The differences in symptom networks of depression, anxiety, and sleep in college students with different stress levels

Abstract

Background

Stress is closely related to depression, anxiety, and sleep problems. However, few studies have explored the complex symptom-level relationships among these variables at different stress levels among college students.

Methods

From March to April 2024, a survey was conducted using a convenience sampling method in three universities in Daqing City, Heilongjiang Province. A total of 7,845 participants (2,447 males and 5,398 females) were assessed using the Psychological Stress Tolerance Index (PSTR), the General Health Questionnaire-20 (GHQ-20), and the Self-Rating Scale of Sleep (SRSS). Based on the GHQ-20 scores, college students were categorized into low, medium, and high-stress levels. Non-parametric tests and Post-hoc tests were conducted to explore the impact of stress levels on depression, anxiety, and sleep. Network analysis methods were used to reveal the differences in the symptom networks of depression, anxiety, and sleep among college students at different stress levels.

Results

Non-parametric test results indicate significant differences in depression, anxiety, and sleep scores among high, medium, and low-stress groups. Post-hoc tests reveal that the high-stress group scores significantly higher in depression, anxiety, and sleep than the medium and low-stress groups. The medium-stress group scored significantly higher than the low-stress group. Network analysis shows that the core symptoms in the low-stress group are “Difficulty falling asleep”, “Anxious and restless”, and “Taking sleeping pills”, with bridging symptoms including “Hopeless future”, “Feeling useless”, “Life is a battlefield”, and “Anxious and restless”. For the medium-stress group, the core symptoms are “Difficulty falling asleep”, “Easily awakened after sleeping”, and “Life is hopeless”, with bridging symptoms including “Feeling useless”, “Life is a battlefield”, “Anxious and restless”, and “Taking sleeping pills”. In the high-stress group, the core symptoms are “Difficulty falling asleep”, “Feeling useless”, and “Anxious and resless”, with bridging symptoms including “Feeling useless”, “Life is a battlefield”, “Anxious and restless”, and “Stress hinders tasks”.

Conclusion

Stress exacerbates depression, anxiety, and sleep problems among college students, with differences in core symptoms and bridging symptoms of depression, anxiety, and sleep disturbances at varying levels of stress. Therefore, precise interventions can be implemented based on the core and bridge symptoms of the three networks, further improving university students’ physical and mental health.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-024-21161-w.

Introduction

Stress is defined as a tense state that occurs when an individual perceives a threat from internal or external stimuli or feels insufficient in their ability to cope with the external environment [1]. College students become vulnerable groups because of their particular life development stage, new living environment and interpersonal communication environment, academic expectations, difficult employment, and other factors. This vulnerability is amplified by the transition from late youth to adulthood [2], So college students face more stress. In a survey that delved into the stressors of college students, the results revealed a remarkable result: 70% of college students suffer from moderate levels of stress [3]. In particular, academic pressure is the most prominent for Chinese students in this pressure picture, accounting for 86.67% of the total stress, becoming an overwhelming major factor [4]. This status quo is deeply rooted in the Chinese education system, which is deeply influenced by Confucian philosophy and tends to regard exams as the core measure of academic achievement, thus increasing the academic pressure on students [5]. This is closely followed by interpersonal stress, which accounts for 70.00% of total stress [4]. That is because, during the university stage, students must independently adapt to new and complex social demands that require the development of complex social skills such as role-taking and conflict management, challenges that can create interpersonal stress in many areas of life [6].

These stresses have a multi-dimensional impact on college students. According to Lazarus’ stress coping model, stress is a key factor leading to a series of physical and mental diseases, such as anxiety, depression, sleep disorders, and risk-taking behavior [7, 8]. A survey involving 500 respondents revealed that with the increase in stress levels, anxiety and depression levels increased significantly [9], and this result was further verified in a cross-country comparison [10]. This confirms the unhealthy shaping effect of stress on mental state. Significantly, during the epidemic period, the life pressure, academic pressure, and economic pressure of college students have increased, which makes the prevalence rate of depression and anxiety as high as 40% [11]. Moreover, from a physiological perspective, these stresses affect remote health outcomes in two main ways. On the one hand, stress changes individual immune and hormonal functions by affecting gene expression and biological regulatory mechanisms, thereby increasing the susceptibility of college students to anxiety, depression, and sleep disorders [6, 12, 13]. A meta-analysis showed that stress increases susceptibility to hormone-induced cognitive impairment by promoting DNA methylation and histone acetylation, altering the genetic change profile of glucocorticoid signaling pathways (e.g., NR3C1 and FKBP5), serotonin signaling pathways (e.g., SLC6A4), and neurotrophic factors (e.g., BDNF). Thus exacerbating symptoms of anxiety, depression, and sleep disorders [13]. On the other hand, stress can lead to the emergence of unhealthy behaviors that adversely affect health outcomes [14]. Stress will affect individual executive function and reduce the self-control of college students. When faced with stress, they often adopt maladaptive coping strategies, such as overeating, smoking, drinking, and smartphone addiction [15–17]. While these behaviors may provide short-term psychological relief, in the long term, they can lead to anxiety, depression, and sleep problems [15, 18].

