Meta-analytic structural equation modeling with mediation role of problematic mobile phone use between solitude and anxiety and depression

Abstract

Background

Problematic mobile phone use (PMPU) is a growing concern linked to adverse mental health outcomes, such as anxiety and depression. While solitude is a known risk factor, the specific roles of its distinct dimensions, which include social avoidance, loneliness, unsociable, and positive solitude, are not well understood within this framework. This study aimed to meta-analytically examine a model clarifying the mediating role of PMPU in the relationships between these facets of solitude and common mental health challenges.

Methods

This meta-analysis synthesized findings from 37 independent studies (N = 27,713), identified through a systematic search of nine databases (e.g., Web of Science, PsycINFO, CNKI). A meta-analytic structural equation modeling (MASEM) approach was used to test the hypothesized relationships among the four types of solitude, PMPU, anxiety, and depression.

Results

The MASEM results revealed that three maladaptive forms of solitude, namely social avoidance, loneliness, and unsociable, were significant positive predictors of both PMPU and adverse mental health outcomes (i.e., anxiety and depression). In contrast, positive solitude was not significantly associated with either PMPU or these outcomes. Importantly, PMPU was identified as a significant mediator that explains the pathway from these maladaptive forms of solitude to anxiety.

Conclusion

This meta-analysis provides evidence that maladaptive forms of solitude are associated with both PMPU and psychological distress. The findings substantiate a theoretical model in which PMPU acts as a key mechanism linking these types of solitude to anxiety. Regarding practical implications, these results suggest that interventions designed to alleviate anxiety in individuals experiencing loneliness or social avoidance should also address and aim to reduce problematic mobile phone use.

Trial registration

Exploring the Link Between solitude, Problematic Mobile Phone Use, and Anxiety and Depression: A Meta-Analytic Structural Equation Modeling Investigation ID: CRD420251021680.

Introduction

Mobile phones have become essential tools in daily life, offering valuable avenues for communication, social connectivity, and entertainment. However, the high-frequency use of mobile phones has been increasingly linked to Problematic Mobile Phone Use (PMPU). PMPU is broadly characterized as an individual’s inability to regulate their mobile phone use, leading to excessive and uncontrolled usage that results in significant dysfunction across various aspects of life [1], such as academic performance, occupational, and social relationships.

It is important to recognize that the definitions and terms related to PMPU are still evolving in the academic field. PMPU serves as an umbrella term that includes related concepts like nomophobia (the fear of being without a mobile phone), mobile phone addiction, and mobile phone dependence [1–3]. Although debates continue about the exact theoretical frameworks and diagnostic criteria for these overlapping constructs, PMPU is utilized as an inclusive term in this meta-analysis. It refers to a cluster of maladaptive behaviors associated with mobile phone engagement, characterized by their problematic and dysfunctional impact. Accordingly, our study integrates findings from primary research that operationalizes these behaviors, including the widespread phenomenon of excessive mobile phone use, to enhance the ecological validity of the findings.

Relationships between different types of solitude and PMPU

Solitude, broadly defined, refers to being alone, a state that involves both physical and emotional separation from others [4]. The impact of solitude on individual development is complex. While some research points to negative outcomes like social withdrawal and distressing emotions, other studies highlight its benefits, including enhanced creativity and positive self-discovery [5]. The distinction of types of solitude driven by different motivations may contribute to understanding these controversies.

Recent research has emphasized the complexity of solitude motivations and their different impacts on psychological well-being [6, 7]. For instance, researchers have pointed to a range of reasons why people seek out solitude. These include the desire for freedom from social pressure, simply being alone due to loneliness, or pursuing intimacy and spirituality [8]. Building on this work, scholars have also created classifications that distinguish between positive and avoidant solitude [9]. These insights suggest that kinds of solitude driven by different motivations are associated with diverse or even contrary outcomes. Consequently, while solitude may offer increased opportunities for mobile phone use, potentially developing into higher frequency usage patterns, it could be inferred that its impact on PMPU largely depends on an individual’s motivations for solitude.

To better clarify the link between solitude and PMPU, we adopt the solitude classification proposed by Chen et al. [10], which builds on Nicol’s [11] typology of self-determined and non-self-determined solitude, itself rooted in Self-Determination Theory. This system categorizes solitude into four types, each driven by distinct underlying motivations and needs: non-self-determined solitude, which includes social avoidance and loneliness; and self-determined solitude, which comprises unsociable and positive solitude.

