Mobile phone dependency and subclinical depressive-anxiety symptom co-occurrence in college students: a cross-lagged panel network analysis

Background

With the widespread use of smartphones among college students, issues related to smartphone dependence have become increasingly prevalent. Existing research has shown that subclinical anxiety and depression often co-occur and are closely associated with smartphone dependence. This study employed a cross-lagged panel network approach to explore the dynamic interplay between smartphone dependence and the co-occurrence of subclinical anxiety and depressive symptoms among college students. Methods: The Mobile Phone Addiction Index (MPAI), Patient Health Questionnaire-9 (PHQ-9), and Generalized Anxiety Disorder-7 (GAD-7) were used as measurement tools. Two waves of longitudinal data were collected from 571 college students (26.3% male; mean age = 19.53 years, SD = 1.12) between March 2024 and September 2024. Results: The results revealed that anxiety symptoms significantly predicted smartphone dependence. The node Uncontrollable worry (A2) exerted the strongest influence on other symptoms, whereas Delayed work (MPAI-16) was more frequently predicted by other symptoms. The three strongest cross-lagged paths in the network were from Spend too much time (MPAI-2) to Complained by others (MPAI-1), from Uncontrollable worry (A2) to Delayed work (MPAI-16), and from Uncontrollable worry (A2) to Anxiety if not used for some time (MPAI-10), all of which played a key role in maintaining the overall network structure. Conclusion: Based on these findings, interventions targeting smartphone dependence among college students should begin with an assessment of their interpersonal relationships and anxiety levels. Enhancing social support, fostering healthy interpersonal connections, and alleviating students’ worries and fears may serve as effective strategies to reduce smartphone dependence in this population.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-025-24142-9.

Introduction

Mobile phone dependency among college students

In March 2024, the China Internet Network Information Center (CNNIC) released its 53rd Statistical Report on China’s Internet Development. As of December 2023, the number of mobile internet users in China had reached 1.091 billion, with individuals aged 10 to 29 accounting for 28.4% of this population [1]. College students—typically in early adulthood—not only have nearly universal access to smartphones, but also exhibit higher usage frequency and stronger psychological attachment to these devices compared to other age groups [2]. According to the dual systems model [3], adolescents and young adults experience a developmental imbalance between the socio-emotional system and the cognitive control system, leading to heightened reward sensitivity and a diminished capacity for impulse control [4]. This neurodevelopmental pattern makes college students particularly susceptible to excessive smartphone use. Consequently, they tend to underestimate their screen time and struggle with regulating their usage behaviors—phenomena broadly categorized as mobile phone dependency or problematic smartphone use (PSU) [5]. A meta-analysis reported that the global prevalence of PSU among college students ranges from 20 to 30% [6], with a national average of 23% reported among Chinese students [7]. However, prevalence rates vary widely across studies due to differences in assessment tools, cut-off criteria, and sample characteristics [8, 9].

Relationship between smartphone dependence and anxiety/depression

An expanding body of empirical research has consistently demonstrated strong associations between mobile phone dependence and symptoms of anxiety and depression [10–12]. This relationship can be understood through the lens of the developmental compensation hypothesis [13], which proposes that individuals who encounter difficulties in meeting key developmental challenges (e.g., establishing interpersonal competence) may resort to maladaptive behaviors—such as excessive smartphone use—as a compensatory mechanism when internal coping resources are insufficient [14]. Among the various mental health correlates of PSU, depressive and anxiety symptoms have garnered the most research attention [15–18]. The Interaction of Person-Affect-Cognition-Execution (I-PACE) model [19] offers a comprehensive framework, suggesting that addictive behaviors emerge through dynamic interactions among individual predispositions, affective and cognitive responses, and executive functioning. Emotional dysregulation and cognitive biases can intensify these interactions. In stressful contexts, smartphones serve as easily accessible tools for immediate emotional relief, offering temporary distraction through social networking, gaming, and other digital content [20]. Both cross-sectional and longitudinal studies have confirmed the positive association between smartphone dependence and depressive symptoms [21]– [22]. Similarly, PSU has been closely linked to anxiety—particularly in the form of fear of missing out (FoMO) [23]. However, the majority of existing studies have relied on latent variable approaches, such as structural equation modeling, which may obscure the fine-grained, symptom-level interrelationships that underlie these comorbid conditions.