It is worth noting that anxiety, depression, and sleep problems are complex interrelationships. Self-reported poor sleep quality is one of the common features and diagnostic criteria of anxiety and depression [13]. Data show that 90% of patients with depression and 70% of patients with anxiety have sleep problems [19]. Other studies have shown that sleep problems are predictors of depression and anxiety. One meta-analysis found that people with insomnia had a two-fold increased risk of depression compared to people without insomnia [20] and had higher scores for somatization, compulsion, and psychological distress. Similarly, Nyer [21] found that compared with college students without sleep disorders, college students with sleep disorders showed more symptoms of depression and anxiety, as well as more cognitive dysfunction. An intervention study conducted by Taylor et al. [22] also provided strong evidence for this, showing that cognitive behavioral therapy for insomnia is also effective in patients with both insomnia and depressive symptoms, and is more effective than antidepressant therapy alone. These findings suggest a bilateral relationship between sleep disorders and anxiety and depression. And this complex relationship seems to be related to personal experiences of stress. On the one hand, the physiological response triggered by stress can directly induce depression and anxiety; on the other hand, stress can indirectly aggravate depression and anxiety symptoms through sleep disorders. This interaction between stress and sleep disorders was hinted at in a recent longitudinal study by Feldman et al. [23], who found that poor sleep quality at baseline predicted an increase in depressive symptoms three months later, while perceived stress levels at baseline did not. This suggests that sleep disturbances mediate the relationship between stress and depression. Second, similar results were found in the study by Matsuda and Kikutani [10], who investigated the relationship between depressive symptoms, sleep disturbances (including insomnia, sleepiness, and nightmares), and stressful life events, testing a hypothetical model that predicted that more life stressors were directly related to an increase in depressive symptoms. However, this relationship is significantly mediated by sleep disturbances.

Previous studies have explored the relationship between stress and depression, anxiety, and sleep based on total scores from scales, often viewing symptoms as isolated and neglecting the interactions and triggering effects between different symptoms [24]. Network analysis methods provide a suitable approach that allows research to go beyond the overall relationship among stress, depression, anxiety, and sleep, instead elucidating the interaction patterns between different symptoms. To our knowledge, no studies have yet used network analysis to investigate the relationships among depression, anxiety, and sleep at varying levels of stress, nor the differences in core and bridge symptoms within the symptom networks of college students under different stress levels. Therefore, to address the limitations of previous studies, we employed network analysis methods, examining the symptom-level interactions among depression, anxiety, and sleep across varying stress levels by calculating the network’s topological structure, thereby filling a crucial research gap. Furthermore, by computing centrality indices, we identified the core symptoms and the bridge symptoms connecting different clusters of depression, anxiety, and sleep symptoms across different stress levels, providing a basis for implementing precise interventions. This research holds significant practical implications, as it not only deepens our understanding of the interactions among depression, anxiety, and sleep under different stress levels but also lays a foundation for developing targeted mental health strategies.

However, It must be pointed out that the anxiety, depression, and sleep in the current study are measured through self-report scales. These scales only reflect the severity levels of anxiety, depression, and sleep problems rather than clinical anxiety, depression, and sleep disorders. Therefore, we aim to achieve two objectives: (1) to further validate the impact of stress on depression, anxiety, and sleep based on previous research, and (2) to explore the core and bridge symptoms of depression, anxiety, and sleep symptom networks at different levels of stress using network analysis methods from a symptom perspective. Based on past studies, our hypotheses are: (1) There are significant differences in depression, anxiety, and sleep scores among college students at different stress levels, with the severity of depression, anxiety, and sleep issues gradually increasing as stress levels rise. (2) The core and bridge symptoms of the depression, anxiety, and sleep symptom networks differ among college students at varying stress levels.