Non-self-determined solitude and PMPU

Social avoidance and loneliness are both classified as forms of non-self-determined solitude [10]. Social avoidance happens when individuals deliberately withdraw from social situations to lessen social anxiety or discomfort. This behavior often stems from feeling pressured externally or wanting to avoid negative social interactions. In contrast, loneliness, within Chen et al.’s motivation-based framework, is a type of solitude not freely chosen but resulting from an individual’s unmet need for social connection. People in this state find themselves alone (or perceive their aloneness) because their social relationships feel unsatisfactory. While traditionally viewed as a distressing emotional state [12, 13], our study conceptualizes loneliness simply as a motivation-driven type of solitude.

According to Self-Determination Theory [14], when individuals lack autonomy, feel externally pressured, or are driven by factors like guilt avoidance, they are generally more prone to experiencing negative emotions such as anxiety, sadness, or boredom. Building on this principle, individuals experiencing non-self-determined solitude, which inherently involves a perceived lack of autonomy, tend to find themselves in such negative emotional states. A primary coping mechanism to escape internal discomfort is often using a mobile phone, due to its easy availability [15, 16]. When mobile phone use exceeds the underlying needs it is intended to fulfill, such compensatory or escapist use may readily escalate into compulsive patterns, and even PMPU [17, 18].

Self-determined solitude and PMPU

Regarding self-determined solitude, Chen et al. [10] categorized it into two types: unsociable and positive solitude. Unsociable stems from an individual’s intrinsic emotional need for tranquility and quiet. Individuals characterized by unsociable typically distance themselves from others voluntarily, not out of anxiety or social deficits, but from a genuine preference for solitude as a peaceful, burden-free experience. They often express a desire to reduce social stimuli in order to maintain emotional balance. In contrast, positive solitude is driven by higher-level needs, such as self-actualization, where individuals actively seek alone time for personal growth, creativity, or focused work, believing that solitude possesses intrinsic value and utility.

According to Self-Determination Theory [14], solitude that is triggered by personal interests or values, rather than by external pressures or guilt, is often associated with more positive experiences. Additionally, studies show that people who truly enjoy being alone often prefer to spend their solitary time on their chosen activities without interruption and are less likely to rely on their phones for engagement [19–21]. For example, Diefenbach et al. found that self-reflection and self-insight when alone were negatively correlated with mobile phone use [15]. Ryan further emphasize that meeting intrinsic motivations is crucial for promoting mental health and reducing behavioral addictions within the Self-Determination Theory framework [22].

These findings suggest that individuals experiencing unsociable and positive solitude tend to engage in activities aligned with their interests and core values. This deliberate and fulfilling engagement is less likely to be disrupted by PMPU.

Relationships between PMPU and mental health

Many studies have consistently shown that PMPU is a significant risk factor for mental health issues, particularly anxiety and depression [23–25]. The mechanisms behind this link are multifaceted [26]. First, Interference Theory suggests that excessive mobile phone use distracts from or replaces crucial face-to-face social interactions [27]. This reduction in real-world social quality often increases feelings of isolation, which are known precursors to anxiety and depression. Second, frequent mobile phone use, particularly on social media, can prompt upward social comparisons, contributing to feelings of inadequacy and low self-esteem [28]. Third, using mobile phones late at night frequently disrupts sleep, a known pathway leading to mood dysregulation [29]. Additionally, the constant alerts and Fear of Missing Out (FoMO) linked with PMPU may cause ongoing stress and anxiety [30].

Building on these ideas, Elhai et al. [17], drawing from Compensatory Internet Use Theory, proposed a model. They suggested that individuals experiencing solitude, perhaps driven by social avoidance, might increase their mobile phone use. This escalation is associated with PMPU, which in turn is related to more severe anxiety symptoms.

The present study

Previous research on the relationship among solitude, PMPU, and mental health has yielded inconsistent findings, particularly when solitude is treated as a single concept [31, 32]. While some studies have explored different subtypes of solitude, their findings have also been contradictory [19, 33]. To our knowledge, no empirical study has yet adopted the four motivation-based types of solitude to comprehensively examine its relationships with PMPU and mental health.

Therefore, to address these inconsistencies and fill this research gap, the present study aims to provide a more integrated conclusion on these relationships by using Meta-Analytic Structural Equation Modeling (MASEM) [34]. Specifically, our primary research question is whether PMPU mediates the relationship between solitude and mental health.

Based on the literature, we developed and tested a model outlining the relationships among different types of solitude, PMPU, and mental health (see Fig. 1).

Fig. 1.

The mediation model diagram of solitude, PMPU, and mental health

First, we examined the link between solitude and PMPU. It is theorized that individuals of non-self-determined solitude, which is often marked by negative emotions, may tend to engage in excessive mobile phone use as a coping mechanism [35]. In contrast, self-determined solitude, which is rooted in personal choice and satisfaction, is less likely to foster such problematic use. Based on this reasoning, we hypothesized that non-self-determined solitude would positively predict PMPU (H1), whereas self-determined solitude would negatively predict PMPU (H2).

