Depression and social media addiction among teenagers in a longitudinal study with dual moderation

Abstract

Depression has been consistently associated with social media addiction, yet the psychological processes underlying this relationship remain insufficiently understood. The present study aimed to examine the longitudinal association between depression and social media addiction among teenagers, with particular attention to the potential moderating roles of difficulty describing feelings and experiential avoidance. A three-wave longitudinal study was conducted with 3,184 teenagers from China. Participants completed self-report measures assessing depression, social media addiction, difficulty describing feelings, and experiential avoidance. Descriptive statistics, correlation analyses, and moderated regression models were used to examine the proposed longitudinal associations. Depression at Time 1 (T1) was positively associated with social media addiction at Time 3 (T3). In addition, difficulty describing feelings and experiential avoidance at Time 2 (T2) significantly moderated the association between T1 depression and T3 social media addiction. Teenagers who experienced greater difficulty describing their feelings or who tended to avoid unpleasant internal experiences appeared more vulnerable to the longitudinal association between depression and social media addiction. These findings highlight the importance of emotional understanding and regulation capacities in shaping patterns of social media use among adolescents and offer implications for theory development as well as targeted prevention and intervention efforts.

Introduction

In today’s digital age, social media has become woven into almost every aspect of daily life. By 2024, the number of active users worldwide had already exceeded 5 billion and is expected to climb beyond 6 billion by 2028¹. Within this massive global community, teenagers stand out as one of the most vulnerable groups². Although relatively young, they tend to spend the most time online, which makes them especially sensitive to the possible risks that accompany social media use³. When engagement with these platforms becomes excessive or poorly regulated, it can lead to a range of negative outcomes. Among these, social media addiction has drawn particular concern in recent years⁴. Andreassen and Pallesen defined social media addiction as a form of psychological dependence marked by an uncontrollable inner drive that pushes individuals to devote excessive time and energy to online interaction⁵. A large meta-analysis covering 64 countries estimated that about 17.42% of the general population shows signs of social media addiction⁶. In China, recent findings suggest that around 15.2% of teenagers experience symptoms consistent with this condition⁷. A growing number of studies—including both cross-sectional research and meta-analyses—consistently show that social media addiction is linked to declines in academic achievement, disrupted sleep, and poorer mental health outcomes^8,9. Given these harmful effects, social media addiction has come to be recognized as a serious issue for public health¹⁰. Understanding the mechanisms and risk factors that contribute to its development is therefore an urgent research priority. Such work can offer an evidence-based foundation for prevention and intervention strategies that reduce teenagers’ vulnerability to this problem.

The determinants of social media addiction are varied and complex, but the relationship between depression and social media addiction among teenagers has received particular attention. Recent data spanning from August 2021 to August 2023 indicates that the prevalence of depression in the past two weeks was 13.1% among adolescents and adults aged 12 and older¹¹. Depression represents not only a leading source of the world’s disease burden but also one of the most common psychological disorders during adolescence¹². Its main clinical symptoms—persistent sadness, anxiety, irritability, and low self-esteem—can substantially interfere with daily functioning^13–18. Evidence increasingly indicates that negative emotional states are closely related to social media addiction¹⁹. According to the Compensatory Internet Use Theory²⁰, individuals who experience persistent negative feelings may rely on online social platforms as a coping mechanism, seeking emotional comfort or temporary escape from real-life pressures^21–26. Within the school environment, teenagers from different cultural and regional backgrounds often live and study together while dealing with a variety of challenges such as academic competition, relationship stress, financial worries, and emotional difficulties²⁷. In such situations, social media can appear to provide an immediate sense of relief or belonging. Yet, when emotional reliance on these platforms becomes excessive, it may gradually develop into addictive use²⁸. In addition, prior research has suggested that certain traits commonly seen in depressed individuals—such as low self-esteem—can serve as strong predictors of addictive behaviors involving the internet or social media²⁹. Drawing on these theories and empirical findings, the present study proposes that depression is prospectively associated with social media addiction among teenagers (H1).

