Internet addiction and anxiety-depressive comorbidity among chinese adolescents: the mitigating effects of physical activity in a large-scale cross-sectional study

Abstract

Objective

This study aimed to examine the associations between internet addiction (IA), physical activity (PA), and mental health comorbidities (depression, anxiety, and their co-occurrence) among Chinese adolescents, and assess the mediating role of PA in the IA-mental health relationship.

Methods

A cross-sectional analysis was conducted using data from the 2024 National Student Common Diseases and Risk Factors Surveillance (NSCDRFS) in Tianjin, China, involving 20,692 adolescents aged 12–19 years. IA was assessed via a DSM-5-derived 9-item scale (cutoff ≥ 5). PA levels were categorized as insufficient (≤ 2 days/week of moderate-to-vigorous activity). Mental health outcomes included depressive symptoms (CES-D ≥ 16), anxiety symptoms (GAD-7 ≥ 10), and anxiety-depression comorbidity (ADC), defined as meeting both depression (CES-D ≥ 16) and anxiety (GAD-7 ≥ 10) diagnostic criteria. Multivariable logistic regression and mediation analyses were performed to evaluate direct and indirect pathways.

Results

Adolescents with IA exhibited significantly higher risks of anxiety (OR = 1.80, 95% CI: 1.68–1.91), depression (OR = 1.99, 95% CI: 1.88–2.11), and ADC (OR = 2.12, 95% CI: 2.00–2.25) compared to non-IA peers (all P < 0.001). Insufficient PA independently increased risks of anxiety (OR = 1.17), depression (OR = 1.27), and ADC (OR = 1.23). Mediation analysis revealed that PA accounted for only 3.5–4.3% of the total effects of IA on mental health outcomes, with direct effects remaining predominant (95.7–96.5%). Subgroup analyses highlighted stronger IA-mental health associations in younger adolescents (12–15 years), non-boarding students, and junior high school cohorts.

Conclusion

IA is an independent risk factor for mental health comorbidities in Chinese adolescents, with PA serving as a minor yet significant mediator. Targeted interventions should integrate IA screening, culturally adapted PA promotion, and digital literacy education to address this triad. Future longitudinal studies are needed to elucidate causal pathways and neurodevelopmental mechanisms.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-025-24137-6.

Introduction

Internet addiction (IA) has emerged as a major public health issue among adolescents globally, driven by the widespread availability and usage of digital technologies [1–3]. The prevalence of Internet addiction among European adolescents has been estimated to average 4.4%, however, East Asian and Middle Eastern countries have significantly higher prevalence rates of Internet addiction, ranging from 8.1 to 26.5% [4, 5]. The 2023 report on Internet Use of Chinese Minors showed that the prevalence of Internet addiction among adolescents was close to 10%, with 12–16 years old being the high-risk group [6]. Factors such as easy access to the internet, social appeal, and poorer self-regulation in adolescents contribute to increased vulnerability to IA [7, 8]. This issue may have been further amplified by pandemic conditions, as lockdown measures increasing time at home and limiting social interaction inherently heighten adolescents’ exposure and reliance on digital connectivity [9–11]. Consequently, Dong et al. empirically demonstrated that Chinese youth developed significantly heightened vulnerability to internet addiction during the COVID-19 pandemic, substantiating its critical impact on both behavioral pathology and mental health outcomes [12].

Adolescent IA has been demonstrated to associate with multiple adverse mental health problems in adolescents, including depression, anxiety, stress, and suicidal ideation [13, 14]. Previous studies have identified significant linkages between internet addiction and adverse mental health outcomes. A study by Peng highlighted that adolescents with internet addiction exhibited higher levels of anxiety and depression symptoms compared to their non-addicted peers [15]. Moreover, problematic online behaviors, such as cyberbullying, exposure to cyber pornography, and internet fraud can serve as critical facilitators of IA, which further deteriorating mental well-being [16–18]. Research by Wang indicated that cyberbullying victimization is associated with increased internet addiction, which in turn exacerbates symptoms of depression and anxiety in youth [19].