Materials and methods

Participants

This study employed a convenience sampling method. According to the theoretical requirements for sample size in network analysis [25], the sample size should be at least higher than the total number of parameters (including threshold and pairwise association parameters). The threshold parameters = the number of nodes, and the pairwise association parameters = the total number of nodes × (total number of nodes − 1)/2. We need to construct 21 nodes in this study, so the threshold parameters are 21, and the pairwise association parameters are 210. Based on this, the minimum sample size requirement is set at 231 cases. However, to ensure the stability and reliability of model construction, it is recommended to include at least 3 to 5 subjects per parameter [25]. Therefore, we further increase the minimum sample size to 697 cases.

This study was conducted among students from three universities in Daqing City, Heilongjiang Province, using a convenience sampling method between March and April 2024. Students were recruited to complete the questionnaire online through the “WenJuanXing” online survey platform. Teachers at the universities sent out the questionnaire links to the students. At the beginning of the questionnaire, we clarified the purpose of the survey and stated that students could voluntarily participate in the study by answering the questionnaire online. The “WenJuanXing” platform allowed participants to complete the questionnaire via their mobile phones or computers. Once students had answered all the questions in the questionnaire, the data was directly uploaded to the backend of the “WenJuanXing” platform. Prior to the formal analysis of the data, we conducted data cleaning. Firstly, we excluded participants who indicated the presence of significant mental and physical illnesses based on the option “Do you have any significant mental or physical illnesses?“, and the results showed that no participants with severe mental or physical illnesses were found. Secondly, we excluded subjects whose ages were obviously outside the range typical for university students (16 to 26 years old) [24]. Furthermore, based on previous response time standards (2 s per item), we eliminated participants who had yet to answer the questions seriously [24]. Ultimately, data from 7,845 participants were included in the analysis.

Measurement

Demographic information

Demographic information encompasses participants’ age, gender, residence (rural or urban), and educational attainment (Junior College or Bachelor’s Degree).

Psychological stress tolerance index (PSTR)

The PSTR was developed by Edwards and revised in Chinese by Zhao Dandi [26]. It consists of 50 items that address stress-related physical and psychological changes. It uses a 0–4 point Likert scale (0 = Never, 1 = Rarely, 2 = Sometimes, 3 = Often, 4 = Always). Scores above 65 indicate a high-stress group, 43–65 indicate a medium-stress group, and below 43 indicate a low-stress group [27]. The questionnaire’s reliability was examined using Cronbach’s alpha (α = 0.82).

General health questionnaire-20 (GHQ-20)

The GHQ-20, initially developed by Goldberg [28] and revised domestically by Li Hong [29], measures an individual’s psychological health. It consists of 20 items covering self-affirmation, depression, and anxiety, with each item scored as “Yes” or “No,” assigned 1 and 0 points, respectively. Items 7 and 10 are scored in reverse. Higher scores on each subscale indicate greater levels of the respective condition. Depression and anxiety were analyzed online in this study. The total scale has a Cronbach’s α coefficient of 0.82, while the depression and anxiety subscales have Cronbach’s α coefficients of 0.63 and 0.64, respectively.

Self-rating scale of sleep (SRSS)

This scale was developed by Professor Li Jianming [30] and aims to assess an individual’s sleep quality over the past month. It consists of 10 items, each rated on a 1–5 scale, with scores ranging from 10 to 50 points. Higher scores indicate poorer sleep quality. The Cronbach’s α coefficient for this scale is 0.64.

Statistical analyses

This study used SPSS 27.0 and R 4.4.0 for data analysis. Categorical data were described using frequencies or percentages, with group comparisons conducted using the chi-square test. For continuous data, which were non-normally distributed, M (P25, P75) was used for description, and non-parametric tests (Kruskal-Wallis H test) were used for group comparisons. If significant differences were found among multiple groups, post-hoc tests were conducted using the Dunn-Bonferroni method, with a significance level set at P < 0.05. Additionally, R 4.4.0 was used for network analysis to explore differences in depression, anxiety, and sleep symptoms network among university students under different levels of stress. The network analysis included network estimation, visualization, centrality metric estimation, and network accuracy and stability assessment.

