e3db-bc9e-4445-807a-808d25ddf576","added_by":"auto","created_at":"2025-01-09 08:28:12","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":108246,"visible":true,"origin":"","legend":"\u003cp\u003eSHAP interaction plots between self-esteem and psychological resilience.\u003c/p\u003e\n\u003cp\u003eThe SHAP dependence plots combine the main and interaction effects of a variable, where the interaction effects, can be interpreted as how two variables affect the model output simultaneously. The x-axis represents the value for self-esteem, while the y-axis is the SHAP values of their interaction. The color of the point represents the value of psychological resilience.\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-5606509/v1/efd0cd81dcca4f37b4e0d8b3.png"},{"id":73357450,"identity":"e8ebaae8-93e2-404e-9c63-926687257dbc","added_by":"auto","created_at":"2025-01-09 08:20:15","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":182096,"visible":true,"origin":"","legend":"\u003cp\u003eThe force plot of two adolescents in the XGBoost model.\u003c/p\u003e\n\u003cp\u003eThe force plot depicts the contribution of each feature to the process of moving the value of the decision score from the base value to the value predicted by the classifier. Red denotes features that make the model score higher, blue denotes features that make the model score lower. The longer the arrow length, the more influence the features.\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-5606509/v1/e8822371d56c5aa18b797459.png"},{"id":96251105,"identity":"28a3e871-ba2c-4ee5-a9c1-f06b3777bab1","added_by":"auto","created_at":"2025-11-19 07:39:20","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1371380,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5606509/v1/84737fd5-abe2-4ec3-ac62-301d43d98263.pdf"},{"id":73357436,"identity":"29737a2d-1e85-46f5-bd43-135ac27426d7","added_by":"auto","created_at":"2025-01-09 08:20:12","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":510256,"visible":true,"origin":"","legend":"","description":"","filename":"supplementary.docx","url":"https://assets-eu.researchsquare.com/files/rs-5606509/v1/e837159d452999395e14982a.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Building Machine Learning Predictive Models for Adolescent Internet Addiction: Key Findings on Self-Esteem and Resilience Interaction","fulltext":[{"header":"Highlights","content":"\u003cp\u003e● This study developed machine learning (ML)-based predictive models to identify key risk factors for adolescent internet addiction (IA).\u003c/p\u003e\n\u003cp\u003e● Negative life events, self-esteem, school connectedness, parent-adolescent cohesion, and psychological resilience were identified as the top five predictors of IA.\u003c/p\u003e\n\u003cp\u003e● SHAP analysis highlighted a strong positive association between negative life events and IA risk. An interaction effect between self-esteem and psychological resilience showed that higher self-esteem transformed low resilience from a risk factor to a protective factor against IA.\u003c/p\u003e\n\u003cp\u003e● The integration of ML and SHAP provided clear and interpretable insights into IA risk factors, offering valuable guidance for prevention strategies.\u003c/p\u003e"},{"header":"1. Introduction","content":"\u003cp\u003eInternet addiction(IA), is defined as an increased desire to use the internet and out-of-control behavior, which is characterized by an intensified desire to reuse the internet (Tams, Legoux (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e).The internet into the daily lives of teenagers has not only brought convenience but also introduced risks. The prevalence estimates for IA exhibit significant variability, current reports in China have shown that there is persistent rise in the prevalence of IA among adolescents, indicating the incidence ranging from 2.2\u0026ndash;21.5%), which surpasses that in the United States and several European countries(\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e). In HK and Macau, about 1.9% of the college students were found to fulfil the criteria of IA(\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e). More recent studies had found that about 12% of the young adults in South Korea are addicted to the internet(\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e). Improper internet use is emerging as a public health issue that seriously influences the physical and mental health of adolescents(\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e). While currently there are no evidence-based treatments for IA, evidence suggests that IA leads to severe social and psychological damage(\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e). Hence, there is an emphasis on the importance of accurate projections and early intervention.\u003c/p\u003e \u003cp\u003eAdolescence constitutes a pivotal phase in human life, marked by the formation of habits, behaviors, and social confidence, during which individual undergoes various physical and mental transformation(\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e). Adolescents experiencing IA are commonly engaged in diverse online activities including gaming, social media usage, and online pornography(\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e). IA in adolescents has been linked to diverse psychological factors(\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e), encompassing personality traits, negative emotions, self-esteem and impulsivity(\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e). Prior studies have demonstrated a correlation between elevated IA levels and increased aggressive behavior, including thoughts of suicide and self-harm among adolescents(\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e). Despite the noted negative consequences of IA among the adolescent population in China, there is a lack of exploration into the potential burden, prevalence and predictors of IA in research studies. Accurate factors predictions for IA in adolescents and a comprehensive understanding of the underlying factors are crucial for designing timely and targeted interventions.