Concurrently, empirical studies have reported an inverse relationship between IA and physical activity (PA), indicating that excessive internet engagement may significantly diminish adolescents’ participation in physical activities [20]. This reduced PA is further implicated in physical health complications, including obesity and cardiovascular diseases, thereby negatively affecting adolescent mental health [21]. The relationship between IA, PA, and mental health can be explained through two primary theoretical lenses. First, the displacement hypothesis posits that excessive internet use displaces time that could otherwise be spent on PA. This theoretical perspective suggests that the more time adolescents spend online, the less time they have for physical activities, which are crucial for mental health [22]. Second, the self-regulation theory highlights that internet addiction may impair an individual’s ability to regulate emotions and behaviors. Adolescents with IA may have difficulty engaging in self-regulated activities like PA, which can exacerbate mental health issues [23]. Conversely, PA has been shown to enhance self-regulation through neurobiological mechanisms such as increased BDNF levels and improved prefrontal cortex function, which can mitigate the negative effects of IA [24]. However, few studies have investigated the association of IA and PA with mental health comorbidities among Chinese adolescents, and those existing studies typically involve small sample sizes. Furthermore, the potential mediating role of PA and the extent of its mediation in the relationship between IA and mental health remains unclear.

To address these critical gaps, utilizing a large dataset comprising 20,692 adolescents, our study aims to comprehensively investigate the interplay among IA, PA, and mental health comorbidities. Specifically, this study seeks to: (1) examine the associations between IA severity and the risks of depression, anxiety, and their co-occurrence; (2) investigate the associations between PA levels and depression, anxiety, and their co-occurrence; and (3) assess the mediating role of PA in the relationship between IA and adolescent mental health outcomes.

Methods

Study population

This cross-sectional study utilized data from the 2024 National Student Common Diseases and Risk Factors Surveillance (NSCDRFS) in Tianjin, China, covering all 16 districts (9 urban, 7 rural). A stratified random cluster sampling method was implemented in September 2024 to ensure representativeness across educational stages and geographic regions. In urban districts, two junior high schools, two senior high schools, and one vocational high school were randomly selected per district, while rural counties included two junior high schools and one senior high school per county. Within each school, two or more classes per grade were randomly chosen using random number tables, with full enrollment of selected classes required to meet a minimum threshold of 100 students per grade. Neighboring schools of equivalent type were recruited as supplements if sample quotas were unmet. From the initial 21,329 recruited participants, we excluded 637 for the following reasons: (a) 92 participants were excluded due to incomplete or invalid responses; (b) 87 had incomplete mental health scale responses (> 20% missing items on CES-D/GAD-7); (c) 458 participants with missing covariates. The final analytical sample comprised 20,692 children and adolescents with complete physical examination records (Fig. 1), all of whom provided written informed consent (or parental consent for minors). Ethical approval was granted by the Tianjin Center for Disease Control and Prevention Ethics Committee (approval number: TJCDC-R-2023-001).

Fig. 1.

Flowchart of the participants

Measures

Assessment of internet addiction

Internet addiction was evaluated using a 9-item scale adapted from DSM-5 criteria for internet gaming disorder, with participants reporting symptom presence over the preceding week [25]. Affirmative responses were summed (range: 0–9), and a threshold of ≥ 5 positive items defined internet addiction [26] (Cronbach’s α = 0.79) (Table S1), with symptoms assessed over the preceding 7-day period.

Assessment of physical activity level

Physical activity levels were assessed using a self-reported questionnaire. This instrument included two items assessing moderate-to-vigorous physical activity (MVPA) over the preceding 7 days: frequency (“How many days did you engage in [MVPA]?“) and duration (“On average, how many minutes did you spend per session?“). MVPA was defined according to cited standards as activities inducing breathlessness and/or elevating heart rate to 60–70% of maximum capacity (e.g., running, basketball, swimming) [27]. Based on participant responses, total weekly minutes of MVPA were calculated.