Next, we considered the link from PMPU to mental health. Extensive research indicates that PMPU is a risk factor for problems such as anxiety and depression [23–25, 36], potentially through mechanisms like impaired social interaction and disrupted sleep [27, 37]. Accordingly, we hypothesized that PMPU would positively predict anxiety and depression (H3).

Finally, drawing from Self-Determination Theory and the Compensatory Internet Use Theory [17], we proposed a mediation model. We suggest that individuals in non-self-determined solitude may rely on their phones to cope with discomfort, which is associated with PMPU and, in turn, is related to outcomes like anxiety and depression. Conversely, those in self-determined solitude are less likely to need such external coping strategies. Therefore, we hypothesized that PMPU would positively mediate the relationship between non-self-determined solitude and anxiety and depression (H4), while it would negatively mediate the relationship between self-determined solitude and anxiety and depression (H5).

While many of the primary studies in this meta-analysis are cross-sectional, MASEM allows us to statistically test these hypothesized directional links across a large, aggregated dataset. This method can provide stronger and more generalizable evidence for the proposed theoretical model than single studies alone [34].

Methods

Research design and data collection

In this study, meta-analytic structural equation modeling (MASEM) is employed to verify the hypothesized mediating model. MASEM offers a comprehensive way to summarize information from multiple studies, effectively compensating for the limitations of traditional meta-analysis (which can typically analyze only a single effect) and standard structural equation models (which lack a robust mechanism for repeated validation across studies) [34, 38].

Moreover, the two-stage structural equation modeling (TSSEM) approach provides particular flexibility in MASEM analysis. It relaxes the stringent requirement for all primary studies to report correlation coefficients among all variables, thereby allowing for the inclusion of studies with missing variables. In the first stage, TSSEM directly assesses heterogeneity and effect sizes. The second stage then uses a combined correlation matrix for structural equation modeling (SEM) analysis, enhancing the accuracy of model estimation [39–41].

MASEM analyzed a path model encompassing the four types of solitude, PMPU, and anxiety and depression, using meta-analysis data. We primarily focused on indirect effects, reflecting PMPU’s mediating role between each of the four types of solitude and mental health. Direct effects among mental health, four types of solitude, and PMPU were also examined.

This meta-analysis was conducted and reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 statement [42]. A systematic literature search was conducted in December 2023, covering a publication period from 2009 to 2023, given the widespread global adoption of smartphones during this period.

We searched the following electronic databases: Web of Science, PsycINFO, EBSCO, Elsevier ScienceDirect, SAGE Journals, and Google Scholar for English-language articles, and the China National Knowledge Infrastructure (CNKI), Wanfang Data, and CQVIP Journal Database for Chinese-language articles. These databases were chosen for their comprehensive coverage of research in psychology, medicine, and the social sciences.

The search strategy combined terms related to the core concepts of the study (solitude, PMPU, anxiety, and depression), and the full search strings are detailed in Table 1. All retrieved records were imported into EndNote X9 for management and duplicate removal.

Table 1.

Retrieval information

Item	Content
Academic database	“Web of Science”, “PsycINFO”, “EBSCO”, “Elsevier”, “SAGE”, “Google Scholar”, “Chinese National Knowledge Infrastructure”, “WANFANG DATA”, and “CQVIP Journal Database”
Keywords	(a) PMPU: “problematic * phone use” OR “*phone addiction” OR “*phone dependence” OR “*phone over use” OR “nomophobia” (b) Solitude: “loneliness” OR “alone”, “avoidance” or “social avoidance” OR “social withdrawal” OR “shyness”, “unsociable” OR “social isolation” OR “isolation behavior” OR “eccentric” OR “oddity”, “positive solitude”, “solitude” (c) Mental health: “anxiety”, “depression”

The inclusion criteria for the literature were as follows: (1)Studies examining the relationships among solitude (referring to any of the four subtypes), PMPU, and mental health (anxiety or depression); (2) Studies reporting correlations between at least two variables of interest; (3) Studies providing essential data for analysis, specifically effect size (e.g., a correlation coefficient, r) and sample size (N); (4) Empirical research articles; (5) Articles published in either English or Chinese; and (6) To test the model’s generalizability, no age limits were imposed. Consequently, participants included primary/secondary students, university students, and community adults.

The exclusion criteria were as follows: (1) Studies that were not empirical or irrelevant to the topic; and (2) Studies involving special populations (e.g., left-behind children, medical professionals).

Coding procedure

The 37 selected articles underwent a coding procedure to identify detailed information regarding nationality, gender, and measurement variables. To ensure statistical independence, each independent sample was coded only once, yielding a single effect size. If an article contained multiple independent samples, they were coded separately as distinct studies.