\u003c/p\u003e \u003cp\u003ePrevious studies have employed statistical models, which are known for their stringent assumptions of a linear relationship between the outcome and explanatory variables, in modeling the predictors of IA across diverse populations. Machine Learning (ML) algorithms offer researchers powerful tools. Emphasizing algorithmic prediction in studies within this field is considered promising because ML algorithms are known for their reliability. ML algorithms provide researchers with robust tools utilized across various medical domains, including diagnosis, outcome prediction, treatment, and interpretation of medical images(\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e). Nevertheless, there remains a deficiency in research concerning ML for the risk prediction of IA. This gap is due to limited evidence supporting real-world clinical applications and the development of interpretable risk prediction models.\u003c/p\u003e \u003cp\u003eTo address these limitations, the current study used ML algorithms (Logistic Regression (LR), Random Forest (RF), Extreme Gradient Boosting Machine (XGBoost), Support Vector Machines (SVM)), combined with SHapley Additive exPlanations (SHAP), to investigate the predictions of IA in adolescents. These methodologies can assess the interpretability of the model in guiding decisions related to interventions.\u003c/p\u003e"},{"header":"2. Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1 Participants\u003c/h2\u003e \u003cp\u003eParticipants were selected between April and May 2023 using a stratified random cluster sampling approach from six junior high schools in Henan Province, China. Within each school, 8 to 10 classes were randomly chosen from each grade. A total of 8,176 valid questionnaires were collected. The protocol received approval from Zhengzhou University, and all participants provided informed consent.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2 Measures\u003c/h2\u003e \u003cdiv id=\"Sec5\" class=\"Section3\"\u003e \u003ch2\u003e2.2.1 Internet addiction\u003c/h2\u003e \u003cp\u003eA 20-item Internet Addiction Test was used to assess Internet addiction (\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e). It was evaluated by a Likert 5-point scale (ranging from 1\u0026thinsp;=\u0026thinsp;rarely to 5\u0026thinsp;=\u0026thinsp;always). Each question was summed up to calculate the total IA score. A total score above 50 was considered indicative of internet addiction (\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e) (Cronbach's α coefficient\u0026thinsp;=\u0026thinsp;0.91).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section3\"\u003e \u003ch2\u003e2.2.2 Negative life events\u003c/h2\u003e \u003cp\u003eThe Adolescent Self-rating Life Events Checklist was adopted to assess the frequency of negative life events ( 刘贤臣 (\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e), where in a higher score denotes a greater prevalence of negative life events. The Cronbach's α coefficient in the study was 0.91.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section3\"\u003e \u003ch2\u003e2.2.3 Parent-adolescent cohesion\u003c/h2\u003e \u003cp\u003eThe 10-item Parent-Adolescent Cohesion Questionnaire was employed to measure parent-child and mother-child cohesion levels(\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e). For questions 3, 4, 8, and 9, reverse coding was applied. The Cronbach's α coefficient in this questionnaire was 0.87.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section3\"\u003e \u003ch2\u003e2.2.4 School connectedness\u003c/h2\u003e \u003cp\u003eBased on prior studies (\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e), school connectedness was evaluated using 10 questions, encompassing three dimensions: teacher support (items 1, 5, and 8), school belonging (items 3, 6, and 9) and classmate support (items 2, 4, 7, and 10) (\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e). Response options ranged from 1 to 5 (1\u0026thinsp;=\u0026thinsp;strongly disagree, 2\u0026thinsp;=\u0026thinsp;strongly disagree, 3\u0026thinsp;=\u0026thinsp;uncertain, 4\u0026thinsp;=\u0026thinsp;strongly agree, 5\u0026thinsp;=\u0026thinsp;strongly agree). Items 1 and 10 were reverse coded. The higher scores indicated higher levels of student connection (Cronbach's α coefficient\u0026thinsp;=\u0026thinsp;0.85).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section3\"\u003e \u003ch2\u003e2.2.5 Psychological resilience\u003c/h2\u003e \u003cp\u003ePsychological resilience was measured by the Connor-Davidson Resilience Scale(\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e) (\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e). 10 items were assessed using a 5-point Likert scale (0\u0026thinsp;=\u0026thinsp;never, 1\u0026thinsp;=\u0026thinsp;rarely, 2\u0026thinsp;=\u0026thinsp;sometimes, 3\u0026thinsp;=\u0026thinsp;often, 4\u0026thinsp;=\u0026thinsp;always). The composite score ranged from 0 to 40, with higher scores indicating better psychological resilience. The Cronbach's α coefficient of the scale was 0.90.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section3\"\u003e \u003ch2\u003e2.2.6 Self-esteem\u003c/h2\u003e \u003cp\u003eSelf-esteem was assessed using the Self-esteem Scale developed by Rosenberg(\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e). 10 items were assessed on a 5-point Likert scale. Items 1, 2, 4, 6, and 7 were reverse coded, and higher scores indicate higher self-esteem (Cronbach's α coefficient\u0026thinsp;=\u0026thinsp;0.90).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section3\"\u003e \u003ch2\u003e2.2.7 Control variables\u003c/h2\u003e \u003cp\u003eControl variables for this study included gender, residence (rural, urban), grade (7th, 8th, 9th), family structure (intact family, others), maternal educational level (primary school and below, junior high school, senior high school, university and above), economic level (poor, moderate, good), study burden (light, moderate, heavy) and academic performance (poor, moderate, good).