Network analysis as a new approach to understanding comorbid symptom structures

To better capture the complexity of symptom co-occurrence, researchers have increasingly adopted network analysis—a methodological framework that conceptualizes mental disorders as systems of interacting symptoms rather than manifestations of underlying latent constructs [24]. Within this framework, psychopathology arises when symptom clusters become causally interconnected and mutually reinforcing [25]. Network analysis provides powerful tools for identifying central or bridge symptoms that play pivotal roles in the development and maintenance of mental disorders. Recent studies have employed symptom-level network analysis to examine the co-occurrence patterns between smartphone dependence and internalizing symptoms such as depression and anxiety [26–28]. However, most of these investigations have used cross-sectional data, which limits the ability to infer temporal and directional relationships among symptoms.

By integrating network analysis with cross-lagged panel models (CLPMs), researchers can construct temporal symptom networks that align with the network theory’s emphasis on dynamic interactions over time [29]. Several longitudinal studies have explored PSU among Chinese college students [20, 30–32], but a notable limitation is that anxiety and depression symptoms are often analyzed separately, despite robust evidence supporting their comorbidity [33]– [34]. To address this gap, the present study applies a cross-lagged panel network model to examine the directional, symptom-level associations between smartphone dependence and comorbid anxiety and depression symptoms. This approach enhances the precision of causal inference and supports the identification of clinically relevant targets—specifically, symptoms with high out-expected influence (those driving other symptoms) and high in-expected influence (those predominantly shaped by other symptoms) [35].

Method

Participants

Since China’s higher education system includes both general undergraduate education and higher vocational education, we adopted a cluster sampling method by selecting one general undergraduate university and one higher vocational college in western China. Within each selected institution, first- and second-year students were randomly recruited. Two waves of data collection were conducted in March 2024 (T1) and September 2024 (T2), respectively.We followed the recommendation of Fried and Cramer [36] to estimate the sample size for the current study. Given that the network comprises 30 nodes and 325 edges, and that each estimated parameter requires at least three participants, a stable network estimation would require a minimum of 1,095 participants. We recruited a total of 1,100 college students from two selected institutions as participants. Informed consent was obtained from all participants prior to distributing the online questionnaires. A total of 1,100 questionnaires were distributed, and after excluding invalid responses, 915 valid questionnaires were collected, resulting in an effective response rate of 83.2%. After excluding invalid responses due to leave of absence, absence from class during the T2 assessment, large amounts of missing data, or patterned/invalid responding, 571 valid samples were retained for longitudinal analysis, yielding an effective follow-up rate of 66.4%. Among the retained participants, 150 were male (26.3%) and 421 were female (73.7%), with a mean age of 19.53 years (SD = 1.12). Independent-samples t-tests and one-way ANOVAs were conducted on T1 variables to compare participants who were retained versus those who were lost to follow-up, in order to assess whether attrition occurred at random. The results indicated no significant differences between retained and attrited participants in terms of age (M _retained =19.53, M _attrited =19.52, t = −0.36, p = 0.97), smartphone dependence (M _retained = 35.08, M _attrited = 34.09, t = −1.17, p = 0.24), depression (M _retained = 12.67, M _attrited = 12.78, t = 0.35, p = 0.73), or anxiety(M _retained = 9.38, M _attrited = 9.53, t = 0.59, p = 0.55). However, significant differences were found between the two groups in terms of gender (F = 20.21, p <0.001) and parents’ average monthly income (F = 19.03, p <0.001), though the effect sizes were small (Partial η²_gender= 0.02, Partial η² _{Parental average monthly income}= 0.02). These findings suggest that there was no systematic attrition in the current study.