Literature quality evaluation

This study adopted the Joanna Briggs Institute (JBI) Critical Appraisal Checklist for Analytical Cross-sectional Studies to evaluate the methodological quality of the original research [43]. The checklist comprises eight items, each with four possible response options: ‘Yes’, ‘No’, ‘Unclear’, or ‘Not applicable’. Affirmative responses (‘Yes’) were assigned 1 point, while all other responses received 0 points, resulting in a total score ranging from 0 to 8 for each study. Study quality was classified based on the percentage of the total score achieved: scores < 50% were identified as low quality, 50–80% as medium quality, and > 80% as high quality.

Data Analysis

In this study, MASEM modeling was performed using the metaSEM(1.4.0) package in R(4.3.3).

Fixed-effects model and random-effects model

When combining effect sizes, there are two methods to choose from: if the participant groups or true effect sizes are consistent across all studies, a fixed-effects model can be used for analysis. The random-effects model assumes that, due to differences in samples, study designs, and measurements, each study may have a specific effect size. Since it is often difficult in practice to differentiate whether the origin of sample heterogeneity is clinical or methodological, in most cases it may be a combination of both, making it challenging to draw these distinctions in interpretations. Therefore, using both fixed-effects and random-effects models is pragmatic. In this study, the fixed effects model was used initially to combine correlation matrices for MASEM fitting. Then, the random effects model was used to combine correlation matrices for MASEM fitting, allowing for comparison between the two methods of effect size combination.

Two stage structural equation model

The Two-Stage Structural Equation Model (TSSEM) was proposed by Cheung [40]. In TSSEM, the correlation coefficients from Stage 1 are pooled across multiple sets of structural equation models, and in Stage 2, these combined correlation matrices are fitted to the structural models using Weighted Least Squares (WLS) estimation. Stage 1 produces an asymptotic variance-covariance matrix, and the WLS procedure uses the inverse of this matrix, ensuring that correlation coefficients estimated with higher precision (due to a larger number of studies) in Stage 1 are given greater weight in the parameter estimation of Stage 2.

Stage 1: Combinatorial correlation matrix

According to Jak, in the study, R_i is a p_i×p_i sample correlation matrix, where p_i is the number of variables observed in the i-th study, and not all studies need to include all variables [41]. For example, to explore the relationships among variables A, B, and C, Study 1 includes all variables, Study 2 is missing variable C, and Study 3 is missing variable A. Therefore, the correlation matrices for Studies 1, 2, and 3 are as follows:

The selection matrices corresponding to studies 1, 2, and 3 are shown below (if a study is missing a particular variable, the row corresponding to that variable will be excluded):

Next, multi-group structural equation modeling (SEM) is used to estimate the combined correlation matrix for all variables, treating each study as a group. The model for each group is expressed as follows:

1

In this model, is the merged correlation matrix with a diagonal of 1, is the selection matrix for each study, and is a diagonal matrix accounting for differences in variables across studies. Maximum likelihood estimation (MLE) is used to estimate the model parameters by maximizing the sum of the likelihood functions across all studies:

2

In the above equation, represents the sample size of the i-th study, where . is the maximum likelihood estimate in each study.

Stage 2: Fitting the structural equation model

In the process of fitting a structural equation model, Stage 1 involves the fitting of a model that provides combined correlation coefficients, as well as the asymptotic variances and covariance estimated matrix V of these coefficients. Therefore, in Stage 2, the WLS estimation is used to fit R into the structural equation model under examination, with the following expression:

3

where represents the column vector in the combined matrix , denotes the latent column vector in the structural model under investigation, and serves as the inverse matrix of the asymptotic variance-covariance matrix used as the weighting matrix.

Results

The initial search yielded 3,253 records. After removing duplicates and screening titles and abstracts, 103 full-text articles were assessed for eligibility (see Fig. 2 for the PRISMA flow diagram). Based on the inclusion criteria, a final sample of 37 studies was retained for the meta-analysis, comprising 27 articles in English and 10 in Chinese. These studies involved a total of 27,713 participants.

Fig. 2.

PRISMA flow diagram

Two coders independently coded all included studies (Cohen’s Kappa = 0.75 for categorical variables; ICC = 0.93 for continuous variables), resolving any discrepancies through discussion. Regarding methodological quality, all studies were rated as medium quality or above; specifically, 8 were classified as medium quality and 29 as high quality. The inter-rater agreement for the quality assessment was substantial (Cohen’s Kappa = 0.704).

The characteristics of the included studies are summarized in Table 2, which presents details on sample characteristics, measurement tools (further specifics available in Supplementary Appendix A), effect sizes, and quality assessment results. The search process is detailed in Supplementary Appendix B.

Table 2.