Symptom network estimated

The “qgraph” package in R was used to estimate the partial correlation networks of depression, anxiety, and sleep symptoms for low, medium, and high-stress groups of university students. The following procedure was applied for each partial correlation network: First, the network structure for categorical and continuous variables was estimated using mixed graphical models (MGM) [31]. In order to sparse the network and make it easier to understand, small correlations were shrunk to exactly zero using the enhanced Least Absolute Shrinkage and Selection Operator (eLASSO) method, with the related tuning parameter set using the Extended Bayesian information criterion (EBIC) [32]. Following previous recommendations, the parameter was set to γ = 0.5 to delete edges with small or unstable correlations between entries, ensuring an accurate network could be estimated [33].

The “qgraph” package was used for network visualization. In the network model, each node represents an item, and each edge represents the relationship between two nodes. Thicker edges indicate stronger relationships, while solid and dashed edges represent positive and negative correlations, respectively [34]. For ease of observation, the average Layout function from the “qgraph” package was used to position the same nodes in the same locations across different networks [35].

Centrality index

Previous research has shown that in psychometrics, closeness and betweenness are not sensitive [36]. Therefore, this study uses strength and expected influence to measure node influence. Strength refers to the sum of the absolute values of the edge weights connected to a node, while expected influence is the sum of edge weights connected to a node. Higher strength and expected influence indicate more significant influence within the network [37]. Following previous research, our study identifies the three symptoms with the highest Expected Influence as core symptoms [38]. Additionally, the “mgm” package is used to estimate predictability. Predictability refers to the variance of a node that can be explained by its neighboring nodes; higher predictability indicates greater influence from neighboring nodes, which can help control symptoms through interventions targeting these neighboring nodes [24]. Since the network model includes symptoms from different diseases, bridge strength and bridge expected influence are calculated for each node to identify those connecting different diseases, termed bridge symptoms. Bridge strength refers to the sum of the absolute values of the edge weights directly connected to other disease nodes. Bridge expected influence is the sum of the edge weights directly connected to nodes from other diseases measured in two ways: bridge expected influence (1-step) is the total sum of edge weights between a node and all nodes not within the same community, and bridge expected influence (2-step) reflects direct influence through another node on other communities [37]. Our study, in line with previous research, uses bridge expected influence (1-step) for analysis [39]. Greater bridge strength and bridge expected influence indicate a greater role of the node in connecting different network groups or communities. Our study adopts the viewpoint of Jones et al., using the 80th percentile cutoff for bridge expected influence to select bridge symptoms [40].

Accuracy and stability of the network

The “bootnet” package in R was used to estimate network accuracy and stability [32]. This study calculated the correlation stability-coefficients (CS-C) of strength, expected influence, bridge strength, and bridge expected influence. The CS-C represents the maximum proportion of reduced sample size where the correlation between centrality measures of the original and reduced samples is at least 0.7. CS-C greater than 0.25 indicates acceptable network stability, while CS-C greater than 0.5 indicates good stability [35]. We conducted a nonparametric bootstrap to evaluate the bootstrapped confidence intervals (95% CI), and a narrow CI means a reliable network [24].

Results

Common method bias test

Our study employed Harman’s single-factor test to assess common method bias. The first factor was found to explain 30.05% of the total variance, which is below the critical threshold of 40% [41]. This indicates that common method bias is not severe.

General demographic information

This study recruited 7845 eligible participants, whose ages ranged from 16 to 26 years, with a mean age of 20.39 ± 1.331 years. In terms of gender distribution, female participants comprised the larger proportion, totaling 5398 individuals, while male participants numbered 2447. Regarding educational background, participants at the undergraduate level formed the majority, amounting to 7479 individuals, with 366 participants at the college level. Furthermore, based on geographical origin, 4741 participants hailed from urban areas, and 3104 originated from rural regions.

Comparison of depression, anxiety, and sleep scores among three groups of university students

The results of the non-parametric test indicated significant differences in depression, anxiety, and sleep scores among university students with different levels of stress (all P-values < 0.01). To further investigate these group differences, Dunn-Bonferroni post-hoc tests were conducted. The results revealed that students in the high-stress group had significantly higher scores in depression, anxiety, and sleep compared to both the medium-stress and low-stress groups (P < 0.05 for all comparisons). Additionally, the medium-stress group exhibited significantly higher scores in these areas compared to the low-stress group (P < 0.05). These findings are summarized in.

Table 1.