When examining how depression relates to social media addiction, one potential factor that deserves attention is the difficulty describing feelings. This variable represents one of the core features of alexithymia³⁰, which involves challenges in identifying and expressing emotions, confusion between emotional states and bodily sensations, and a tendency to focus on concrete, externally oriented thinking³¹. Evidence from previous studies shows that people who have greater difficulty describing feelings often show avoidant attachment patterns, which makes emotional regulation and interpersonal communication more difficult³². From the viewpoint of the cognitive–behavioral model, such individuals are more inclined to avoid dealing with distressing internal experiences³³. The immersive and relatively anonymous environment of social media offers these individuals an easy route to escape from everyday pressures and regulate unpleasant emotions³⁴. Frequent online engagement brings momentary relief, and this temporary comfort can, through negative reinforcement, gradually strengthen the habit, eventually turning it into an addictive pattern³⁵. Furthermore, individuals with high levels of difficulty describing feelings typically have limited interpersonal sensitivity and struggle with emotional closeness, which may create challenges in direct, face-to-face communication³⁶. The features of social media—its distance, anonymity³⁷, and the lack of face-to-face demands³⁸—reduce emotional pressure and allow for more controlled expression. These characteristics compensate for their social limitations and provide a sense of belonging that is hard to achieve offline³⁹. Consequently, such individuals are more prone to rely heavily on social media, thereby increasing their risk of addiction^40–44. Based on this reasoning, the present study proposes that difficulty describing feelings moderates the relationship between depression and social media addiction among teenagers (H2).

In addition to emotional expression difficulties, experiential avoidance may also play a moderating role. Rooted in acceptance and commitment therapy (ACT), experiential avoidance refers to a psychological tendency to manage distress by avoiding or suppressing negative inner experiences—such as unwanted thoughts, painful emotions, or bodily sensations⁴⁵. In the short term, avoidance may serve as a coping strategy that helps individuals maintain emotional stability. Yet, when it becomes a habitual and rigid response, it tends to produce the opposite effect, aggravating internal distress and obstructing the pursuit of meaningful life goals^46,47. Studies have consistently found that experiential avoidance correlates positively with social media addiction⁴⁸. People who often suppress or escape from their emotional discomfort are less capable of effective problem-solving, which may amplify negative emotions and promote repetitive online engagement^49,50. Empirical findings further support that experiential avoidance is an important predictor of social media addiction⁵¹. Based on these insights, we propose that experiential avoidance moderates the relationship between depression and social media addiction among teenagers (H3).

Although previous research has revealed a connection between depression and social media addiction, several gaps remain. First, most existing studies rely on cross-sectional data, which cannot clarify the temporal sequence or potential causal directions. Depression may act both as a cause and a consequence of social media addiction, yet this dynamic process has not been sufficiently verified. To address this issue, the current study adopted a three-wave longitudinal design spanning six months, which allows us to examine the sustained effects of depression on social media addiction among teenagers from a developmental perspective. This approach provides stronger empirical support for understanding how negative emotions interact with addictive behaviors over time. Second, earlier studies have mostly focused on emotion regulation or coping strategies, while the roles of emotional processing and experiential avoidance have received less integrated attention. Difficulty describing feelings, as a key aspect of alexithymia, captures impairments in recognizing and expressing internal emotional states. Experiential avoidance, on the other hand, reflects the tendency to evade or suppress distressing emotions. Both are closely related to depression and addiction tendencies. By combining these two constructs in one framework, this study investigates how they jointly moderate the depression–social media addiction link, extending existing theory through the interaction between emotional processing deficits and avoidance-based coping styles. Third, adolescence is a unique developmental stage characterized by emotional volatility, identity exploration, and sensitivity to social feedback. Teenagers are particularly prone to both excessive social media use and mental health problems. Yet, longitudinal research focusing specifically on this group remains limited, with most work centered on college students or adults. By targeting teenagers, this study examines the long-term connection between depression and social media addiction through a developmental lens, offering deeper insights into how emotional regulation abilities and online behaviors evolve during this critical life stage.

In summary, this study offers three key contributions:

1. It employs a three-wave longitudinal design to clarify the temporal dynamics between depression and social media addiction among teenagers.

2. It introduces difficulty describing feelings and experiential avoidance as key moderating factors, revealing how emotional processing and avoidance jointly shape addictive behaviors.

3. It focuses specifically on the teenage population, enriching the understanding of psychological mechanisms underlying social media addiction in adolescence.

The Present Study Based on the theoretical frameworks and empirical evidence discussed above, the present study proposes the following hypotheses:

Hypothesis 1 (H1): Depression will be positively associated with social media addiction.

Hypothesis 2 (H2): Difficulty describing feelings will moderate the relationship between depression and social media addiction.

Hypothesis 3 (H3): Experiential avoidance will moderate the relationship between depression and social media addiction.

Based on these hypotheses, a theoretical model was constructed (see Fig. 1).