Literature coding table for MASEM

Author	Year	Nationality	Female Proportion	Measured Variables
Bitar et al. [44]	2023	Lebanon	0.92	unsociable、PMPU、depression、anxiety
Borg et al. [45]	2022	Canada	0.50	positive solitude、unsociable、avoidance、depression、anxiety
Chen et al. [46]	2021	China	0.60	Positive solitude、PMPU
Diefenbach et al. [15]	2019	United Kingdom	0.80	Positive solitude、PMPU
Enez Darcin et al. [47]	2016	Turkey	0.62	avoidance、loneliness、anxiety、PMPU
Estévez et al. [48]	2017	Spain	0.83	unsociable、avoidance、PMPU、depression、anxiety
Han et al. [49]	2017	China	0.59	avoidance、PMPU、anxiety
Houghton et al. [50]	2016	Australia	0.52	positive solitude、Unsociable
Jiang et al. [51]	2023	China	0.39	avoidance、PMPU、depression
Keisari et al. [52]	2022	Israel	0.76	positive solitude、depression
Kim et al. [53]	2018	South Korea	0.58	avoidance、PMPU、anxiety
Lane et al. [54]	2021	China	0.86	avoidance、PMPU、depression、anxiety
Lapierre et al. [55]	2019	United States	0.66	loneliness、PMPU、depression
Lay et al. [56]	2019	Canada	0.73	positive solitude、anxiety
Lee et al. [57]	2018	South Korea	0.00	unsociable、PMPU、depression、anxiety
Li et al. [58]	2022	China	0.38	social avoidance、PMPU、anxiety
Li et al. [59]	2023	China	0.50	positive solitude、loneliness、PMPU
Lian et al. [60]	2021	China	0.48	positive solitude、PMPU
Lin et al. [61]	2020	China	0.52	positive solitude、depression、anxiety
Liu et al. [62]	2015	China	N/A	positive solitude、loneliness、depression
Liu et al. [63]	2019	China	0.52	loneliness、PMPU、anxiety
Lu et al. [64]	2022	China	0.52	positive solitude、unsociable、social avoidance、loneliness、PMPU
Peng et al. [65]	2022	China	0.60	loneliness、avoidance、PMPU、depression
Tian et al. [66]	2023	China	0.57	loneliness、PMPU、depression
Wang et al. [67]	2019	China	0.43	loneliness、PMPU、depression、anxiety
Wang et al. [68]	2018	China	0.22	loneliness、PMPU、depression、anxiety
Xiong et al. [69]	2023	China	0.50	avoidance、PMPU、anxiety
Ying Zeng [70]	2015	China	0.60	avoidance、PMPU、anxiety
Xiaomeng Gao et al. [71]	2019	China	0.60	avoidance、unsociable、PMPU、anxiety
Shanshan Hu [72]	2015	China	0.51	avoidance、PMPU、anxiety
Haibin Huang et al. [73]	2018	China	0.67	avoidance、PMPU、anxiety
Xue Liu [74]	2014	China	0.57	avoidance、PMPU、anxiety
Junjie Sun et al. [75]	2017	China	0.43	avoidance、loneliness、PMPU、anxiety
Yimeng Wang [76]	2018	China	0.61	avoidance、PMPU、anxiety
Qin Wu et al. [77]	2019	China	0.60	loneliness、PMPU、depression
Bin Zhang et al. [78]	2019	China	0.49	loneliness、PMPU、depression
Huan Zhu et al. [79]	2019	China	0.61	unsociable、PMPU、anxiety

For the first stage of the analysis, a fixed-effects model was employed to combine the correlation matrices. The results of this stage are presented in Table 3.

Table 3.

Combined correlation matrix in the fixed effects model

variable	positive solitude	unsociable	social avoidance	loneliness	PMPU	depression	anxiety
positive solitude	1
unsociable	0.27***	1
social avoidance	0.27***	0.37***	1
loneliness	0.10***	0.37***	0.37***	1
PMPU	0.09***	0.23***	0.36***	0.31***	1
depression	0.09***	0.35***	0.39***	0.60***	0.30***	1
anxiety	0.17***	0.35***	0.51***	0.42***	0.42***	0.51***	1

Significance levels:* p < .05, ** p < .01, *** p < .001

The fit indices for the fixed-effects model were: = 4537.07, p < 0.001, RMSEA = 0.24, SRMR = 0.14, CFI = 0.74 and TLI = 0.68. Furthermore, the Q test was statistically significant, and the I² values exceeded 75%, confirming substantial heterogeneity across studies. Given these results, the random-effects model was subsequently used to pool the correlation matrices (see Table 4).

Table 4.