Comparison of depression, anxiety, and sleep scores among three groups of college students [M(P25, P75)]

Symptoms	Low Stress Group (n = 3558)	Medium Stress Group (n = 2154)	High Stress Group (n = 2133)	H	P
Depression	0.49 (0.00, 1.00)	0.27 (0.00, 0.00)a	17.12 (13.00, 19.00)ab	1211.56	<0.01
Anxiety	0.61 (0.00, 1.00)	0.61 (0.00, 1.00)a	19.83 (17.00, 22.00)ab	2104.84	<0.01
Sleep	1.96 (0.00, 3.00)	1.96 (0.00, 3.00)a	24.07 (20.00, 28.00)ab	214.70	<0.01

Note Compared with the Low Stress Group, ^aP<0.05; compared with the Moderate Stress Group, ^bP<0.05

Network structure

The symptoms networks of depression, anxiety, and sleep among university students with different stress levels are illustrated in Fig. 1. Figure 1a shows the network for the low-stress group, which includes 21 nodes and 65 non-zero edges out of 210 possible edges. In this symptom network, the edge with the highest weight is between D4 (Life is hopeless) and D5 (Feeling useless) (r = 0.43), followed by the edge between X1 (Insomnia caused by worry) and X2 (Irritability affects sleep) (r = 0.40). Figure 1b presents the network for the medium stress group, consisting of 21 nodes and 58 non-zero edges out of 210 possible edges. In this symptom network, the edge with the highest weight is between nodes X1 (Insomnia caused by worry) and X2 (Irritability affects sleep) (r = 0.35), followed by the edge between X3 (Anxious and restless) and X4 (High mental stress) (r = 0.23). Figure 1c depicts the network for the high-stress group, which includes 21 nodes and 69 non-zero edges out of 210 possible edges. In this symptom network, the edge with the highest weight is between nodes X1 (Insomnia caused by worry) and X2 (Irritability affects sleep) (r = 0.40), followed by the edge between D3 (Losing confidence) and D5 (Feeling useless) (r = 0.34).

Fig. 1.

Depression, anxiety, and sleep symptom networks of college students with different stress levels. Note (a) shows the network of low-stress groups; (b) shows the network of the medium-stress group; (c) For the high-stress group network. Each node represents an item, and the circular pie chart outside the node represents the predictability of that node. The line connecting two nodes represents the partial correlation between the two nodes; the thicker the edge means the stronger the partial correlation, and the thinner the edge means the weaker the partial correlation. The solid blue line represents the positive correlation, and the dotted red line represents the negative correlation

Centrality and predictability estimates

The predictability of each symptom is represented as a ring-shaped pie chart on the outside of the nodes. In the low-stress group symptom network, node D4 (Life is hopeless) has the highest predictability, followed by D5 (Feeling useless) and X3 (Anxious and restless). In the medium-stress group symptom network, node D4 (Life is hopeless) also has the highest predictability, with D5 (Feeling useless) and D3 (Losing confidence) coming next. In the high-stress group symptom network, the highest predictability is observed for node D5 (Feeling useless), followed by S5 (Difficulty falling asleep) and D3 (Losing confidence).

The non-standardized estimates of strength and expected influence of depression, anxiety, and sleep symptom networks in college students with different stress levels are shown in Fig. 2. In the low-stress group symptom network, the highest expected influence is from S5 (Difficulty falling asleep), followed by X3 (Anxious and restless) and S9 (Taking sleeping pills). Nodes S5 (Difficulty falling asleep), X3 (Anxious and restless), and S9 (Taking sleeping pills) also have high strength, indicating that these symptoms are central in the low-stress group symptom network. In the medium-stress group symptom network, S5 (Difficulty falling asleep) has the highest expected influence, followed by S6 (Easily awakened after sleeping) and D4 (Life is hopeless). Nodes S5 (Difficulty falling asleep), S6 (Easily awakened after sleeping), and D4 (Life is hopeless) also have high strength, indicating they are central in the medium-stress group symptom network. In the high-stress group symptom network, S5 (Difficulty falling asleep) has the highest expected influence, followed by D5 (Feeling useless) and X3 (Anxious and restless). These three symptoms also have high strength, indicating they are central in the high-stress group symptom network.

Fig. 2.