Fig. 1.

Hypothetical model diagram.

Methods

Participants

Participants completed the questionnaire within approximately 10 min.The study was approved by the Medical Ethics Committee of the institution where the author is affiliated, ensuring that the research design and data collection procedures adhered to ethical and legal standards. All procedures followed the guidelines set by the Ethics Committee, enhancing the study’s credibility and participants’ trust. During data cleaning, questionnaires displaying consecutive identical responses, regular response patterns (e.g., wave-like answering), or failing the built-in validity checks were excluded. At Stage 1 (T1), a total of 3,812 valid questionnaires were obtained. Stage 2 (T2) was conducted in February 2025 and assessed difficulty describing feelings and experiential avoidance. After data matching using student identification numbers and excluding invalid or unmatched responses, 3,436 valid questionnaires were retained at T2, corresponding to an effective response rate of 90.13% relative to T1. Stage 3 (T3) was conducted in May 2025 and focused on social media addiction. Following the same matching and data cleaning procedures, 3,184 participants provided complete and valid data across all three waves, yielding an effective response rate of 92.67% relative to T2 and an overall retention rate of 83.52% from T1 to T3. The primary reasons for attrition included absenteeism, leave of absence, transfer to a different major, and exclusion due to careless responding. The final sample had a mean age of 14.83 years (SD = 1.40). Detailed descriptive statistics for demographic variables are presented in Table 1.

Table 1.

Descriptive statistics of population variables.

Items	Category	N	Percent
Gender	Boys	1577	49.53%
	Girls	1607	50.47%
Grade	Primary School Grade 5	37	1.16%
	Primary School Grade 6	109	3.42%
	Junior Secondary Grade 1	381	11.97%
	Junior Secondary Grade 2	734	23.05%
	Junior Secondary Grade 3	118	3.71%
	Senior Secondary Grade 1	1206	37.88%
	Senior Secondary Grade 2	576	18.09%
	Senior Secondary Grade 3	23	0.72%
Place of residence	Village	1649	51.79%
	Towns	1535	48.21%
Living on campus	Live on campus	2086	65.51%
	Not live on campus	1098	34.49%
Only child status	Only children	360	11.31%
	Non-only children	2824	88.69%

Attrition analysis was conducted to examine potential systematic differences between participants who completed all three waves of data collection (retained sample) and those who did not (dropout sample). The two groups were compared on baseline demographic variables (age and gender) and T1 depression scores. Independent-samples t tests were used for continuous variables, and chi-square tests were applied for categorical variables. The results showed no significant differences between the retained and dropout groups on age, gender, or T1 depression (all p > 0.05), indicating that attrition was unlikely to introduce substantial systematic bias.

Measures

Depression

Depressive symptoms were assessed using the Patient Health Questionnaire-2 (PHQ-2), which measures the frequency of depressed mood and anhedonia over the past two weeks. The PHQ-2 is a brief and widely used screening instrument for depressive symptoms and has been applied in adolescent populations in both Western and Chinese contexts^52–54. Each item is rated on a 4-point Likert scale ranging from 1 (not at all) to 4 (nearly every day), yielding a total score from 2 to 8, with higher scores indicating greater depressive symptom severity⁵⁵. In the present sample, the PHQ-2 demonstrated acceptable internal consistency (Cronbach’s α = 0.706). To further evaluate its construct validity, convergent validity was examined by correlating PHQ-2 scores with theoretically related variables measured in the study. As expected, PHQ-2 scores were positively and significantly associated with social media addiction at subsequent waves, as well as with difficulty describing feelings and experiential avoidance (all p < 0.001), supporting the scale’s convergent validity in this sample of Chinese adolescents. The PHQ-2 was selected over longer measures such as the PHQ-9 to reduce participant burden and fatigue. Prior research has shown that the PHQ-2 captures the core affective components of depression and performs comparably to longer versions in screening contexts. Taken together, these findings support the appropriateness of using the PHQ-2 as an indicator of depressive symptoms in the present longitudinal study.

Social media addiction

Social media addiction was measured using the Bergen Social Media Addiction Scale (BSMAS)⁵⁶. The BSMAS consists of 6 items⁵⁷, each rated on a 5-point Likert scale ranging from 1 (very rarely) to 5 (very often). Higher total scores reflect higher levels of social media addiction. In the present sample, the Cronbach’s α was 0.801, indicating good reliability.