Summary of meta-analytic results for relationships among variables

Variables	k	N	heterogeneity assessments			combined correlation	Egger’s Test
			Q	I²	p	r	z	p
positive solitude and unsociable	2	2215	114.55	97.83%	< 0.001	0.27**	0.58	0.562
positive solitude and social avoidance	2	1750	5.55	81.99%	0.019	0.31***	-1.85	0.064
positive solitude and loneliness	2	2144	153.25	97.34%	< 0.001	0.19*	0.50	0.616
positive solitude and PMPU	5	4278	71.88	94.70%	< 0.001	0.07	-0.55	0.583
positive solitude and depression	5	3418	43.13	95.77%	< 0.001	0.15	-0.92	0.360
positive solitude and anxiety	4	2711	71.88	94.70%	< 0.001	0.08	-1.30	0.192
unsociable and social avoidance	3	2193	19.09	83.54%	< 0.001	0.40***	-0.50	0.617
unsociable and loneliness	1	678	N/A	N/A	N/A	0.45***	N/A	N/A
unsociable and PMPU	3	1799	17.48	81.10%	< 0.001	0.29***	1.10	0.272
unsociable and depression	2	1607	244.53	99.07%	< 0.001	0.39***	0.69	0.489
unsociable and anxiety	2	1121	93.55	96.78%	< 0.001	0.39***	1.22	0.221
social avoidance and loneliness	2	2488	40.19	95.25%	< 0.001	0.35***	1.13	0.258
social avoidance and PMPU	18	14,197	763.23	97.06%	< 0.001	0.30***	-1.74	0.082
social avoidance and depression	4	3359	332.67	99.28%	< 0.001	0.26	-1.68	0.094
social avoidance and anxiety	11	9294	1794.14	99.49%	< 0.001	0.42***	-0.44	0.663
loneliness and PMPU	12	10,362	241.83	96.05%	< 0.001	0.32***	-0.26	0.798
loneliness and depression	6	4774	52.38	90.64%	< 0.001	0.58***	-0.50	0.615
loneliness and anxiety	3	2854	13.73	78.23%	< 0.001	0.37***	1.49	0.136
PMPU and depression	13	9510	46.63	81.67%	< 0.001	0.30***	-1.13	0.257
PMPU and anxiety	20	14,577	2407.42	97.93%	< 0.001	0.37***	-1.41	0.160
depression and anxiety	5	3097	201.98	97.48%	< 0.001	0.54***	1.89	0.059

k number of effect sizes, N total sample size, r pooled correlation coefficient (random-effects model), N/A Not applicable due to insufficient number of studies (k = 1), * p < .05, ** p < .01, *** p < .001

The results of the random-effects model showed significant positive correlations between pairs of positive solitude, unsociable, social avoidance, and loneliness. Unsociable was significantly positively correlated with PMPU, depression, and anxiety. Social avoidance showed a significant positive correlation with PMPU and anxiety. Loneliness exhibited significant positive correlations with PMPU, depression, and anxiety. Moreover, there were significant positive correlations between pairs of PMPU, anxiety, and depression.

Transferring the results of the random-effects model from Stage 1 to the analysis in Stage 2, the MASEM model, as depicted in Fig. 1, utilized solitary behavior as the independent variable, PMPU as the mediating variable, and anxiety and depression as the dependent variables. The model fit statistics were: = 10.04, p < 0.01, RMSEA = 0.02, SRMR = 0.07, CFI = 0.99, TLI = 0.87.

As shown in Fig. 3, unsociable, social avoidance, and loneliness all significantly and positively predicted PMPU and anxiety. Unsociable and loneliness additionally predicted depression. Notably, social avoidance did not significantly predict depression, nor did PMPU directly predict depression. Additionally, positive solitude showed no significant direct prediction for PMPU, anxiety, or depression.

Fig. 3.

Meta-analytic mediation model of positive solitude, PMPU and anxiety and depression

Regarding indirect effects (Table 5), PMPU significantly and positively mediated the relationships between unsociable, social avoidance, and loneliness with anxiety. However, no mediating role of PMPU was observed between positive solitude and anxiety, nor between any solitude type and depression.

Table 5.

Indirect effects in the meta-analytic mediation model

Indirect effect	estimated value	95% CI
βpositive solitude−PMPU × βPMPU−depression	-0.002	-0.02 ~ 0.02
βunsociable−PMPU × βPMPU−depression	0.004	-0.02 ~ 0.02
βsocial avoidance−PMPU × βPMPU−depression	0.01	-0.02 ~ 0.33
βloneliness−PMPU × βPMPU−depression	0.01	-0.02 ~ 0.03
βpositive solitude−PMPU × βPMPU−anxiety	-0.01	-0.04 ~ 0.02
βunsociable−PMPU × βPMPU− anxiety	0.02	0.001 ~ 0.05
βsocial avoidance−PMPU × βPMPU− anxiety	0.03	0.006 ~ 0.07
βloneliness−PMPU × βPMPU− anxiety	0.03	0.01 ~ 0.07

Publication bias assessment

Publication bias was assessed through visual inspection of the funnel plot and Egger’s regression test. As shown in Fig. 4; Table 4, no significant publication bias was detected (p > 0.05).