Non-standardized estimates of strength and expected influence of depression, anxiety, and sleep symptom networks in college students with different stress levels

The non-standardized estimates of bridge strength and bridge expected influence of depression, anxiety, and sleep symptom networks among college students with different levels of stress are shown in Fig. 3. Our study used the 80th percentile of bridging expected influence thresholds to select bridge symptoms. In the low-stress group symptom network, D1 (Hopeless future), D5 (Feeling useless), D6 (Life is a battlefield), and X3 (Anxious and restless) are identified as bridge symptoms connecting different disease groups. In the medium-stress group symptom network, D5 (Feeling useless), D6 (Life is a battlefield), X3 (Anxious and restless), and S9 (Taking sleeping pills) are identified as bridging symptoms. In the high-stress group symptom network, D5 (Feeling useless), D6 (Life is a battlefield), X3 (Anxious and restless), and X5 (Stress hinders tasks) are identified as bridge symptoms connecting different disease groups.

Fig. 3.

The non-standardized estimates of bridge strength and bridge expected influence of depression, anxiety, and sleep symptom networks among college students with different levels of stress. (a) shows the non-standardized estimates of bridging strength for the symptom networks of the three groups; (b) displays the non-standardized estimates of bridging expected impact for the symptom networks of the three groups

Network accuracy and stability testing

The confidence intervals for the edge weight bootstrap results of the three symptom networks are very narrow, indicating that all three symptom networks have high accuracy (see Additional file 1, Fig. S1). Additionally, the CS-C for the strength, expected influence, bridge strength, and bridge expected influence of the three symptom networks are all 0.75, which means that the strength, expected influence, bridge strength, and bridge expected influence are still correlated with the original data after discarding 75% of the data (see Additional file 1, Fig. S2), suggesting that the networks exhibit strong stability.

Discussion

The university stage is critical for students transitioning from adolescence to adulthood, facing exam stress, interpersonal stress, and many uncertainties [42]. This stress can negatively impact their mental and physical health and future career potential [43]. However, despite the widespread recognition of the significance of these stressors, research on the differences in the networks of depressive, anxiety, and sleep symptoms among college students under varying levels of stress remains relatively scarce. Therefore, this study employs network analysis methods to delve deeper into this area and provide a new perspective for understanding the mental health status of college students. Here are some of our findings.

The non-parametric test results show significant differences in depression, anxiety, and sleep scores among the three groups of students. Post-hoc tests revealed that students in the high-stress group had significantly higher scores in depression, anxiety, and sleep disturbances compared to both the medium and low-stress groups, while the medium-stress group had significantly higher scores than the low-stress group. This indicates that as stress levels increase, emotional and sleep problems among college students become more severe, consistent with previous research [42, 44]. On the one hand, stress can elevate systemic inflammation through pathways such as the hypothalamic-pituitary-adrenal axis, sympathetic nervous system, and vagus nerve, with inflammatory mediators like pro-inflammatory cytokines potentially inducing depression and anxiety symptoms [45, 46]. Other studies also suggest that individuals facing multiple stressors are more likely to react negatively to daily events [47] and exhibit higher emotional instability [48]. This emotional instability can predict a range of health issues, such as low self-esteem, increased inflammation, decreased sleep quality, depression, and anxiety [49].

Our research has found similarities in the core symptoms among the three symptom networks. “Difficulty falling asleep” was a common core symptom across all three symptom networks and was strongly associated with “Easily awakened after sleep” and “Difficulty falling asleep after waking”. This association was powerful in the medium stress group symptom network, where “Difficulty falling asleep” and “Easily awakened after sleep” were both core symptoms, indicating a close relationship between them. This suggests that sleep problems among university students are widespread, affecting both the process of falling asleep and maintaining sleep even at low-stress levels. Palagini et al. suggest that stress exposure can disrupt sleep-wake processes and neurobiological feedback mechanisms [50], leading to an imbalance in sleep regulation and increased sleep reactivity, which has been shown to be highly correlated with sleep problems and psychological health levels [51].