Participant categorization strictly followed World Health Organization guidelines defining insufficient physical activity (IPA) as ≥ 60 min of MVPA achieved on ≤ 2 days weekly [28]. Based on two components of this instrument (days engaged and mean session duration), participants were dichotomized into two groups: those meeting IPA criteria constituted the IPA group, while those exceeding the minimum threshold comprised the adequate moderate-to-vigorous physical activity (MVPA) group.

Ascertainment of mental health outcomes

Depressive symptoms were assessed via the 20-item Center for Epidemiologic Studies Depression Scale (CES-D) [29], evaluating symptom frequency over the past 2 weeks, scored on a 4-point Likert scale (0 = rarely to 3 = mostly) with a clinical cutoff of ≥ 16 (Cronbach’s α = 0.82).

Anxiety symptoms were assessed using the 7-item Generalized Anxiety Disorder Scale (GAD-7) [30], with items referencing the prior 2-week period, and scored from 0 (not at all) to 3 (almost daily). The GAD-7 and its cutoff score of ≥ 10 have been well validated for identifying anxiety symptoms.

Anxiety-Depression Comorbidity Meeting both positive anxiety symptoms and positive depression symptoms was defined as anxiety-depression comorbidity (ADC).

Covariates

Covariates included age (12–15 years, 16–19 years), gender (boy, girl), education level (middle school, high school), residence (urban, rural), boarding status (categorized as “yes” if residing in school dormitories ≥ 4 nights/week during academic terms, based on official school registry records; otherwise “no”). Anthropometric measurements included height (recorded to 0.1 cm precision using mechanical stadiometers) and weight (measured to 0.1 kg accuracy via electronic scales). Body mass index (BMI = weight (kg)/height (m)²) was calculated, and the population was divided into binary variables (obese and non-obese) according to China’s Screening Standard for Overweight and Obesity in School-Age Children and Adolescents (WS/T 586–2018) [31].

Statistical analysis

In the analysis, categorical variables were presented as frequencies and relative percentages (%), the continuous variable (“Age, years”) was tested for normal distribution and presented as means (standard deviations). Chi-square test and the t test were used for descriptive statistical analysis to summarize the differences in the prevalence of anxiety, depressive and ADC among people with different sociological characteristics. Multivariable logistic regression models were hierarchically constructed across three adjustment tiers: model 1 adjusted for gender, age group, education level, residence, boarding status; Model 2 added adjustment for obesity. Model 3 further adjusted for IAD and PA simultaneously from Model 2, and odds ratio (OR) and 95% confidence interval (95%CI) were used to predict the magnitude of risk. Subgroup analysis and interaction test were conducted by grouping variables such as age, education level and boarding status, and the significance of interaction terms was evaluated by likelihood ratio test. Mediation analysis was used to test the mediating effect of physical activity (PA) on IA and mental health. A counterfactual framework was used to decomposition the direct effect (IA→ outcome) and indirect effect (IA→PA→ outcome), and the proportion of effect (indirect effect/total effect) was calculated by Bootstrap with 1000 times. All analyses were performed using R 4.3.0, with statistical significance defined as a two-tailed P-value < 0.05.

Result

Differences in the prevalence of health outcomes across characteristics

Table 1 summarizes the prevalence of anxiety, depression, and ADC across sociodemographic and health-related factors. Older adolescents (15–18 years) exhibited significantly higher rates of all outcomes compared to younger peers (11–14 years) (e.g., anxiety: 24.2% vs. 14.6%, P < 0.001). Girls had elevated prevalence relative to girls (anxiety: 22.7% vs. 16.4%; depression: 17.2% vs. 12.9%, both P < 0.001). High school students reported higher rates of anxiety (24.5% vs. 14.9%), depression (18.6% vs. 11.6%), and ADC (13.8% vs. 8.5%) than middle school students (all p < 0.001). Boarding students, those from single-parent families, and individuals with internet addiction (IA) demonstrated disproportionately higher prevalence across outcomes (e.g., IA: anxiety 32.5% vs. 18.3%, P < 0.001). Urban-rural differences were limited to depression (urban: 16.0% vs. rural: 13.7%, P < 0.001), while physical activity and obesity showed mixed associations. All analyses met statistical significance (P < 0.05), with key findings robust at P < 0.001.