Fig. 4.

Funnel plot of the study on the relationship between four types of solitude, PMPU and mental health

Given that several associations were based on a limited number of studies, Rosenthal’s fail-safe N was additionally calculated for all pairwise associations. Most associations exhibited large fail-safe N values, indicating robustness against potential publication bias. However, a few associations based on a small number of studies (e.g., positive solitude and anxiety) had relatively small fail-safe N values, suggesting these effects may be more sensitive to unpublished null results.

Discussion

This study examined the relationship between various factors, including solitude, PMPU, mental health, using the MASEM method. Our findings indicate a positive association between certain types of solitude (namely, social avoidance, loneliness, and unsociable) and both PMPU and anxiety. Additionally, loneliness and unsociable were positively associated with depression. In contrast, we found no significant relationship between positive solitude and PMPU, anxiety, or depression. Our results also confirmed a positive relationship between PMPU and anxiety. Regarding mediation, the results showed that PMPU mediated the relationship between three types of solitude(social avoidance, loneliness, and unsociable) and anxiety. However, this mediating role was not found for positive solitude. Finally, we did not find evidence that PMPU mediated the relationship between any type of solitude and depression.

Solitude and PMPU

Our results show that non-self-determined solitude was a significant positive predictor of PMPU, which supports H1 and aligns with previous research [80, 81]. A possible explanation for this is that when individuals experience non-self-determined solitude due to social avoidance or loneliness, they may turn to their mobile phones to satisfy unmet emotional or belongingness needs. This compensatory use is often linked to the excessive or uncontrolled patterns that characterize PMPU.

However, our findings regarding self-determined solitude’s relationship with PMPU were mixed, contrary to H2’s predicted negative link. Notably, unsociable, a subtype of self-determined solitude, unexpectedly predicted PMPU positively and significantly. The unexpected positive link between unsociable and PMPU may stem from its nature as a stable personality trait often associated with clinical psychiatric conditions [82]. Individuals high in this trait frequently struggle with inattentiveness and poor self-regulation in managing mobile phone use. Consistent with the Deficient Self-Regulation Model (DSRM) [83], such self-regulatory challenges are associated with uncontrolled increases in mobile phone use, which in turn are linked to PMPU.

Conversely, positive solitude showed no significant relationship with PMPU. This may be because individuals who engage in positive solitude are often focused on self-improvement or intrinsically interesting activities. For them, mobile phones are less frequently used, or may simply serve as a functional tool to support these activities, rather than being a primary source of engagement.

Mediating Role of PMPU

Regarding H3, our results provide partial support, showing that PMPU positively predicts anxiety. However, no significant predictive effect was found for depression.

Regarding H4, which predicted that non-self-determined solitude would increase PMPU and thereby worsen anxiety and depression, our results provide partial support. Non-self-determined solitude positively predicted PMPU, and this indirect pathway significantly increased anxiety symptoms [84, 85]. However, no statistical significance emerged for the indirect effect on depression.

Elhai et al.’s theoretical model [17] helps contextualize these findings. Their model suggests that individuals experiencing anxiety may increase mobile phone use while in solitude due to factors like social avoidance, FoMO, boredom, and rumination. This increased use is associated with PMPU, which in turn is related to more severe anxiety symptoms. Our findings imply that solitary individuals with varied motivations experience distinct PMPU-related pathways affecting their mental health. For instance, socially avoidant individuals may use phones to alleviate social anxiety and fulfill belonging needs through online interactions [86, 87], potentially fostering problematic use. Similarly, lonely individuals, dissatisfied with social connections, may use mobile phones to cope with boredom and loneliness [88], often checking notifications due to FoMO [30]. This pattern is associated with mobile phone use and PMPU, contributing particularly to anxiety.

The differential effects on anxiety versus depression warrant further discussion. Anxiety, which is characterized by heightened vigilance, worry, and physiological arousal [89], may be more immediately triggered by mobile phone stimuli (e.g., constant notifications, social comparison, FoMO). Depression, marked by chronic low mood, anhedonia, and social withdrawal [90], might be less directly influenced by these mobile phone-related factors. Alternatively, its development may involve more complex, longer-term processes not fully captured by our model, or its effects could be mediated by other unexamined factors.

Regarding H5, which proposed that self-determined solitude would reduce PMPU and thus alleviate anxiety and depression, the results did not support this. Instead, findings revealed a more complex and partly opposing pattern. Specifically, unsociable positively predicted PMPU, and this indirect effect increased anxiety symptoms; no significant effect emerged for depression. Positive solitude showed no significant indirect effects on either PMPU or mental health outcomes.