\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e2.3 Statistical analysis\u003c/h2\u003e \u003cp\u003eStatistical analysis and data visualization were conducted using Python version 3.7. Categorical variables were presented as numbers and proportions, while continuous variables were reported in Mean\u0026thinsp;\u0026plusmn;\u0026thinsp;SD. Binary logistic regression analysis was employed to explore factors related to IA. Additionally, the data were split into training and test sets with an 8:2 ratio, where 80% of the data were used for training the models and 20% for testing the models employing four algorithms, including Logistic LR, RF, XGBoost, and SVM. The LR model forecasts the probability of the binary dependent variable through the application of maximum likelihood estimation for ascertaining the regression coefficient. Both RF and XGBoost are algorithms grounded in tree-based learning. SVM serves as a generalized linear classifier within the framework of supervised learning. SHAP values were utilized to assess the contribution of each feature within each prediction model(\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e). \u003cem\u003eP\u003c/em\u003e\u0026lt;0.05 were considered statistically significant.\u003c/p\u003e \u003c/div\u003e"},{"header":"3. Results","content":"\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\n \u003ch2\u003e3.1 Sample characteristics\u003c/h2\u003e\n \u003cp\u003eIn total, 8176 adolescents (average age 14.42 years; 53.1% men) were included in this study. The prevalence of IA was observed in 1584 (19.4%) participants. Significant statistical differences were observed in gender, age, grade, family structure, maternal educational levels, family economic status, residence, study burden, academic performance, negative life events, psychological resilience, school connectedness, parent-adolescent cohesion, and self-esteem to IA. (See Table \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e\n \u003cp\u003e\u003c/p\u003e\u0026nbsp;\u003ctable id=\"Tab1\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eDescriptive statistics of the sample (n\u0026thinsp;=\u0026thinsp;8176)\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003eVariables\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colspan=\"2\"\u003e\n \u003cp\u003eInternet addiction\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n \u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eNo (%)/Mean\u0026thinsp;\u0026plusmn;\u0026thinsp;SD\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eYes (%)/Mean\u0026thinsp;\u0026plusmn;\u0026thinsp;SD\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTotal\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eP\u003c/em\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGender\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMen\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3576(82.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e769(17.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e4345(53.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eWomen\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3016(78.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e815(21.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3831(46.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGrade\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.017\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e7th\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2528(82.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e523(17.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3051(37.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e8th\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2290(78.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e635(21.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2925(35.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e9th\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1774(80.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e426(19.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2200(26.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eResidence\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.014\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRural\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1964(79.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e522(21.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2586(30.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eUrban\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e4628(81.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1062(18.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e5690(69.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOnly-child\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.422\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e6031(80.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1439(19.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e7470(91.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e561(79.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e145(20.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e706(8.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFamily structure\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eIntact family\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e6116(81.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1410(18.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e7526(92.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOthers\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e476(73.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e174(26.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e650(8.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMaternal educational levels\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePrimary school and below\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e517(74.