Difficulty describing feelings

Difficulty describing feelings was assessed using the Difficulty Describing Feelings subscale of the Toronto Alexithymia Scale (TAS), originally developed by Bagby et al.⁵⁸and revised for the Chinese population by Zhu et al.⁵⁹. The subscale includes five items, each rated on a 5-point Likert scale from 1 (strongly disagree) to 5 (strongly agree). Total scores range from 5 to 25, with higher scores indicating a greater level of difficulty describing feelings. The Cronbach’s α for this subscale in the current study was 0.893, reflecting excellent internal consistency.

Experiential avoidance

Experiential avoidance was measured using the Acceptance and Action Questionnaire–II (AAQ-II), developed by Fledderus et al.⁶⁰and adapted and validated in Chinese by Cao et al.⁶¹. The scale comprises seven items, rated on a 7-point Likert scale ranging from 1 (never true) to 7 (always true). The sum of all items represents the participant’s level of experiential avoidance, with total scores ranging from 7 to 49. Higher scores indicate stronger tendencies toward experiential avoidance. In this study, the Cronbach’s α was 0.915, suggesting excellent reliability.

Data processing and analysis

All statistical analyses were conducted using SPSS 26.0, while AMOS software was used to construct and evaluate the structural equation model (SEM).First, common method bias was tested; when the variance explained by the first factor was below 40%, it indicated no significant bias⁶². Next, multicollinearity was examined, and all variance inflation factors (VIF) were below 5, suggesting no serious multicollinearity issues⁶³.Descriptive statistics and correlation analyses were performed to examine the demographic characteristics and bivariate relationships among key variables. Before model testing, all study variables were standardized.A structural equation model was then constructed in AMOS to examine the relationship between depression and social media addiction, and to test the moderating roles of difficulty describing feelings and experiential avoidance. Model parameters were estimated using the maximum likelihood estimation (MLE) method. Model fit was evaluated using multiple indices, including the Comparative Fit Index (CFI) and the Root Mean Square Error of Approximation (RMSEA)⁶⁴.To enhance the robustness of the findings, bootstrapping with 5,000 resamples was applied to estimate the significance of indirect effects, with 95% confidence intervals (95% CI) reported⁶⁵. Demographic variables, including age, grade, only-child status, residential location, and boarding status, were included as covariates to control for potential confounding effects. The significance threshold was set at p < 0.05.

Results

Common method bias and multicollinearity test

Given that all data in this study were obtained through self-report questionnaires, there was a potential risk of common method bias. To address this issue, Harman’s single-factor test was conducted by performing an exploratory factor analysis on all measurement items. The results revealed that four factors with eigenvalues greater than 1 were extracted, with the first factor accounting for only 22.17% of the total variance—well below the commonly accepted threshold of 40%.This finding indicates that no significant common method bias was present in the data, thereby supporting the reliability of the measurements and enhancing the robustness of the subsequent analyses.

A multicollinearity test was conducted for the main independent variables. The results showed that all variance inflation factor (VIF) values were below 5, indicating that no severe multicollinearity existed among the variables.

Descriptive analysis

Table 2 presents descriptive statistics and t-test results for the four core variables—depression, difficulty describing feelings, experiential avoidance, and social media addiction—across different demographic groups.Regarding gender differences, female teenagers scored significantly higher than male teenagers in depression (t = − 3.97, p < 0.001), difficulty describing feelings (t = − 2.81, p < 0.01), experiential avoidance (t = − 2.76, p < 0.01), and social media addiction (t = − 4.65, p < 0.001).For residential location, teenagers from rural areas scored significantly higher than those from urban areas in difficulty describing feelings (t = − 2.07, p < 0.05) and social media addiction (t = − 2.31, p < 0.05), while no significant differences were found in depression and experiential avoidance.Regarding only-child status, non-only children exhibited significantly higher scores in depression (t = − 4.15, p < 0.001), difficulty describing feelings (t = − 3.18, p < 0.01), and experiential avoidance (t = − 2.55, p < 0.05) compared to only children, whereas no significant difference was observed in social media addiction.For boarding status, boarding students scored significantly higher than non-boarding students in depression (t = 2.45, p < 0.05) and difficulty describing feelings (t = 3.07, p < 0.01), while no significant differences were found in experiential avoidance and social media addiction. Overall, gender, residential location, and only-child status emerged as key demographic factors influencing the primary variables in this study.

Table 2.

Describes the analysis.