Table 1.

Differences in the prevalence of health outcomes across characteristics

	Overall, N (%)	Anxiety, N (%)	P	Depression, N (%)	P	ADC, N (%)	P
Age (years), mean(SD)	14.9(1.7)	15.3(1.8)	< 0.001	15.3(1.7)	< 0.001	15.3(1.7)	< 0.001
11–14	10,092 (48.8)	1469 (14.6)	< 0.001	1156 (11.5)	< 0.001	839 (8.3)	< 0.001
15–18	10,600 (51.2)	2570 (24.2)		1949 (18.4)		1450 (13.7)
Gender
Boy	10,497 (50.7)	1721 (16.4)	< 0.001	1352 (12.9)	< 0.001	945 (9.0)	< 0.001
Girl	10,195 (49.3)	2318 (22.7)		1753 (17.2)		1344 (13.2)
Residence
Urban	11,622 (56.2)	2320 (20.0)	0.077	1861 (16.0)	< 0.001	1328 (11.4)	0.059
Rural	9070 (43.8)	1719 (19.0)		1244 (13.7)		961 (10.6)
Boarding status
Yes	1677 (8.1)	489 (29.2)	< 0.001	333 (19.9)	< 0.001	256 (15.3)	< 0.001
No	19,015 (91.9)	3550 (18.7)		2772 (14.6)		2033 (10.7)
Education level
Middle school	10,092 (48.8)	1585 (14.9)	< 0.001	1239 (11.6)	< 0.001	907 (8.5)	< 0.001
High school	10,600 (51.2)	2454 (24.5)		1866 (18.6)		1382 (13.8)
Internet addiction(IA)
Yes	1722 (8.3)	560 (32.5)	< 0.001	475 (27.6)	< 0.001	375 (21.8)	< 0.001
No	18,970 (91.7)	3479 (18.3)		2630 (13.9)		1914 (10.1)
Physical activity(PA)
IPA	9673 (46.7)	1918 (17.4)	< 0.001	1422 (12.9)	< 0.001	1053 (9.6)	< 0.001
MVPA	11,019 (53.3)	2121 (21.9)		1683 (17.4)		1236 (12.8)
Obesity
Yes	4375 (21.1)	812 (18.6)	0.071	661 (15.1)	0.847	488 (11.2)	0.843
No	16,317 (78.9)	3227(39.8)		2444 (15.0)		1801 (11.0)

ADC anxiety-depression comorbidity, IPA insufficient physical activity, MVPA moderate-to-vigorous physical activity

Multivariable logistic regression analyses

Multivariable logistic regression analyses demonstrated robust associations between IA, PA, and mental health outcomes among adolescents (Table 2). After adjusting for sociodemographic factors (Model 1), adolescents with IA exhibited significantly higher risks of anxiety (OR = 1.80, 95% CI:1.68–1.91), depression (OR = 1.99, 95% CI:1.88–2.11), and ADC (OR = 2.12, 95% CI:2.00–2.25) compared with their non-IA counterparts (all P < 0.001). These associations remained robust following additional adjustment for obesity (Model 2) and simultaneous inclusion of physical activity (Model 3), with only minimal attenuation in effect estimates (e.g., anxiety: OR = 1.77, 95% CI:1.66–1.88; depression: OR = 1.95, 95% CI:1.84–2.07; ADC: OR = 2.09, 95% CI:1.96–2.21; all P < 0.001).

Table 2.