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e174(25.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e691(8.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eJunior high school\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2991(81.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e695(18.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3686(45.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSenior high school\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1440(80.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e353(19.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1793(21.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eUn University and above\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1644(82.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e362(18.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2006(24.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFamily economic status\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePoor\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e363(72.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e138(27.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e501(6.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eModerate\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e5062(81.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1160(18.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e6222(76.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGood\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1167(80.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e286(19.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1453(17.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eStudy burden\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLight\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e399(83.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e81(16.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e480(5.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eModerate\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3792(85.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e636(14.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e4428(54.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHeavy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2401(73.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e867(26.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3268(40.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAcademic performance\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePoor\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1528(72.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e572(27.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2100(25.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eModerate\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3600(83.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e716(16.6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e4316(52.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGood\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1464(83.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e296(16.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1760(21.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAge(years)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e14.40\u0026thinsp;\u0026plusmn;\u0026thinsp;0.94\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e14.46\u0026thinsp;\u0026plusmn;\u0026thinsp;0.91\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e14.42\u0026thinsp;\u0026plusmn;\u0026thinsp;0.94\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.020\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNegative life events\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e39.25\u0026thinsp;\u0026plusmn;\u0026thinsp;11.76\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e52.56\u0026thinsp;\u0026plusmn;\u0026thinsp;16.39\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e41.83\u0026thinsp;\u0026plusmn;\u0026thinsp;13.83\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePsychological resilience\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e24.23\u0026thinsp;\u0026plusmn;\u0026thinsp;8.15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e19.50\u0026thinsp;\u0026plusmn;\u0026thinsp;8.03\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e23.32\u0026thinsp;\u0026plusmn;\u0026thinsp;8.34\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSchool connectedness\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e38.00\u0026thinsp;\u0026plusmn;\u0026thinsp;6.77\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e32.86\u0026thinsp;\u0026plusmn;\u0026thinsp;7.42\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e37.01\u0026thinsp;\u0026plusmn;\u0026thinsp;7.19\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eParent-adolescent cohesion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e36.29\u0026thinsp;\u0026plusmn;\u0026thinsp;8.32\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e31.47\u0026thinsp;\u0026plusmn;\u0026thinsp;8.40\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e35.36\u0026thinsp;\u0026plusmn;\u0026thinsp;8.55\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSelf-esteem\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e29.95\u0026thinsp;\u0026plusmn;\u0026thinsp;5.38\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e25.98\u0026thinsp;\u0026plusmn;\u0026thinsp;6.15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e29.18\u0026thinsp;\u0026plusmn;\u0026thinsp;5.75\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003cp\u003e\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\n \u003ch2\u003e3.2 Model evaluation\u003c/h2\u003e\n \u003cp\u003eFour ML models, LR, RF, XGBoost, and SVM were utilized to predict the occurrence of IA in adolescents (See Table \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e). XGBoost had the highest AUC (Area under the curve,0.790) and precision (0.795), and its accuracy (0.822), recall (0.608), and F1(0.792) were second high among the four ML models. Figure \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e shows the ROC (Receiver operating characteristic) curves for all models in test sets, while the ROC curves in the training set are supplied as Supplemental Fig. 1. Therefore, we selected XGBoost for further analysis.