Variables		T1 depression		T2 difficulty describing feelings		T2 experiential avoidance		T3 social media addiction
		Mean	Sd	Mean	Sd	Mean	Sd	Mean	Sd
Gender	Boys	4.00	1.53	14.31	3.99	21.67	10.44	13.93	5.09
	Girls	4.21	1.49	14.70	3.85	22.68	10.12	14.74	4.77
	t	−3.97***		−2.81**		−2.76**		−4.65***
Place of Residence	Towns	4.06	1.55	14.35	4.07	22.31	10.50	14.13	4.86
	Village	4.41	1.49	14.64	3.79	22.06	10.08	14.54	5.01
	t	−1.62		−2.07*		0.67		−2.31*
Only child status	Only children	3.79	1.42	13.84	4.27	20.84	10.62	13.93	5.03
	Non-only children	4.14	1.52	14.59	3.87	22.35	10.23	14.39	4.93
	t	−4.15***		−3.18**		−2.55*		−1.67
Live on campus	Live on campus	4.15	1.52	14.66	3.74	22.15	10.42	14.35	4.89
	Not live on campus	4.01	1.50	14.20	4.25	22.24	10.04	14.33	5.03
	t	2.45*		3.07**		−0.24		0.11

*: p<0.05;**: p<0.01; ***: p<0.001.

Correlation analysis

As shown in Table 3, depression was significantly positively correlated with difficulty describing feelings (r = 0.403, p < 0.001), experiential avoidance (r = 0.485, p < 0.001), and social media addiction (r = 0.314, p < 0.001).Additionally, difficulty describing feelings was positively correlated with experiential avoidance (r = 0.405, p < 0.001) and social media addiction (r = 0.228, p < 0.001). Experiential avoidance was also positively associated with T3 social media addiction (r = 0.319, p < 0.001).Overall, the core variables exhibited moderate, significant positive correlations, suggesting that higher levels of depression, difficulty describing feelings, and experiential avoidance are associated with an increased tendency toward social media addiction.

Table 3.

Correlation analysis.

Variables	VIF	1	2	3
1 T1 depression	1.402	-	-
2 T2 difficulty describing feelings	1.128	0.403***	-	-
3 T2 experiential avoidance	1.404	0.485***	0.405***	-
4 T3 social media addiction	-	0.314***	0.228***	0.319***

***: p<0.001.

Moderated mediation analysis

Table 4 presents the fit indices, criteria, and evaluation results for the structural equation model (SEM). The overall model fit was acceptable (χ²/df = 2.41, RMSEA = 0.062, RMR = 0.057, GFI = 0.902, CFI = 0.924, IFI = 0.927). Although some indices (e.g., GFI) were slightly below ideal thresholds, the overall pattern of fit indices met commonly accepted standards, suggesting that the proposed SEM adequately represented the observed data.

Table 4.

Model fitness test.

Model fit indices	Model fit criteria	Model results	Model fit evaluation
χ2/df	1–3	2.41	Acceptable
RMSEA	< 0.08	0.062	Good fit
RMR	< 0.08	0.057	Good fit
GFI	> 0.90	0.887	Acceptable
CFI	> 0.90	0.924	Good fit
IFI	> 0.90	0.927	Good fit

Table 5; Fig. 2 present the associations of depression, difficulty describing feelings, and experiential avoidance with social media addiction, as well as their interaction terms. The results indicated that T1 depression (β = 0.178, p < 0.001), T2 difficulty describing feelings (β = 0.077, p < 0.001), and T2 experiential avoidance (β = 0.194, p < 0.001) were each positively associated with T3 social media addiction. Moreover, the interaction analyses revealed significant moderating effects: the depression × difficulty describing feelings interaction (A × B) was positively associated with social media addiction (β = 0.045, p = 0.008), and the depression × experiential avoidance interaction (A × C) was also significant (β = 0.041, p = 0.017). These findings indicate that difficulty describing feelings and experiential avoidance moderate the longitudinal association between depression and social media addiction, such that the strength of this association increases at higher levels of these moderators.

Table 5.

Moderation model testing.

Outcome variables	Predictor variables	Standardized path coefficient	S.E.	C.R.	P
T3 social media addiction	T1 depression (A)	0.178	0.016	10.505	<0.001
	T2 difficulty describing feelings (B)	0.077	0.016	4.566	<0.001
	T2 experiential avoidance (C)	0.194	0.016	11.444	<0.001
	A × B	0.045	0.014	2.671	0.008
	A × C	0.041	0.014	2.394	0.017

Fig. 2.