Measures

Mobile phone addiction index (MPAI)

The Mobile Phone Addiction Index (MPAI), originally developed by Leung [37] and subsequently revised by Huang et al. [38], was used to assess participants’ level of mobile phone addiction. The scale comprises 17 items rated on a 5-point Likert scale, and includes four subscales: loss of control, withdrawal, escape, and inefficiency. The Chinese version of the total scale demonstrated good internal consistency, with a Cronbach’s alpha coefficient of 0.87. In the present study, the Cronbach’s alpha coefficients at Time 1 (T1) and Time 2 (T2) were 0.94 and 0.91, respectively.

PHQ-9 (patient health questionnaire-9)

The Chinese version of the Patient Health Questionnaire-9 (PHQ-9) revised by Zhang et al. [39] was used to assess participants’ depressive symptoms over the past two weeks. The PHQ-9 consists of nine items, each corresponding to one of the diagnostic criteria for Major Depressive Disorder (MDD) as outlined in the DSM-IV. Items are rated on a 4-point scale ranging from 0 to 3, indicating “not at all,” “several days,” “more than half the days,” and “nearly every day,” respectively. Higher scores indicate more severe depressive symptoms. The Chinese version of the total scale demonstrated good internal consistency, with a Cronbach’s alpha coefficient of 0.85. In the present study, the Cronbach’s alpha coefficients of the PHQ-9 at Time 1 (T1) and Time 2 (T2) were 0.93 and 0.91, respectively.

GAD-7 (generalized anxiety disorder 7-item questionnaire)

The Chinese version of the Generalized Anxiety Disorder-7 (GAD-7) scale, revised by He Xiaoyan [40], was used to assess participants’ anxiety symptoms over the past two weeks. The scale consists of seven items, each rated on a 4-point scale ranging from 0 to 3, indicating “not at all,” “several days,” “more than half the days,” and “nearly every day,” respectively. Higher total scores reflect more severe symptoms of generalized anxiety disorder. The Chinese version of the scale demonstrated good internal consistency, with a Cronbach’s alpha coefficient of 0.90. In the present study, the Cronbach’s alpha coefficients of the GAD-7 at Time 1 (T1) and Time 2 (T2) were 0.93 and 0.94, respectively.

Data collection

We used Wenjuanxing (www.wjx.cn), a widely used online survey platform in China, to present all questionnaire items and collect participants’ responses. Before completing the survey, participants were instructed to carefully read the survey instructions and use their student ID numbers as a unique identifier for tracking across time points. Participant recruitment was conducted offline. Specifically, research team members visited key campus locations such as academic buildings, dormitory areas, and cafeterias to introduce the study in person, explain its purpose and procedures, and distribute informed consent forms. Some authors also set up promotional booths on campus to invite eligible students to participate. All participants volunteered to join the study and completed written or electronic informed consent prior to participation. Throughout the recruitment and data collection process, each author was actively involved in on-site coordination, verbal instruction, data management, and maintaining procedural integrity, ensuring the rigor and validity of the research implementation.

Data analysis

We used SPSS 24.0 to perform descriptive statistics, correlation analyses, and group difference tests. For reliability analysis of the scales, we selected all items of each instrument at both T1 and T2 time points and conducted separate reliability analyses, thereby obtaining the internal consistency coefficients (Cronbach’s α) for each measure at both time points. Cross-lagged panel network estimation was performed in R (version 4.3.3), with regularized regression conducted using the glmnet package and network visualization generated using the qgraph package. The procedure for cross-lagged panel network analysis followed the methodological guidelines proposed by Rhemtulla [29].

We first computed the unstandardized and regularized autoregressive coefficients both within and across time points. Regularization was performed using the least absolute shrinkage and selection operator (LASSO) to reduce the number of false positives. To more clearly highlight the cross-lagged paths of primary interest [41], autoregressive paths were set to zero in the main figure presented in the text. A full cross-lagged panel network including autoregressive paths is provided in the Appendix.

The directed edges in the cross-lagged panel network allow for the computation of two centrality indices: out-expected influence (out-EI) and in-expected influence (in-EI). Out-EI reflects the extent to which a node predicts other nodes within the network, while in-EI represents the degree to which a node is predicted by other nodes.