In addition, we have found that the core symptoms differ among the three symptom networks. The core symptom of “Anxious and restless” in the low-stress group reflects significant nervousness among university students even under relatively low-stress conditions. This nervousness is primarily manifested in stress perception, as the edge weight between “Anxious and restless” and “High mental stress” is the highest. The core symptom of “Taking sleeping pills” as a common measure to address sleep issues is most strongly connected to “polyphonic nightmares”, confirming Waldman’s research and indicating that university students experience a certain degree of sleep problems even at low-stress levels [52]. In the symptom network of the medium-stress group, the core symptom of “Life is hopeless” reflects university students’ negative expectations and pessimistic views about life, as well as their perception of being unable to change their future [53]. This perception intensifies doubts about their self-doubt, with the strongest connection between “Life is hopeless” and “Feeling useless”. In the symptom network of the high-stress group, the core symptom of “Feeling useless” is most strongly associated with “Losing confidence”, as “Feeling useless” acts as an irrational belief that undermines students’ self-efficacy and increases negative self-evaluation, leading to a gradual loss of confidence. The strongest connection between the core symptom “Anxious and restless” and “High mental stress” further validates the profound impact of perceived stress on university students’ emotional states. Additionally, we have also found that the importance of irrational beliefs gradually increases as stress levels rise. Specifically, in the low-stress group’s symptom network, the core symptoms are more focused on sleep issues and nervousness. However, in the medium-stress group’s symptom network, the irrational belief “Life is hopeless” becomes the third core symptom of the network. In the high-stress group’s network, the significance of irrational beliefs is even more prominent, with “Feeling useless” rising to become the second core symptom of the network. This discovery further uncovers the negative impact of stress on university students’ cognition. This aligns with Chi’s research findings, which show a positive correlation between stress levels and irrational beliefs [54].

Our research has found similarities in the bridge symptoms among the three symptom networks. Common bridging symptoms include “Feeling useless”, “Life is a battlefield”, and “Anxious and restless”. “Feeling useless” and “Life is a battlefield” are significant indicators of depression, strongly linked to depression, anxiety, and sleep symptoms through negative cognition and, despair and helplessness about life. “Anxious and restless” serve as bridge symptoms connecting anxiety, depression, and sleep symptoms because prolonged tension activates the hypothalamic-pituitary-adrenal axis, increasing cortisol secretion. Excess cortisol affects the secretion of monoamine neurotransmitters such as serotonin and dopamine, whose abnormal expression can lead to depressive moods and sleep problems [55–57].

Additionally, there are also differences in the bridge symptoms among the three symptom networks. The bridging symptom of the low-stress group symptom network “Hopeless future” reflects college students’ pessimism and hopelessness about the future. This negative impact on future expectations is significantly associated with depression, anxiety, and sleep symptoms. Research indicates that such hopelessness exacerbates cognitive distortions, resulting in self-loathing, self-blame, and low self-esteem [58], further intensifying emotional and sleep problems among college students. The bridging symptom of medium stress group symptom networks “Taking sleeping pills” connects sleep with anxiety and depression symptoms. A survey of 60,000 people showed that long-term use of sedative-hypnotic drugs affects attention and memory, causing cognitive impairment and reducing neural sensitivity, which further aggravates depressive and anxious moods. The bridging symptom of the high-stress group symptom network “Stress hinders tasks” reflects the impact of stress on executive function, significantly associating anxiety, depression, and sleep symptoms. According to the ego depletion theory, each person’s cognitive resources are limited, and the long-term existence of stress will lead to insufficient cognitive resources, which will affect individual executive function [59, 60] and make individuals more inclined to use non-adaptive emotion regulation strategies, thus aggravating negative emotional experience [61]. In the subsequent process, more negative cognitive bias and negative attribution will be generated, which will aggravate sleep problems [62].

This study provides inspiration for intervention work in colleges and universities. In the low-stress group, “Anxious and restless” is both a core and bridge symptom of the symptom network with high predictability. Therefore, intervening in it can not only disrupt the influence of core symptoms on other symptoms within the symptom network but also sever the communication of information between bridge symptoms across different disease groups. The stress coping model shows that anxious emotions, as a reaction to stress, are often related to inadequate coping strategies [8]. To improve anxious emotions among college students in the low-stress group, interventions such as psychological drama, group discussions, and stress management workshops can enhance individual coping skills. Additionally, as anxious emotion is an emotional response, structured breathing training can help alleviate it by enhancing parasympathetic nervous system activity, maintaining gas balance, activating brain relaxation centers such as the prefrontal cortex and insula, and reducing muscle tension [63]. In the medium stress group, the core symptom “Life is hopeless” exhibits the highest predictability within this symptom network and is strongly linked to the bridging symptom of “Feeling useless”. Therefore, improving the core symptom “Life is hopeless” can disrupt the relationships between symptoms in the original network and further mitigate the mutual influences among different symptom clusters by affecting the bridge symptom “Feeling useless”. Positive psychology suggests that hopelessness is mainly due to a lack of positive resources such as positive cognition and emotions [64]. To reduce hopelessness among college students, positive psychology techniques like gratitude exercises, positive emotion cultivation, and mindfulness can be introduced to enhance overall psychological resilience and optimism. Additionally, providing social support and creating a positive learning environment for students in the medium-stress group can help alleviate hopelessness and boost their confidence in facing future challenges. In the high-stress group, “Feeling useless” is both a core and bridging symptom with the highest predictability in the network. Therefore, intervening in it can rapidly activate the network and improve the overall symptom level. According to cognitive-behavioral theory, individuals’ emotions and behaviors stem from their cognitive interpretations of events [65]. Therefore, cognitive-behavioral therapy can be used to reduce stress responses by identifying and challenging irrational beliefs about events and improving emotional regulation and sleep quality.