Logistic regression analysis of the association between internet addiction, physical activity and anxiety, depression, ADC

	model1	model2	model3
	OR (95% CI)	OR (95% CI)	OR (95% CI)
Anxiety
Internet addiction (IA)
No	1 (ref)	1 (ref)	1 (ref)
Yes	1.80***(1.68, 1.91)	1.79***(1.68, 1.91)	1.77***(1.66, 1.88)
Physical activity(PA)
MVPA	1 (ref)	1 (ref)	1 (ref)
IPA	1.17***(1.10, 1.24)	1.17***(1.10, 1.24)	1.15***(1.08, 1.22)
Depression
Internet addiction (IA)
No	1 (ref)	1 (ref)	1 (ref)
Yes	1.99***(1.88, 2.11)	1.99***(1.87, 2.11)	1.95***(1.84, 2.07)
Physical activity(PA)
MVPA	1 (ref)	1 (ref)	1 (ref)
IPA	1.27***(1.19, 1.34)	1.26***(1.18, 1.34)	1.24***(1.16, 1.32)
anxiety-depression comorbidity(ADC)
Internet addiction (IA)
No	1 (ref)	1 (ref)	1 (ref)
Yes	2.12***(2.00, 2.25)	2.11***(1.99, 2.25)	2.09***(1.96, 2.21)
Physical activity(PA)
MVPA	1 (ref)	1 (ref)	1 (ref)
IPA	1.23***(1.14, 1.32)	1.23***(1.14, 1.32)	1.20***(1.11, 1.29)

model1 adjusted for gender, age group, education level, residence, boarding status. model2: further adjusted for obesity, based on model (1) model3, included IAD and PA simultaneously, based on model (2) 

***P<0.001

Insufficient physical activity (IPA) was independently associated with elevated risks of anxiety (OR = 1.17, 95% CI:1.10–1.24), depression (OR = 1.27, 95% CI:1.19–1.34), and ADC (OR = 1.23, 95% CI:1.14–1.32) relative to moderate-to-vigorous PA (MVPA) in fully adjusted models (all P < 0.001). The inclusion of IA and PA in Model 3 did not substantially alter these relationships, indicating stable and independent contributions of both factors to mental health outcomes.

Subgroup analysis of the association between internet addiction and mental health outcomes

IA demonstrated a robust association with incident anxiety symptoms, depressive symptoms, and ADC (All P < 0.001) across distinct demographic strata.

Anxiety Symptom

Stratified analyses revealed marked effect heterogeneity across age groups, with a heightened association in younger adolescents (12–15 years: OR = 2.59) compared to their older counterparts (16–19 years: OR = 1.68; interaction P = 0.001). While both genders exhibited significant associations (boys: OR = 2.02; girls: OR = 2.18), no significant effect modification by gender was observed (P = 0.49). Geographical residency further differentiated the relationship, with rural populations showing a more pronounced linkage (OR = 2.39 vs. urban: OR = 2.05; P = 0.218). Notably, non-boarding students manifested stronger effects (OR = 2.20) relative to boarding peers (OR = 1.46; interaction P = 0.009), as did junior high school cohorts (OR = 2.66) compared to senior high school groups (OR = 1.64; interaction P < 0.001; Figure S1).

Depressive Symptoms

Age-specific analyses again highlighted elevated vulnerability in early adolescence (12–15 years: OR = 2.76 vs. 16–19 years: OR = 1.93; interaction P = 0.011). Gender-stratified estimates demonstrated comparable associations (boys: OR = 2.26; girls: OR = 2.38; interaction P = 0.654). Strikingly, boarding status emerged as a critical modifier, with non-boarding students exhibiting nearly twofold greater odds (OR = 2.54) than their boarding counterparts (OR = 1.31; interaction P < 0.001). Educational stage further amplified disparities, as junior high school students displayed a 2.84-fold risk vs. 1.88-fold in senior high school cohorts (interaction P = 0.002; Figure S2).

Anxiety-Depressive Comorbidity (ADC)

The IA-ADC association was most pronounced in younger adolescents (12–15 years: OR = 2.95 vs. 16–19 years: OR = 2.01; interaction P = 0.013). Non-boarding enrollment (OR = 2.65 vs. boarding: OR = 1.40; interaction P = 0.001) and junior educational level (OR = 3.11 vs. senior: OR = 1.96; interaction P = 0.002; Figure S3) remained robust modifiers, paralleling trends observed in anxiety and depression phenotypes.