Moderating model (*: p<0.05; **: p < 0.01; ***: p < 0.001).

Table 6; Figs. 3, 4 and 5 present simple slope analyses at low, medium, and high levels of difficulty describing feelings and experiential avoidance. The results showed that the association between depression and social media addiction was statistically significant across all levels of both moderators (p < 0.001) and became stronger as moderator levels increased. Specifically, for difficulty describing feelings, the strength of the association increased from 0.100 (low) to 0.133 (medium) and 0.166 (high). For experiential avoidance, the association increased from 0.178 (low) to 0.210 (medium) and 0.243 (high). When both moderators were at high levels, the association was strongest (β = 0.243, 95% CI [0.196, 0.290]), suggesting a synergistic pattern in which emotional expression difficulties and experiential avoidance jointly intensified the depression–social media addiction association. Overall, these results provide support for the proposed hypotheses, indicating that the longitudinal association between depression and social media addiction is amplified at higher levels of difficulty describing feelings and experiential avoidance.

Table 6.

The moderating effect of different levels of difficulty describing feelings and experiential avoidance.

Moderating variable (T2 difficulty describing feelings)	Moderating variable (T2 experiential avoidance)	Effect size	SE	t	Lower limit 95%CI	Upper limit 95%CI
Low	Low	0.100	0.028	3.552***	0.045	0.156
	Medium	0.133	0.026	5.025***	0.081	0.185
	High	0.166	0.033	5.026***	0.101	0.230
Medium	Low	0.139	0.026	5.288***	0.088	0.191
	Medium	0.172	0.020	8.707***	0.133	0.210
	High	0.204	0.024	8.564***	0.158	0.251
High	Low	0.178	0.033	5.352***	0.133	0.243
	Medium	0.210	0.024	8.613***	0.163	0.258
	High	0.243	0.024	10.154***	0.196	0.290

*: p<0.05;***: p<0.001.

Fig. 3.

Simple slope diagram (low experiential avoidance).

Fig. 4.

Simple slope diagram (medium experiential avoidance).

Fig. 5.

Simple slope diagram (high experiential avoidance).

Discussion

This study examined the longitudinal association between T1 depression and T3 social media addiction among teenagers and further explored whether difficulty describing feelings and T2 experiential avoidance moderated this association. The findings indicated that higher levels of T1 depression were positively associated with higher levels of T3 social media addiction. Moreover, both T2 difficulty describing feelings and T2 experiential avoidance functioned as significant moderators of this association. These results help clarify the psychological processes linking depressive symptoms with social media addiction during adolescence. Specifically, they suggest that teenagers who experience greater difficulty articulating emotions or who tend to avoid negative internal experiences may show a stronger association between depressive moods and subsequent addictive patterns of social media use. By illustrating how emotional processing deficits and avoidance-oriented coping tendencies interact over time, this study contributes to a more nuanced understanding of the risk mechanisms underlying adolescent social media addiction and highlights potential directions for early psychological intervention.

The observed positive association between depression and social media addiction among teenagers is consistent with previous research⁶⁶. Prior studies have suggested that a central motivation underlying addictive behaviors, including excessive social media use, involves attempts to regulate negative emotional states⁶⁷. Depression, anxiety⁶⁸, and loneliness⁶⁹ have been widely identified as psychological correlates of maladaptive digital behaviors. Although emotional problems may also emerge as consequences of addiction⁷⁰, existing evidence indicates that they often precede problematic use, as individuals may turn to social media for emotional relief or compensatory regulation⁷¹. When teenagers experience persistent sadness or emotional tension, online platforms may offer temporary distraction or perceived social support, which can gradually foster habitual and potentially compulsive engagement⁷². Cross-sectional studies among adolescents consistently document associations between depression, anxiety, and problematic patterns of social media use⁷³. In line with the I-PACE model, depression has been conceptualized as an important antecedent associated with problematic internet-related behaviors⁷⁴. Empirical findings further suggest that depressive symptoms are linked to greater severity of social media dependence⁷⁵ and a higher likelihood of addictive use patterns among teenagers⁷⁶. Taken together, these findings provide support for Hypothesis 1 (H1), indicating a positive longitudinal association between depression and social media addiction in adolescent populations.