\u003c/p\u003e\n \u003cp\u003e\u003c/p\u003e\u0026nbsp;\u003ctable id=\"Tab2\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eEvaluation of the machine learning model performance.\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eAlgorithm\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eAUC\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eAccuracy\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ePrecision\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eRecall\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eF1\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eP\u003c/em\u003e\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLR\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.782\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.828\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.749\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.618\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.799\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRF\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.773\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.811\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.765\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.525\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.739\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eXGBoost\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.790\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.822\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.795\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.608\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.792\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSVM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.724\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.816\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.760\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.549\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.757\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003ctfoot\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"7\"\u003eNote: AUC: Area under the curve. a: \u003cem\u003eP\u003c/em\u003e value is the result of one-way analysis of variance for the AUC of the five models. LR: logistic. RF: Random Forest. XGBoost: Extreme Gradient Boosting. SVM: Support Vector Machine\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tfoot\u003e\n \u003c/table\u003e\n \u003cp\u003e\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\n \u003ch2\u003e3.3 SHAP model interpretation\u003c/h2\u003e\n \u003cp\u003eThe SHAP value with less than 0 indicates a negative contribution, equal to 0 indicates no contribution, and greater than 0 indicates a positive contribution. The top five features are negative life events, self-esteem, school connectedness, parent-adolescent cohesion, and psychological resilience. The higher the SHAP value of a feature, the higher the probability of developing IA. Adolescents with elevated levels of negative life events (depicted as red dots) were more prone to developing IA compared to those with lower levels (depicted as blue dots). Conversely, adolescents with low levels of self-esteem, school connectedness, parent-adolescent cohesion, and psychological resilience were more likely to develop IA (see Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e\n \u003cp\u003eSupplemental Fig. 2 depicts the top five variables of feature importance on the model output, revealing a nearly monotonic increase in local SHAP values for negative life events. The SHAP interaction plot (see Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e) demonstrates the interaction effects between self-esteem and psychological resilience. A low value for psychological resilience (depicted as blue dots) poses a risk factor. However, as self-esteem improves, a low value for psychological resilience undergoes a transition from being a risk to exhibiting a somewhat protective effect.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\n \u003ch2\u003e3.4 SHAP values of individual prediction for interpretation\u003c/h2\u003e\n \u003cp\u003eThe force plot illustrates predictions for two randomly selected adolescents No. 5 and adolescents No. 6692, explaining the individual predictions in this study. The function f(x) represents the model output, indicating the predicted probability for each adolescent, while E[f(X)] means the average of the model predictions. Adolescents No. 5, a boy from rural, demonstrates a low risk of IA (-0.023) attributed to protective factors, including self-esteem (\u003cspan class=\"CitationRef\"\u003e10\u003c/span\u003e), psychological resilience (\u003cspan class=\"CitationRef\"\u003e21\u003c/span\u003e), grade (\u003cspan class=\"CitationRef\"\u003e9\u003c/span\u003e), school connectedness (\u003cspan class=\"CitationRef\"\u003e12\u003c/span\u003e), rural residence, and moderate family economic status (see Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e(a)).\u003c/p\u003e\n \u003cp\u003eIn contrast, Adolescent No. 6692, a girl diagnosed with IA in the study, exhibits a high probability of IA (0.442) due to risk factors such as negative life events (55), parent-adolescent cohesion (\u003cspan class=\"CitationRef\"\u003e33\u003c/span\u003e), and good academic performance. (see Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e(b)).