Collectively, IA consistently correlated with adverse mental health outcomes across all subgroups, though effect magnitudes exhibited marked heterogeneity, being substantially modified by age, educational stage, and boarding status. These findings underscore the importance of targeted preventive strategies tailored to high-risk adolescent subpopulations.

Mediation effect analysis

The mediatory pathway of physical activity (PA) in the association between internet addiction (IA) and mental health outcomes—encompassing anxiety symptoms, depressive symptoms, and ADC—was examined through formal mediation modeling. Analysis revealed statistically significant yet modest mediation effects attributable to PA across all three mental health phenotypes.

Anxiety Symptoms

PA mediated a small yet statistically significant proportion of the total effect (indirect effect: β = 0.006, 95% CI: 0.004–0.008), accounting for 4.15% of the total variance. The direct effect of IA remained robust (β = 0.136, 95% CI: 0.113–0.159), contributing 95.85% to the overall effect (total effect: β = 0.142, 95% CI: 0.118–0.165; Fig. 2).

Fig. 2.

Mediation Analysis of the Relationship Between IA and Anxiety Symptoms with PA as the Mediator

Depressive Symptoms

A parallel pattern emerged for depressive symptoms, with PA mediating 4.29% of the total effect (indirect effect: β = 0.006, 95% CI: 0.004–0.008). The direct pathway predominated (β = 0.131,95% CI: 0.111–0.153), constituting 95.71% of the aggregate association (total effect: β = 0.137, 95% CI: 0.117–0.159; Fig. 3).

Fig. 3.

Mediation Analysis of the Relationship Between IA and Depressive Symptoms with PA as the Mediator

Anxiety-Depressive Comorbidity (ADC)

For ADC symptoms, mediation by PA explained 3.50% of the total effect (indirect effect: β = 0.004, 95% CI: 0.003–0.006). The direct effect retained substantial magnitude (β = 0.113, 95% CI: 0.093–0.136), representing 96.50% of the total pathway (total effect: β = 0.117, 95% CI: 0.097–0.139; Fig. 4).

Fig. 4.

Mediation Analysis of the Relationship Between IA and ADC Symptoms with PA as the Mediator

These findings collectively demonstrate that PA serves as a partial mediator in the IA–mental health relationship, though its proportional contribution to total effects is minor (3.50–4.29%) relative to the overwhelming predominance of direct pathways. The results underscore the necessity of investigating additional mediating mechanisms beyond PA to fully elucidate IA’s psychiatric sequelae.

Discussion

Our study demonstrates significant associations between internet addiction (IA), insufficient physical activity (PA), and mental health comorbidities in adolescents. The dose-response relationship between IA severity and mental health risks (ORs = 1.77–2.09) persists after adjusting for sociodemographic confounders and obesity status, suggesting IA’s independent contribution to psychological distress. Notably, the ADC demonstrated the strongest association with IA (OR = 2.09), indicating adolescents with IA had over double the risk of comorbid anxiety and depression symptoms. This substantiates IA as a critical target in adolescent mental health interventions. The inverse PA-mental health relationship (IPA ORs = 1.22–1.39) reveals PA’s potential protective role, consistent with previous neurobiological evidence that exercise modulates stress response systems [32, 33].

The results of this study show that adolescents aged 12–15 years have a significantly higher vulnerability to IA (OR = 3.15) than those aged over 16 years, this developmental pattern mirrors findings about heightened neuroplasticity during puberty onset [34, 35], meaning that the brain is more responsive to experience, may lead to positive cognitive development but may also increase the risk of psychopathology. While European studies show that IA peaks at the age of 16–18 years [36]. This difference may be related to the characteristics of China’s education system: high-intensity exam-oriented education starts in junior high school (grade 7), which leads to younger adolescents facing academic pressure earlier and turning to online for emotional compensation [37, 38].