The moderating role of difficulty describing feelings identified in this study (H2) is consistent with prior research suggesting that deficits in emotional expression are an important vulnerability factor associated with social media addiction^77,78. From a theoretical perspective, Maslow’s hierarchy of needs posits that mental well-being depends on the fulfillment of basic psychological needs⁷⁹. When such needs are insufficiently satisfied, individuals are more likely to experience negative emotions and psychological distress, which may increase their tendency to seek alternative forms of emotional compensation. In this context, social media—owing to its anonymity, convenience, and immersive features—may function as an accessible avenue for emotional regulation and temporary relief⁸⁰. This process is further supported by uses and gratifications theory⁸¹, which proposes that repeated reliance on social media for immediate emotional satisfaction may gradually develop into maladaptive or addictive patterns of use⁸². Although the interaction between depression and difficulty describing feelings was statistically significant, the effect size was relatively small (β = 0.045), indicating a modest but meaningful moderating influence. Several factors may account for this magnitude. From a temporal perspective, the extended intervals between measurement waves in the longitudinal design may have attenuated interaction effects, given that adolescents’ emotional states and coping strategies are subject to developmental and contextual fluctuations. Methodologically, the reliance on self-report measures and standardized variables in a large sample may have further constrained variability in interaction terms, resulting in smaller standardized coefficients despite clear statistical significance. From a theoretical standpoint, difficulty describing feelings is unlikely to function as a single dominant moderator; instead, it may operate as one element within a broader constellation of emotional processing and regulatory vulnerabilities. Adolescents with higher levels of difficulty describing feelings often experience challenges in identifying and verbalizing emotions and in interpreting others’ emotional cues⁸³, which may be associated with interpersonal difficulties⁸⁴, reduced self-confidence⁸⁵, and a stronger reliance on external emotional outlets⁸⁶. Social media, being easily accessible and relatively less emotionally demanding, may offer a seemingly safe and controllable environment for communication⁸⁷. Over time, however, increasing dependence on online interactions may be associated with diminished real-world social engagement, heightened social isolation, and an incremental elevation in the risk of social media addiction⁸⁸. Thus, even a modest moderation effect may carry meaningful implications at the population level, particularly given the high prevalence of depressive symptoms and social media use during adolescence.

The present findings indicate that experiential avoidance functions as a significant moderator in the association between depression and social media addiction among teenagers, providing support for Hypothesis 3 (H3) and converging with prior research⁸⁹. Previous studies have suggested that rigid and inflexible patterns of experiential avoidance are positively associated with the severity of internet-related addictive behaviors⁹⁰. Because social media offers an easily accessible means of disengaging from distressing emotions or maladaptive thoughts, adolescents with higher levels of experiential avoidance may be more inclined to rely on social media as an emotion-focused coping strategy, which may, over time, be associated with more problematic patterns of use⁹¹. These findings are consistent with the addiction steady-state model and compensatory internet use theory, both of which emphasize that maladaptive online behaviors can emerge when individuals repeatedly turn to digital environments to cope with psychological distress. Importantly, the present results highlight the distinctive role of experiential avoidance as a vulnerability factor that may intensify the longitudinal association between depression and social media addiction. Moreover, the instant feedback and social comparison features inherent to social media platforms may place adolescents high in experiential avoidance at greater risk of becoming embedded in a self-reinforcing cycle. Exposure to upward social comparisons, particularly among individuals with lower self-acceptance, may amplify negative emotions and feelings of inadequacy, which can, in turn, be associated with more frequent social media engagement as a means of short-term emotional relief. Although such compensatory use may temporarily alleviate distress, it may also be linked to an increased risk of persistent and maladaptive social media use over time.

It is important to interpret the moderating effects identified in this study in light of the timing of measurement. Difficulty describing feelings and experiential avoidance were assessed at T2, after depression was measured at T1. Although these constructs are often conceptualized as relatively stable emotion regulation tendencies, accumulating evidence suggests that they also include state-like components that may fluctuate in response to psychological distress. Consequently, the levels of difficulty describing feelings and experiential avoidance observed at T2 may have been partially shaped or intensified by earlier depressive symptoms. From this perspective, the present findings should not be interpreted as reflecting only pre-existing trait vulnerabilities. Rather, they suggest that adolescents who, following depressive experiences, exhibit greater difficulty in emotional expression or stronger tendencies toward experiential avoidance show a stronger temporal association between earlier depression and later social media addiction. In other words, emotion regulation difficulties and avoidant coping strategies may emerge or become more pronounced in the context of depression and, in turn, amplify adolescents’ reliance on social media as a compensatory or avoidant coping mechanism.