The accuracy of edge weights was assessed by estimating 95% confidence intervals (CIs) using nonparametric bootstrapping (bootstrapped samples = 1000). Less overlap between CIs indicates higher accuracy. In addition, a subsetting bootstrap procedure was conducted to evaluate the stability of centrality estimates. If the rank order of centrality estimates in networks constructed from subsets of the data remains highly correlated with that of the original network—even after removing a large proportion of the sample—then the centrality estimates can be considered stable. Finally, the centrality stability coefficient (CS-coefficient) was calculated as a reference index; values above 0.5 indicate good stability, and values above 0.25 indicate acceptable stability.

Results

Descriptive statistics of smartphone dependence, anxiety, and depression among college students

The means and standard deviations for all items related to smartphone dependence, depression, and anxiety at both time points were calculated and are presented in Table 1. At T1, the average scores for smartphone dependence, depression, and anxiety were 35.08 (SD = 11.18), 12.67 (SD = 4.14), and 9.38 (SD = 3.34), respectively. At T2, the corresponding average scores were 34.15 (SD = 11.84), 12.96 (SD = 4.40), and 9.73 (SD = 3.43). A comparison of cross-sectional networks at T1 and T2 revealed no significant differences in overall network structure (p = 0.37) or global strength (p = 0.49) between the two time points. Network analysis results for both time points are provided in the Supplementary Materials, Figure S1.

Table 1.

Abbreviation, means, standard deviation (SD) of each variable in the present network at two-time points

Items	Abbreviation	T1 M(SD)	T2 M(SD)
Symptoms of PHQ-9
Anhedonia	D-1	0.52(0.66)	0.51(0.64)
Sad mood	D-2	0.45(0.61)	0.49(0.64)
Sleep	D-3	0.47(0.64)	0.54(0.68)
Fatigue	D-4	0.53(0.65)	0.54(0.66)
Appetite	D-5	0.37(0.57)	0.46(0.63)
Guilt	D-6	0.44(0.62)	0.41(0.57)
Concentration	D-7	0.43(0.62)	0.47(0.64)
Motor	D-8	0.28(0.52)	0.34(0.56)
Suicide	D-9	0.18(0.46)	0.21(0.45)
Symptoms of GAD-7
Nervousness	A-1	0.42(0.60)	0.47(0.61)
Uncontrollable worry	A-2	0.35(0.58)	0.38(0.59)
Excessive worry	A-3	0.41(0.62)	0.43(0.62)
Trouble relaxing	A-4	0.37(0.58)	0.43(0.60)
Restlessness	A-5	0.24(0.48)	0.33(0.52)
Irritability	A-6	0.35(0.56)	0.39(0.57)
Feeling afraid	A-7	0.24(0.48)	0.30(0.57)
Symptoms of MPAI
Complained by others	MPAI-1	2.04(0.97)	1.98(0.88)
Spend too much time	MPAI-2	2.04(0.88)	2.01(0.92)
Hiding the time spent	MPAI-3	1.59(0.81)	1.71(0.86)
Phone bill overrun	MPAI-4	1.53(0.82)	1.55(0.84)
Use longer than planned	MPAI-5	2.30(1.04)	2.18(1.02)
Failure to cut down the time	MPAI-6	2.13(0.99)	2.04(0.96)
Never enough time on the phone	MPAI-7	1.95(0.95)	1.84(0.88)
Preoccupied by no network signal	MPAI-8	2.06(1.07)	2.03(1.02)
Hard to switch off	MPAI-9	2.26(1.26)	2.17(1.20)
Anxiety if not used for some time	MPAI-10	1.79(0.94)	1.87(0.99)
Feel lost without the phone	MPAI-11	1.88(0.99)	1.88(0.97)
Hard to contact	MPAI-12	2.50(1.23)	2.36(1.19)
Avoid isolation	MPAI-13	2.09(1.09)	2.01(1.03)
Alleviate loneliness	MPAI-14	2.28(1.11)	2.12(1.04)
Uplift low mood	MPAI-15	2.53(1.10)	2.36(1.09)
Delayed work	MPAI-16	2.04(0.95)	2.02(0.96)
Reduced productivity	MPAI-17	2.07(0.98)	2.02(0.95)