The study identified notable rural-urban disparities in mental health outcomes. Contrary to global WHO trends [39], urban adolescents exhibited higher depressive symptoms prevalence (11.4%) versus rural peers (10.6%). This reflects China’s unique digital landscape: while rural device ownership rates surpass urban areas by 38%, this connectivity often occurs without adequate safeguards. Rural youth, frequently with less parental supervision, may use devices to compensate for social gaps (aligning with studies on China’s left-behind children [40, 41] and rural India [42]). Unlike the US digital divide (where rural coverage lags [37]), China’s policy-driven rural internet penetration creates distinct risks for unregulated overuse [43]. Urban adolescents, conversely, face elevated competitive pressures contributing to depression vulnerability.

A emerges as a culturally salient protective factor, with daily moderate-vigorous exercise attenuating IA’s mental health impacts by 21–29%. While consistent with PA’s established neuroprotective effects [44, 45], the mediation magnitude (3.5%−4.3% of total effects) underscores the need for multi-modal interventions. In this study, the mediating effect of PA on the negative effects of IA was only 3.5%−4.3%, which was much lower than the 11–15% in the United States [46]. On one hand, the granularity of PA measurement may be insufficient. The lack of distinction between school physical education classes and voluntary exercise makes it challenging to accurately reflect the true state of PA and its impact on mental health. In China, mandatory group exercises (e.g., recess exercises) are predominant, while in the United States, adolescents are more likely to participate in voluntary team sports (e.g., school teams), which also provide opportunities for social capital accumulation [22, 38]. On the other hand, some studies indicate that boosting the frequency and intensity of PA is more effective in enhancing mental health. For example, one study revealed that multimodal interventions incorporating various forms of PA can significantly improve college students’ mental health, outperforming single strategies [23]. This suggests that the effects of PA may vary with different frequencies and intensities. However, due to the questionnaire survey design, quantitative PA data could not be obtained, which is also an area for improvement in future research.

Emerging evidence suggests a neurobiological triad linking these phenomena: excessive internet use dysregulates dopaminergic reward pathways [24, 47], while PA enhances prefrontal inhibition of amygdala hyperactivity via BDNF-mediated neuroplasticity [48, 49]. The competing displacement vs. compensation hypotheses frame their interaction: screen time may displace PA opportunities, yet PA could buffer internet-induced stress through endocannabinoid signaling [50]. This dynamic is amplified in China’s high-pressure educational ecosystems, where academic stress increases vulnerability to both IA and PA deprivation [51]. While IA poses significant risks, certain factors can act as buffers or protective mechanisms. Community bonds may moderate the relationship between cyberbullying, internet pornography, internet fraud, and problematic internet use [18]. Promoting healthy coping styles, encouraging physical activity, and providing mental health support are crucial interventions [11, 52] Furthermore, educating adolescents about responsible internet use and fostering a balanced lifestyle can mitigate the adverse effects of IA [53, 54].

Implications for policy and practice

Our findings provide key insights for policy and practice. The strong association between Internet addiction (IA) and mental health problems underscores the need for routine IA screening in school health programs. This facilitates early identification of at-risk adolescents and timely intervention. Physical activity (PA) programs that promote acculturation are essential. Given the important role that physical activity plays in reducing the adverse effects of physical activity, schools and communities should create accessible, attractive physical activities, such as tai chi or badminton, that are appropriate for Chinese adolescents. Strengthening digital literacy education can help adolescents develop healthy online habits. This can be achieved through school curricula that teach students to think critically about online content, time management, and the psychological effects of excessive Internet use. Targeted intervention measures should be taken for young adolescents (12–15 years old), non-boarding students and junior high school students. These include providing after-school programs for supervised Internet access and structured physical activity to reduce screen time and improve mental health.