In addition, this study explored the longitudinal association between depression and social media addiction among teenagers, with particular attention to the moderating roles of difficulty describing feelings and experiential avoidance—areas that have received relatively limited attention in prior research. The findings indicate that depressive symptoms are prospectively associated with later levels of social media addiction, and that this association varies as a function of adolescents’ emotional processing difficulties and avoidant coping tendencies. Teenagers who experience greater difficulty describing feelings or who tend to avoid emotional discomfort may be more inclined to engage in social media use as a form of emotional disengagement or psychological compensation when experiencing depressive symptoms, which is associated with higher levels of addictive use over time. From a theoretical perspective, this study contributes in several ways. First, by adopting a three-wave longitudinal design, the findings provide evidence for temporal ordering between depression and subsequent social media addiction, extending insights beyond what cross-sectional studies can offer. However, it should be noted that social media addiction was not assessed at baseline, which limits the ability to control for autoregressive effects. As such, the observed longitudinal associations may partially reflect the stability of social media addiction over time rather than changes attributable solely to depressive symptoms. Second, the inclusion of both difficulty describing feelings and experiential avoidance helps clarify how deficits in emotional awareness and avoidant coping strategies jointly shape adolescents’ susceptibility to problematic social media use, thereby extending emotion regulation theory within the context of behavioral addiction. Third, by integrating emotional and behavioral factors within a longitudinal framework, the study highlights how individual differences in emotion regulation capacity and avoidance tendencies interact with depressive symptoms to influence patterns of social media engagement. From a practical standpoint, these findings underscore the importance of early identification of depressive symptoms and maladaptive emotion regulation patterns in adolescents. Interventions targeting emotional expression skills and acceptance-based coping strategies, such as emotional awareness training and acceptance and commitment therapy, may help reduce experiential avoidance and reliance on social media as an emotional coping tool. In addition, collaboration between schools and families to foster supportive emotional environments and open communication may help address the psychological processes underlying excessive social media use.

Despite these contributions, several limitations should be acknowledged. First, all variables were assessed via self-report questionnaires, which may introduce recall bias and shared method variance. Second, some potentially relevant demographic variables, such as socioeconomic status, were not collected and should be considered in future research. Third, depression was assessed using the PHQ-2, which may not fully capture the multidimensional nature of depressive symptoms. Importantly, because social media addiction was not measured at the first wave, the study could not control for baseline levels or autoregressive effects, limiting causal inference and raising the possibility that observed associations reflect stability in social media addiction over time. Future studies should incorporate baseline assessments of social media addiction and employ longer-term designs with more measurement points to better disentangle directionality and developmental trajectories. Finally, as the sample consisted of Chinese teenagers, caution is warranted when generalizing the findings to other cultural or developmental contexts.

Conclusion

This research employed a three-wave longitudinal design and identified a prospective association between depression and later levels of social media addiction among teenagers. In addition, both difficulty describing feelings and experiential avoidance were found to play significant moderating roles in this longitudinal association. Adolescents who struggled to articulate emotions or who tended to avoid psychological distress appeared more likely to engage in social media use as a way of coping with depressive moods, which was associated with higher levels of addictive use over time. These findings extend existing theoretical perspectives on the relationship between depression and addictive behaviors by providing longitudinal evidence that highlights the role of emotional processing difficulties and avoidant coping tendencies in shaping adolescents’ vulnerability to social media addiction. From a practical perspective, the results underscore the importance of supporting adolescents in developing emotional awareness and expression skills. Encouraging the adoption of more adaptive emotion regulation strategies may help reduce reliance on social media as an emotional coping outlet. Furthermore, incorporating emotional expression exercises and interventions targeting experiential avoidance into school-based mental health programs may foster resilience and strengthen adolescents’ capacity to manage negative emotions and stress. Such approaches may contribute to a lower likelihood of problematic social media use and promote better overall psychological well-being and social adjustment.

Author contributions

Pingfan Liu12345, Jingbo Wang235, Qiangzhi Zuo235, Tong Han35, Junwei Zhang125. 1 Conceptualization; 2 Methodology; 3 Data curation; 4 Writing—Original Draft; 5 Writing—Review & Editing; 6 Funding acquisition.

Data availability

The datasets generated and/or analysed during the current study are not publicly available due [our experimental team’s policy] but are available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

The study was approved by the Biomedicine Ethics Committee of Jishou University before the initiation of the project (JSDX-2024-0135). And informed consent was obtained from the participants and their guardians before starting the program. We confirm that all the experiment is in accordance with the relevant guidelines and regulations such as the declaration of Helsinki.