SD standard deviation, PHQ-9 the nine-item Patient Health Questionnaire, GAD-7 the seven-item Generalized Anxiety Disorder scale, MPAI The Mobile Phone Addiction Index

Network estimation

The cross-lagged panel network estimating the relationships from T1 to T2 between smartphone dependence and symptoms of depression and anxiety is presented in Fig. 1. The results showed that the node with the highest out-expected influence (out-EI) was Uncontrollable worry (A2) (OEI = 34.28), followed by Spend too much time (MPAI-2) (OEI = 34.19) and Concentration (D7) (OEI = 33.15). The nodes with the highest in-expected influence (in-EI) were Delayed work (MPAI-16) (IEI = 32.74), Failure to cut down the time (MPAI-6) (IEI = 32.72), and Anxiety if not used for some time (MPAI-10) (IEI = 32.71). In the current network, the three most influential cross-lagged edges were: Spend too much time (MPAI-2) → Complained by others (MPAI-1) (OR = 1.31), Uncontrollable worry (A2) → Anxiety if not used for some time (MPAI-10) (OR = 1.27), and Uncontrollable worry (A2) → Delayed work (MPAI-16) (OR = 1.24). The cross-lagged panel network including autoregressive paths for all variables is presented in Appendix Figure S2.

Fig. 1.

Cross-lagged panel network of smartphone dependence and depression-anxiety among college students

The symptom centrality (z-scores) in the cross-lagged panel network of smartphone dependence and depression-anxiety is shown in Fig. 2. The node with the highest out-expected influence (out-EI) was Uncontrollable worry (A2), while the node with the highest in-expected influence (in-EI) was Delayed work (MPAI-16). Results from the edge weight bootstrapping procedure (see Appendix Figure S3) indicated moderate accuracy of the cross-lagged network estimation: there was considerable overlap in the 95% confidence intervals (CIs) for most edge weights, although some of the strongest edges did not have overlapping CIs. The subsetting bootstrap results (see Appendix Figure S4) suggested that the estimates of in-EI, out-EI, and bridge-EI in the cross-lagged network were stable and generalizable. Additionally, the centrality stability coefficients for out-EI, in-EI, and bridge-EI were 0.36, 0.52, and 0.67, respectively. Centrality difference tests revealed that, within the T1→T2 network, most symptom nodes differed significantly from one another in terms of centrality (see Appendix Figure S5).

Fig. 2.

Centrality estimates of the t1→t2 cross-lagged panel network

Further analysis of bridge expected influence (bridge-EI; z-scores) revealed that the node Uncontrollable worry (A2) exhibited significantly higher bridge expected influence compared to all other nodes (see Fig. 3).

Fig. 3.

Bridge centrality estimates of the T1→T2 cross-lagged panel network

Taken together, based on the results of bridge expected influence (bridge-EI) and the strength of cross-lagged edges, the present study identified two primary bridging edges in the longitudinal network linking smartphone dependence and depression-anxiety symptoms among college students: Uncontrollable worry (A2) → Delayed work (MPAI-16) and Uncontrollable worry (A2) → Anxiety if not used for some time (MPAI-10). These two edges may play a critical role in sustaining the overall network.

Discussion

This study is the first to apply a cross-lagged panel network analysis to examine the symptom-level predictive pathways between smartphone dependence and depression-anxiety symptoms among college students over a six-month period. The results revealed that, within the cross-lagged network, the node Uncontrollable worry (A2) exerted the most predictive influence on other symptoms, while Delayed work (MPAI-16) was most frequently predicted by other nodes. The edges from Uncontrollable worry (A2) to Delayed work (MPAI-16) and from Uncontrollable worry (A2) to Anxiety if not used for some time (MPAI-10) emerged as the strongest bridging connections in the network. These findings suggest that these two bridge edges may serve as a key foundation for the comorbidity between smartphone dependence and depression-anxiety symptoms in college students.