\u003c/p\u003e\n\u003c/div\u003e"},{"header":"4. Discussion","content":"\u003cp\u003eIn this study, we proposed a prediction model developed based on ML algorithms to accurately identify IA in adolescents. Our study provides a meaningful explanation based on the SHAP model. As previously described, the incidence of IA varied significantly under the influence of different social, cultural, and economic backgrounds(\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e). After the COVID-19 pandemic, IA may become a common problem for society, particularly affecting adolescents(\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e). Hence, reducing prevalence has become a crucial goal in the management of IA. In the present investigation, comprising 8716 adolescents, and ML prediction models were constructed utilizing 13 distinct features. The significance of a prediction is contingent upon its precision, thereby providing substantial contributions to clinical application.\u003c/p\u003e \u003cp\u003eOur study suggests that compared with other features, psychological features play a more significant role in the ML prediction of IA in adolescents. Specifically, the top five features in the prediction models are negative life events, self-esteem, school connectedness, parent-adolescent cohesion and psychological resilience. The ranking of variable importance based on SHAP values revealed that negative life events were the most significant factor influencing IA, a finding that has been rarely reported in previous literature on IA. More specifically, compared to adolescents with lower negative life events, adolescents with higher negative life events were more likely to experience social media addiction. These results indicated that negative life events may play an important role in IA(\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eHowever, our results indicate that lower self-esteem, school connectedness, and parent-adolescent cohesion are associated with the likelihood of IA occurrence. Studies have shown an association between IA and lower self-esteem(\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e), aligning with our findings. Additionally, lower school connectedness is related to increased IA has been reported(\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e). It seems plausible that the phenomenon is linked to the promotion of school connectedness, which enhances children's sense of belonging within the school environment. Improving positive interactions between teachers and students may reduce the risks of IA among adolescents. Furthermore, a study has reported parent\u0026ndash;adolescent cohesion association with IA, which is consistent with our research. Worthy to pay attention, these points toward different psychological factors are closely related to IA in adolescents(\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e). The research found that adolescents with psychosocial problems are more prone to IA. A study from Turkey showed that the risk of psychosocial problems in adolescents was 19.8% and 18.8%, respectively(\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e). It was worth noting that the current findings were aligned with existing evidence suggesting that participants with poor mental health were more likely to cause IA.\u003c/p\u003e \u003cp\u003eMore importantly, the SHAP model can be used to investigate the individual effects of risk factors and their interaction effects. The SHAP dependence plot depicts the top five variables of feature importance on the model output is presented in Supplementary Fig.\u0026nbsp;2, revealing a discernible positive correlation between negative life events and IA. Moreover, our study proves the observed SHAP interaction between self-esteem and psychological resilience in adolescents. As self-esteem improves, a lower value for psychological resilience undergoes a shift from being a risk factor to displaying a slightly protective effect.\u003c/p\u003e \u003cp\u003eTo provide a detailed explanation and interpretation of IA prediction, we presented an individualized explanation of the model predictions through a force plot, illustrating predictions for two randomly selected adolescents. Remarkably, our ranking of variable importance closely aligns with observed differences in variables between subjects with and without IA. For example, adolescents without IA tend to exhibit higher levels of self-esteem, psychological resilience, and school connectedness. Conversely, Adolescent No. 6692 is associated with a high probability of IA risk (f(x)\u0026thinsp;=\u0026thinsp;0.442) due to factors that elevate the prediction, including higher negative life events, lower parent-adolescent cohesion, good academic performance, and a tendency to be in grade 9 and reside in a rural area.\u003c/p\u003e \u003cp\u003eThe study contains the following advantages. To date, no predictive models for IA in adolescents have been developed using ML algorithms and SHAP models. Unlike previous research in China, which predominantly relies on traditional regression models, our study represents an innovative approach to addressing the challenges associated with predicting IA in adolescents (\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e). The utilization of advanced ML models holds the potential to significantly enhance prediction accuracy. The application of ML algorithms to the medical field is a new trend, demonstrated in studies exploring connections between posttraumatic stress disorder (PTSD) and emotion regulation(\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e), predicting alcohol use(\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e) in adolescents, and other clinical applications(\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e). Additionally, the interaction analysis between self-esteem and psychological resilience is a novel approach to examining the interplay between variables for IA.\u003c/p\u003e \u003cp\u003eHowever, it is imperative to acknowledge the limitations of this study, when interpreting its findings. Firstly, Firstly, adolescents answered self-report questionnaires have recall bias. Secondly, analysis of cross-sectional data sets cannot be used to draw arbitrary conclusions. Finally, other factors affecting IA in adolescents, such as genetic factors(\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e), were not taken into account in the current study. Future research endeavors may incorporate genetic factors and other potential molecular level predictors, providing insights into the role of genetic mechanisms in IA.