The positive association between unsociable and PMPU is consistent with prior work linking this trait to nomophobia [91] and its classification as a stable personality feature tied to various psychiatric disorders [82]. Individuals high in unsociable may struggle with mobile phone self-regulation. Following the Deficient Self-Regulation Model (DSRM) [83], such deficits are associated with excessive use, which in turn can be accompanied by anxiety.

The lack of a significant mediating role for PMPU in the relationship between positive solitude and outcomes of anxiety and depression supports H5. A likely explanation is that individuals who engage in positive solitude are, by definition, using this time to meet higher-level psychological needs [92, 93]. Consequently, they are less likely to experience anxiety and depression in the first place, and are also less prone to developing PMPU.

Implications

In sum, these findings highlight the nuanced roles that different forms of solitude play in relation to PMPU and mental health, underscoring the importance of motivation and self-regulation in these complex interactions. Whereas previous meta-analyses have typically examined only one or two types of solitude [94, 95], our study utilized a classification of four motivation-based types. This approach helps to clarify the previously inconsistent findings on the relationships among solitude, PMPU, and mental health. Our results suggest that future empirical studies should consider these different types of solitude to better understand their unique effects. Furthermore, while traditional meta-analyses typically examine the relationship between two variables, our use of MASEM allowed us to test a path model involving multiple variables. This provides a more comprehensive understanding of how these factors are interrelated.

While associations between solitude, PMPU, and depression were identified, PMPU did not serve as a clear mediating mechanism for depression. This highlights the need to explore alternative pathways explaining links between solitude and depression, and PMPU and depression. Finally, although positive solitude showed no statistically significant predictive effects on PMPU, anxiety, or depression, a negative trend was observed. Future studies should further investigate positive solitude’s potential role in mitigating PMPU related mental health problems.

This study provides insights with practical relevance. Specifically, our findings suggest a practical implication for clinical and educational settings. Interventions for individuals who experience anxiety related to social avoidance, loneliness, or unsociability could adopt a dual focus by not only targeting PMPU but also promoting more adaptive solitude experiences to reduce anxiety.

Limitations

This study has several limitations. First, we did not examine specific mobile phone activities engaged in by the four types of solitary individuals. Exploring these could offer deeper insights into PMPU motivations across solitude groups and clarify underlying mechanisms. Second, although PMPU includes various subcategories, we treated it as a single, unified construct. This approach, while necessary for the meta-analysis, may obscure which specific aspects of PMPU (e.g., compulsive checking vs. escapist use) are most strongly related to different types of solitude. Third, potential moderating effects of demographic variables (e.g., age, region) on relationships among solitude types, PMPU, and mental health remain unexplored and warrant further investigation, as factors like age and cultural context could theoretically alter the meaning and impact of solitude. Additionally, while our MASEM approach allowed for theory testing, reliance on primary studies’ correlational data limits definitive causal inference and robust empirical testing of alternative models. Regarding publication bias, funnel plots and Egger’s regression tests did not indicate significant asymmetry. However, fail-safe N analyses showed that while most associations were robust, a few based on a small number of studies (e.g., positive solitude and anxiety) may be more sensitive to unpublished null results. These findings should therefore be interpreted with caution, and future studies with larger samples are needed to further validate them. Finally, some paths in our model had a relatively small number of included effect sizes (K ≤ 3), particularly those involving positive solitude, unsociability, and the relationship between loneliness and anxiety. This limited the statistical power for these specific analyses, which may account for some of the non-significant results observed. Therefore, these specific findings should be interpreted with caution, and more primary studies are needed to validate these pathways.

Conclusion

Recent research regarding the relationships among solitude, PMPU, anxiety, and depression has yielded inconsistent results. By synthesizing existing empirical studies, this study confirms that specific subtypes of solitude (unsociability, social avoidance, and loneliness) are significantly associated with both PMPU and anxiety. These findings not only provide robust evidence clarifying the link between PMPU and anxiety and validate the theoretical model proposed by Elhai et al., but also emphasize a critical implication for educational and public health practices, suggesting that interventions should adopt a dual focus targeting both maladaptive solitude and PMPU to effectively alleviate anxiety associated with mobile phone use.

Authors’ contributions

T.L., X.L. and G.N. conceived this work. T.L., X.L. and N.Z. obtained resources and undertook data curation. G.N. and X.L. supervised the project. T.L., N.Z., Y.Y., X.L. and G.N. were were involved in reviewing and editing the manuscript.

Funding

This work was supported by the Key Laboratory of Adolescent Cyberpsychology and Behavior, Ministry of Education (CCNUCYPSYLAB2324A01).

Data availability

The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

Tianjin Normal University: 2024030410.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.