Consistent with previous cross-sectional and cross-lagged panel network studies, Uncontrollable worry (A2) emerged as the most influential symptom in the longitudinal network [42]. This finding aligns with the assumptions of the Interaction of Person-Affect-Cognition-Execution (I-PACE) model [19], which suggests that negative emotions such as anxiety drive individuals to seek immediate and effective.

emotion regulation strategies. Due to their portability and multifunctionality—combining entertainment, social networking, and communication—smartphones have become the preferred tool for college students to obtain instant gratification, helping them alleviate negative emotions in the short term. However, this also leads to increased reliance on smartphones as a coping mechanism for stress and emotional difficulties [43]. In addition, the symptom nodes Spend too much time (MPAI-2) and Concentration (D7) also showed high predictive influence within the network. Prior research on smartphone addiction has demonstrated that it shares similar features with substance addiction, such as excessive use, withdrawal symptoms, and impaired daily functioning [44]. As smartphone use becomes a primary emotion regulation strategy, individuals inevitably spend more time on their devices. Furthermore, difficulty in concentration (D7) also showed strong predictive links to other symptoms in the network. This is supported by existing evidence indicating that self-perceived attentional deficits are associated with smartphone addiction [45], and that adolescents with smartphone addiction often exhibit impairments in executive functioning [46].

In the current cross-lagged panel network, the nodes most strongly influenced by other symptoms were Delayed work (MPAI-16), Failure to cut down the time (MPAI-6), and Anxiety if not used for some time (MPAI-10). All of these nodes belong to the smartphone dependence symptom cluster, suggesting that smartphone dependence is more likely to be predicted by emotional problems. Previous research has found a moderate positive correlation between smartphone addiction and procrastination behaviors among students [47]. Difficulty in controlling smartphone use often leads to reduced time for sleep, academic tasks, and social activities, thereby contributing to problems in these areas [48]. Additionally, a cross-sectional network analysis conducted among rural Chinese adolescents identified Failure to cut down the time (MPAI-6) and Anxiety if not used for some time (MPAI-10) as the most influential symptom nodes in the network, which is consistent with the findings of the present study [26]. According to the Interaction of Person-Affect-Cognition-Execution (I-PACE) model [49], the interplay between personal factors (e.g., negative emotions) and situational factors contributes to increased dependence on smartphones among college students. Moreover, weaker self-control [50] and higher levels of fear of missing out (FoMO) [51] may lead individuals to struggle with reducing their smartphone use and to experience anxiety when separated from their devices.

This study identified three strongest cross-lagged paths: Spend too much time (MPAI-2) → Complained by others (MPAI-1), Uncontrollable worry (A2) → Anxiety if not used for some time (MPAI-10), and Uncontrollable worry (A2) → Delayed work (MPAI-16). The path from Spend too much time (MPAI-2) to Complained by others (MPAI-1) lies within the domain of smartphone dependence symptoms, suggesting that excessive time spent on smartphones may lead to interpersonal problems. This finding is consistent with previous research showing that smartphone addiction among adolescents is significantly negatively associated with social support [52], while social anxiety is positively related to problematic social media use [53]. These results support the developmental compensation hypothesis [13], which posits that when individuals lack sufficient psychological resources to cope with interpersonal difficulties, they may resort to maladaptive compensation—such as problematic smartphone use. Excessive smartphone use can damage interpersonal relationships, and poor social relationships in turn reinforce dependence on smartphones, creating a vicious cycle. Therefore, interventions targeting smartphone dependence among college students should pay particular attention to the quality of their interpersonal relationships. The other two paths both originate from the anxiety symptom Uncontrollable worry (A2) and point to the smartphone dependence symptoms Anxiety if not used for some time (MPAI-10) and Delayed work (MPAI-16), consistent with prior findings [20]. This suggests that, compared to depression, anxiety symptoms are stronger predictors of smartphone dependence among college students. Previous cross-lagged panel network analyses on adolescent anxiety and depression have also shown that anxiety prospectively predicts depression [34]. Therefore, interventions specifically targeting anxiety symptoms—particularly the node Uncontrollable worry (A2) identified in this study—may be especially effective in reducing smartphone dependence among college students.