\u003c/p\u003e"},{"header":"5. Conclusion","content":"\u003cp\u003eIn summary, our study combined the ML models and the explanation model to reliably predict the risk of IA in adolescents. This approach could assist physicians in intuitively understanding the influence of key features and detecting IA risks early by observing signs of psychological health in individuals. The findings have the potential to promote the identification of IA factors and the development of subsequent strategies for the follow-up care of adolescents.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eConflicts of Interest\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that there are no conflicts of interest.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eHuman Ethics and Consent to Participate declarations\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthical approval\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe protocol received approval from Zhengzhou University.\u003cbr\u003e\u003cstrong\u003eInformed consent\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll participants provided informed consent.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eDue to the confidentiality of the data, our data will not be publicly released. Our data will be provided by the corresponding author if required.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding Statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eCollaborative Innovation System Research on Drug Intervention \u0026amp; Non-drug Intervention in Proactive Health Context (20220518A); Research on Cardiovascular Disease Screening and Healthy Lifestyle Intervention in Children and Adolescents(20230014B) and Platform for Dynamic Monitoring and Comprehensive Evaluation of Healthy Central Plains Action (20220134B).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors wish to express their gratitude to all investigators and all participants and thank all their colleagues for their valuable inputs to the study design and data collection.\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eTams S, Legoux R, Leger PM. Smartphone withdrawal creates stress: A moderated mediation model of nomophobia, social threat, and phone withdrawal context. Computers in Human Behavior 2018;81:1-9.\u003c/li\u003e\n\u003cli\u003eMihara S, Higuchi S. Cross-sectional and longitudinal epidemiological studies of Internet gaming disorder: A systematic review of the literature. Psychiatry Clin Neurosci 2017;71:425-444.\u003c/li\u003e\n\u003cli\u003eDing YJ, Lau CH, Sou KL, et al. Association between internet addiction and high-risk sexual attitudes in Chinese university students from Hong Kong and Macau. Public health 2016;132:60-63.\u003c/li\u003e\n\u003cli\u003eNa E, Lee H, Choi I, et al. Comorbidity of Internet gaming disorder and alcohol use disorder: A focus on clinical characteristics and gaming patterns. The American journal on addictions 2017;26:326-334.\u003c/li\u003e\n\u003cli\u003eChristakis DA. Internet addiction: a 21st century epidemic? BMC Medicine 2010;8.\u003c/li\u003e\n\u003cli\u003eBoer M, Stevens G, Finkenauer C, et al. Attention Deficit Hyperactivity Disorder-Symptoms, Social Media Use Intensity, and Social Media Use Problems in Adolescents: Investigating Directionality. Child development 2020;91:e853-e865.\u003c/li\u003e\n\u003cli\u003eKaraer Y, Akdemir D. Parenting styles, perceived social support and emotion regulation in adolescents with internet addiction. Comprehensive psychiatry 2019;92:22-27.\u003c/li\u003e\n\u003cli\u003eSeider S, Jayawickreme E, Lerner RM. Theoretical and Empirical Bases of Character Development in Adolescence: A View of the Issues. Journal of youth and adolescence 2017;46:1149-1152.\u003c/li\u003e\n\u003cli\u003eGriffiths MD, van Rooij AJ, Kardefelt-Winther D, et al. Working towards an international consensus on criteria for assessing internet gaming disorder: a critical commentary on Petry et al. (2014). Addiction (Abingdon, England) 2016;111:167-175.\u003c/li\u003e\n\u003cli\u003eZhang W, Pu J, He R, et al. Demographic characteristics, family environment and psychosocial factors affecting internet addiction in Chinese adolescents. Journal of affective disorders 2022;315:130-138.\u003c/li\u003e\n\u003cli\u003eGao YX, Wang JY, Dong GH. The prevalence and possible risk factors of internet gaming disorder among adolescents and young adults: Systematic reviews and meta-analyses. Journal of psychiatric research 2022;154:35-43.\u003c/li\u003e\n\u003cli\u003eKuang L, Wang W, Huang Y, et al. Relationship between Internet addiction, susceptible personality traits, and suicidal and self-harm ideation in Chinese adolescent students. Journal of behavioral addictions 2020;9:676-685.\u003c/li\u003e\n\u003cli\u003eRajkomar A, Dean J, Kohane I. Machine Learning in Medicine. The New England journal of medicine 2019;380:1347-1358.\u003c/li\u003e\n\u003cli\u003eYoung KS. Internet Addiction: The Emergence of a New Clinical Disorder. Mary Ann Liebert, Inc 1998.\u003c/li\u003e\n\u003cli\u003eZhang J H, Zhang X Q, Lu X Y, et al. The mediating role of depression in the relationship between childhood abuse and Internet addiction in adolescents. Chinese Journal of Health Psychology 2022;30:6.\u003c/li\u003e\n\u003cli\u003eLiu Xianchen. Development and reliability and validity test of adolescent life Events Scale. Shandong Psychiatry 1997;10: 5.\u003c/li\u003e\n\u003cli\u003eZhang W X, Wang M P, Andrew, et al. Adolescents\u0026apos; expectation of autonomy, attitude towards parental authority and parent-child conflict and affinity. Journal of Psychology 2006.\u003c/li\u003e\n\u003cli\u003eMcneely CA, Nonnemaker JM, Blum RW. Promoting School Connectedness: Evidence from the National Longitudinal Study of Adolescent Health. (Research Papers). Journal of School Health 2002; 72.