Limitations and future direction

This study has several limitations: (1) This study adopted a convenience sampling method, which may have impacted the representativeness of the sample. Future research could consider using a more rigorous random sampling method to provide stronger support for the study. (2) It employs a cross-sectional design, making it challenging to reveal causal relationships between symptoms. Future research could utilize longitudinal networks to explore the dynamic relationships between variables. (3) This study assessed stress, depression, anxiety, and sleep symptoms solely through questionnaire surveys, which may introduce social desirability bias. Future studies could adopt alternative assessment methods to validate the results further. (4) The majority of participants in this study were female, which may have affected the gender neutrality and universal applicability of the research findings. To more accurately assess the impact of gender on the results, future research should include gender as an important covariate in the analysis and consider comparisons and analyses across different gender samples. (5) When exploring the relationships between stress, anxiety, depression, and sleep issues, this study did not fully consider other variables that may influence these relationships, such as social support, personality traits, family functioning, the use of social media, and coping styles. The omission of these variables may limit the interpretation of the research results. Therefore, future research could further incorporate these variables into the symptom network. (6) Our samples are mainly from three universities in Daqing City, Heilongjiang Province, China, and the representativeness of the samples is limited. Therefore, future research can further expand the sample sources and consider cross-cultural comparison.

Conclusions

This study employs non-parametric tests and post-hoc analyses to investigate the impact of stress on depression, anxiety, and sleep issues among college students. The results indicate that higher stress levels are associated with more severe depression, anxiety, and sleep problems. Furthermore, we utilized network analysis to delve into the interrelationships among depression, anxiety, and sleep from the perspective of symptomatic manifestations at varying stress levels. By calculating centrality indices, we identified the core and bridge symptoms within the symptom networks. Notably, the results reveal differences in the core and bridge symptoms across the depression, anxiety, and sleep symptom networks at different stress levels. These findings provide a solid foundation for implementing precise and targeted interventions to address the mental health needs of college students under various stress conditions.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

PSTR

Psychological Stress Tolerance Index

GHQ-20

General Health Questionnaire-20

SRSS

Self-Rating Scale of Sleep

CI

Confidence intervals

CS-C

Correlation stability coefficient

eLASSO

Enhanced Least Absolute Shrinkage and Selection Operator

EBIC

The Extended Bayesian Information Criterion

Author contributions

Wei Li, Shuhui Huo, Fei Yin, and Zhengjun Wang contributed to the study design, analysis, and interpret data and drafted the manuscript. Wei Li, Zhengyu Wu, and Xueqi Zhang performed data collection, analysis, and interpretation. Jianqin Cao participated in the design and coordination of the study and revised the manuscript. All authors read and approved the final manuscript.

Funding

The present study is supported by the Humanities and Social Sciences Project, Ministry of Education (24YJAZH005), China; the Natural Science Foundation of Heilongjiang Province (LH2024H031), China; the Philosophy and Social Science Research Planning Project of Heilongjiang Province (19EDB091), China; the Fundamental Research Funds for the Provincial Universities (JFQN202103), China; the National Natural Science Foundation of China (72204068).

Data availability

The data that support the findings of this study are available from the corresponding author, upon reasonable request.

Declarations

Ethics approval and consent to participate

The study was approved by the Ethics Committee of Harbin Medical University, ensuring that all procedures adhered to strict academic research ethical standards (HMUDQ20240711001). Participants were also informed of their right to withdraw from the survey at any time. Informed consent was obtained prior to the survey, guaranteeing their anonymity and confidentiality throughout the process.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.