The present study also has several limitations. First, the study did not include a clinical sample, which may limit the generalizability of the findings. Second, all data were collected through self-report measures, making it difficult to avoid self-report bias or socially desirable responding. Finally, although no systematic attrition was observed during the study, the final sample size was still relatively small. Future research should aim to address these limitations through improved sampling and methodology.

Conclusion

This study is the first to employ a cross-lagged panel network approach to examine the symptom-level interactions within the comorbidity network of smartphone dependence, depression, and anxiety among college students. The results revealed that, within the directed network, the node Uncontrollable worry (A2) exerted the strongest influence on other symptoms, while Delayed work (MPAI-16) was most frequently predicted by other nodes. The three strongest cross-lagged paths were: Spend too much time (MPAI-2) → Complained by others (MPAI-1), Uncontrollable worry (A2) → Delayed work (MPAI-16), and Uncontrollable worry (A2) → Anxiety if not used for some time (MPAI-10). These connections played a crucial role in maintaining the integrity of the overall network. Therefore, interventions targeting smartphone dependence among college students should begin with a comprehensive understanding of their interpersonal relationships and anxiety levels. Enhancing social support, fostering healthy interpersonal connections, and alleviating students’ worries and fears may serve as effective strategies to reduce smartphone dependence in this population.

In addition to its theoretical significance, this study offers practical implications across clinical, educational, technological, and policy domains. By identifying anxiety symptoms—especially Uncontrollable worry—as central predictors in the network, the findings support prioritizing anxiety-focused psychological interventions (e.g., cognitive-behavioral therapy, mindfulness-based techniques) to indirectly reduce problematic smartphone use. Educators and university counselors may apply these insights to develop early screening tools and prevention programs, while policymakers and campus administrators can use them to design digital wellness strategies that promote structured screen time and peer support. Furthermore, commercial developers of digital well-being tools may incorporate the identified symptom-level predictors into algorithmic screening and personalized feedback systems for early intervention. From a research perspective, this study advances the integration of cross-lagged modeling and network analysis to capture dynamic symptom interactions, laying a foundation for future investigations into digital behaviors and emotional comorbidities in broader populations. Overall, this work provides a novel framework for identifying key symptom targets and informing multidimensional approaches to mental health support among college students.

Authors’ contributions

Author Z.Z., L.L., X.D., C.C., J.W., N.H., J.X. and J.L. were responsible for drafting of the manuscript and data analysis.Author L.L., Z.Z., C.C., J.W., N.H., J.X., J.L.and X.D. include contributed to data collection and analysis.Author Z.Z., L.L., X.D. handled experimental operations and review manuscript.

Funding

This research was supported by the Scientific Research Planning Project of the Sichuan Psychological Society: “Longitudinal Network Relationships Between Smartphone Addiction, Depression, and Anxiety Among College Students in Vocational Colleges” (Project No. SCSXLXH202403105) and Northwest Normal University Graduate Research Grant Program (2023KYZZ-B036) and Hainan College of Foreign Studies Scientific Research Project (Hwyky2019-07).

Data availability

This study was approved by the Ethics Review Committee of Northwest Normal University (Approval Number 2024321). All of the procedures were performed in accordance with the Declaration of Helsinki and relevant policies in China. All participants agreed to participate voluntarily, with informed consent when they fled in the survey. This study did not involve any clinical trials or experiments; therefore, clinical trial registration and consent for publication are not applicable. The authors declare no competing interests. The datasets used in the current study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

This study was approved by the Ethics Review Committee of Northwest Normal University (Approval Number 2024321). All of the procedures were performed in accordance with the Declaration of Helsinki and relevant policies in China. All participants agreed to participate voluntarily, with informed consent when they fled in the survey.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.