\u003c/li\u003e\n\u003cli\u003eYu C F, Zhang W, Zeng Y Y, et al. The relationship between gratitude and problem behavior in adolescents: the mediating role of school bonding. Psychological Development and Education 2011;27;9. \u003c/li\u003e\n\u003cli\u003eCampbell-Sills L, Stein MB. Psychometric analysis and refinement of the Connor-davidson Resilience Scale (CD-RISC): Validation of a 10-item measure of resilience. Journal of Traumatic Stress 2010;20:1019-1028.\u003c/li\u003e\n\u003cli\u003eConnor KM, Davidson JRT. Development of a new resilience scale: The Connor‐Davidson Resilience Scale (CD‐RISC). Depression and Anxiety 2003;18.\u003c/li\u003e\n\u003cli\u003eRosenberg M. Rosenberg Self-Esteem Scale (RSES). . APA PsycTests 1965.\u003c/li\u003e\n\u003cli\u003eLundberg S, Lee SI. A Unified Approach to Interpreting Model Predictions. Nips; 2017; 2017.\u003c/li\u003e\n\u003cli\u003eTwh C, Smy S, Mwl C. Adolescent Internet Addiction in Hong Kong: Prevalence, Psychosocial Correlates, and Prevention. The Journal of adolescent health : official publication of the Society for Adolescent Medicine 2019;64:S34.\u003c/li\u003e\n\u003cli\u003eLi YY, Sun Y, Meng SQ, et al. Internet Addiction Increases in the General Population During COVID-19: Evidence From China. The American journal on addictions 2021;30:389-397.\u003c/li\u003e\n\u003cli\u003eWang X, Ding T, Lai X, et al. Negative Life Events, Negative Copying Style, and Internet Addiction in Middle School Students: A Large Two-year Follow-up Study. International journal of mental health and addiction 2023:1-11.\u003c/li\u003e\n\u003cli\u003eSevelko K, Bischof G, Bischof A, et al. The role of self-esteem in Internet addiction within the context of comorbid mental disorders: Findings from a general population-based sample. Journal of behavioral addictions 2018;7:976-984.\u003c/li\u003e\n\u003cli\u003eLiu S, Yu C, Conner BT, et al. Autistic traits and internet gaming addiction in Chinese children: The mediating effect of emotion regulation and school connectedness. Research in developmental disabilities 2017;68:122-130.\u003c/li\u003e\n\u003cli\u003eOzturk F, Ayaz-Alkaya S. Internet addiction and psychosocial problems among adolescents during the COVID-19 pandemic: A cross-sectional study. Archives of psychiatric nursing 2021;35:595-601.\u003c/li\u003e\n\u003cli\u003eZhang X, Zhang J, Zhang K, et al. Effects of different interventions on internet addiction: A meta-analysis of random controlled trials. Journal of affective disorders 2022;313:56-71.\u003c/li\u003e\n\u003cli\u003eChrist NM, Elhai JD, Forbes CN, et al. A machine learning approach to modeling PTSD and difficulties in emotion regulation. Psychiatry research 2021;297:113712.\u003c/li\u003e\n\u003cli\u003eAfzali MH, Sunderland M, Stewart S, et al. Machine-learning prediction of adolescent alcohol use: a cross-study, cross-cultural validation. Addiction (Abingdon, England) 2019;114:662-671.\u003c/li\u003e\n\u003cli\u003eLi W, Wang J, Liu W, et al. Machine Learning Applications for the Prediction of Bone Cement Leakage in Percutaneous Vertebroplasty. Frontiers in public health 2021;9:812023.\u003c/li\u003e\n\u003cli\u003eTereshchenko S, Kasparov E. Neurobiological Risk Factors for the Development of Internet Addiction in Adolescents. Behavioral sciences (Basel, Switzerland) 2019;9.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"adolescents, internet addiction, machine learning, SHAP, self-esteem, psychological resilience","lastPublishedDoi":"10.21203/rs.3.rs-5606509/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5606509/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eObjective:\u003c/strong\u003e Internet addiction (IA) is a significant mental health concern among adolescents. This study aimed to develop machine learning (ML)-based predictive models to identify and explain key risk factors for IA.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethod: \u003c/strong\u003eA total of 8176 junior high school students from Henan Province were surveyed from April to May 2023. The dataset was randomly divided into training and test sets in an 8:2 ratio. Four ML algorithms were used to predict IA, and feature importance was determined using SHapley Additive exPlanations (SHAP). The XGBoost model, which achieved the highest area under the curve (AUC), was selected for detailed analysis and individualized prediction explanations.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults:\u003c/strong\u003e The five most important predictors of IA were negative life events, self-esteem, school connectedness, parent-adolescent cohesion, and psychological resilience. Importantly, an interaction effect was found between self-esteem and psychological resilience: as self-esteem increased, the influence of low resilience transitioned from being a risk factor to a protective factor against IA.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusion:\u003c/strong\u003e This study demonstrates the power of ML models combined with SHAP for predicting IA and identifying its psychosocial determinants. The findings highlight the critical interplay of self-esteem and psychological resilience, offering valuable insights for clinicians and educators in addressing IA among adolescents.\u003c/p\u003e","manuscriptTitle":"Building Machine Learning Predictive Models for Adolescent Internet Addiction: Key Findings on Self-Esteem and Resilience Interaction","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-01-09 08:20:07","doi":"10.21203/rs.3.rs-5606509/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"d3accdca-2108-4ba1-a63c-1f7ca10ba72c","owner":[],"postedDate":"January 9th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":42486348,"name":"Humanities/Medical humanities"},{"id":42486349,"name":"Social science/Psychology"}],"tags":[],"updatedAt":"2025-11-18T14:24:34+00:00","versionOfRecord":[],"versionCreatedAt":"2025-01-09 08:20:07","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-5606509","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5606509","identity":"rs-5606509","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}