Strengths and limitations

This study has several notable strengths. First, the stratified random cluster sampling across 16 urban-rural districts ensured population representativeness, with oversampling of school types (junior/senior/vocational high schools) enhancing generalizability to diverse educational contexts. Second, we employed validated instruments, including DSM-5-derived IA criteria and clinically calibrated mental health scales (CES-D, GAD-7), which improved diagnostic accuracy and cross-study comparability. Analytical robustness was achieved through hierarchical adjustment for sociodemographic, familial, and anthropometric confounders across three multivariable models, effectively isolating the independent effects of IA and PA. Third, subgroup analysis confirmed the consistency of results across different population groups. Finally, mediation analysis elucidates the partial mediating role of physical activity in the Internet addiction-mental health association.

However, several limitations warrant consideration. First, the temporal ambiguity inherent to cross-sectional data precludes definitive causal inference between IA, PA, and mental health outcomes, though observed dose-response gradients support biological plausibility, longitudinal data are needed to disentangle temporal sequences. Second, potential non-response bias merits attention: Exclusions (637/21,329; 3.0%) due to incomplete data (n = 92), missing outcomes (n = 87), or absent covariates (n = 458) could introduce selection bias if excluded adolescents differed systematically (e.g., higher IA severity). To mitigate these concerns, the use of clinical interviews or digital tools to collect data may improve response rates and data quality among hard-to-reach populations. Third, the binary categorization of internet use (addictive/non-addictive) overlooks nuances in online activities (e.g., educational vs. recreational use), which may differentially impact mental health. Additionally, the binary classification of PA may oversimplify the complexity of physical activity behaviors. This categorization does not account for variations in PA intensity, duration, or type, which could lead to measurement error. Future studies would benefit from more nuanced PA assessment methods, such as accelerometer data, to capture a more comprehensive picture of PA behaviors and their mental health impacts. Fourth, unmeasured confounders like sleep quality, academic stress, and genetic predispositions may partially explain the observed associations. Finally, while the sampling strategy ensured Tianjin’s representativeness, regional socioeconomic and cultural specificities (e.g., higher urbanization rate than national average) may limit direct extrapolation to other Chinese regions. Nevertheless, these limitations reflect common challenges in large-scale epidemiological research rather than unique design flaws, and our rigorous methodology mitigates their impact on core conclusions. Despite these limitations, this study provides timely, policy-actionable insights into the IA-PA-mental health triad, laying groundwork for culturally tailored interventions in China’s evolving digital landscape.

Conclusion

This study explores the intricate links between internet addiction, physical activity, and mental health in Chinese adolescents. Our results show that internet addiction is a key risk factor for mental health issues like anxiety and depression. The data reveals a dose-response link: higher addiction levels correlate with greater mental health risks. Physical activity offers some protection against these issues, though its mediating role is modest.

These results point to several important implications. Targeted interventions are needed, such as internet addiction screening in schools, promoting culturally relevant physical activity, and improving digital literacy education. Longitudinal research is also crucial to better understand the causal links between these factors and explore related neurodevelopmental mechanisms.

Despite providing valuable insights, our study has limitations like its cross-sectional nature and use of self-reported data. Future research should address these by using more objective measurement tools and longitudinal designs.

Abbreviations

IA

Internet addiction

PA

Physical activity

ADC

Anxiety-Depression Comorbidity

BMI

Body mass index

IPA

Insufficient physical activity

MVPA

Moderate-to-vigorous physical activity

Authors’ contributions

XYW and FQL investigated and collected data. XWZ and NJ statistically processed the data and wrote the manuscript. KX and ZYS consulted relevant literature. ZHL revised the article. All authors contributed to the article and approved the submitted version.

Funding

This work was sponsored by Tianjin Health Research Project (Grant No. TJWJ2024QN089).

Data availability

The datasets analyzed in the current study are available from the corresponding author, Zhonghui Liu (liuzhonghui2012@126.com), upon reasonable request.

Declarations

Ethics approval and consent to participate

This study was approved by the Tianjin Center for Disease Control and Prevention Ethics Committee (approval number: TJCDC-R-2023-001 granted on 02/21/2023). The study was conducted in accordance with the Declaration of Helsinki. Written informed consent were obtained from all students and their parents or guardians before participating in the study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.