INTRODUCTION
Stress responses are essential for the maintenance of homeostasis and survival, and maladaptive stress responses are considered a cause of various physiological and mental disorders. According to stress appraisal theories, perceived stress depends on two factors: the frequency and intensity of the stress load, as well as the cognitive appraisal and coping with the stress [1,2]. Consequently, individual differences in how one responds to stress can significantly affect physical and mental health [3].
Stress sensitivity, also known as stress reactivity, refers to the intensity of an individual’s physiological and psychological responses to stressors [4]. Individuals with high stress sensitivity are characterized by strong negative emotional reactivity and intense physical experiences in response to minor stressors [5]. Increased stress sensitivity has been associated with various mental disorders and related traits, such as depression, anxiety, and addictive disorders [5-8]. For example, high stress sensitivity has long been recognized as a risk factor for affective disorders [9]. Additionally, increased stress sensitivity has been proposed as a predisposing factor for the development and maintenance of behavioral addictions, such as internet gaming disorder (IGD) [10,11].
Existing measures of stress sensitivity typically operate by inducing stressful situations and then assessing the stress response of subjects. The commonly used induction protocols include the Trier Social Stress Test [12] and ScanSTRESS [13], among others. Measurements typically include Visual Analogue Scales, salivary cortisol, and heart rate [14]. Behaviorally, stress sensitivity can be measured using the Experience Sampling Method, which requires participants to report their daily appraisals of activities and mood states multiple times over a period of several days [15]. The above measures are focused on the “state” intensity of stress reactivity to stressors. To the best of our knowledge, however, no existing measure is capable of capturing trait stress sensitivity in individuals.
Instead of treating it as a state feature, stress reactivity may be seen as a relatively stable trait-like tendency reflecting a long-lasting tendency toward stressful situations. Longitudinal Experience Sampling research indicates that higher levels of stress reactivity at baseline were associated with more functional health limitations 18 years later. On the other hand, although the average stress reactivity significantly declined with age over the 18-year period, the magnitude of this decline was very limited (Δstress reactivity=-0.02 per 10 years, p<0.001) [16]. Another longitudinal study indicates that while stress reactivity declined among adults across a 20-year time span, it remained stable for those aged 54 years or older at baseline [17]. These findings therefore suggest that sensitivity may be seen as a relatively stable trait that shows a sustained impact on one’s life. In support of this concept, individual differences in stress sensitivity have been associated with variance in one’s genetic characteristics [18] or early experiences [19]. For instance, individuals who had adverse life experiences in childhood tend to show higher stress sensitivity than those without [20].
The Stress Sensitivity Inventory (SSI) is a self-report scale developed by Moore et al. [21] that assesses trait stress sensitivity by evaluating the pattern of emotional, social, cognitive (e.g., self-condemnation), physiological, and behavioral stress responses in individuals. The original SSI has been demonstrated to have excellent reliability and validity. However, its application in the Chinese context is currently unknown.
Research has shown a significant association between stress sensitivity and IGD. Enhanced stress vulnerability has been recognized as a potential predisposing factor for IGD [11], suggesting that individuals with high trait stress sensitivity may engage in problematic gaming behaviors as a coping strategy. This association is supported by the diathesis-stress model, which posits that individuals with certain personality traits, such as high neuroticism, may be more vulnerable to developing IGD when faced with stressful life events [22]. In support of this perspective, previous research showed that stressful life events are positively associated with the severity of IGD symptoms, and this link is moderated by neuroticism [23]. A meta-analysis showed that the severity of IGD symptoms was positively correlated with neuroticism, a personality trait characterized by excessive negative feelings [24]. Additionally, high stress sensitivity is often accompanied by intensified negative emotions, such as depression and anxiety, which are also linked to an increased risk of IGD [25]. Therefore, understanding the interplay between trait stress sensitivity and IGD can inform the genesis and development of IGD.
The network approach is particularly useful for analyzing and visualizing complex relationships among psychopathology symptoms that would otherwise be difficult to disentangle using traditional approaches. In contrast with the traditional “common cause model,” which posits that the co-occurrence of symptoms is driven by a latent common cause [26], the network theory asserts that the covariance of symptoms is fundamental to understanding mental disorders [27]. A primary application of network analysis is the identification of “core symptoms” for specific conditions and “bridge symptoms” that link comorbid disorders [28]. Both symptom types represent key therapeutic targets for interventions aimed at treating individual disorders or their comorbidities. In recent years, the network approach has been applied to investigate the complex structure of various mental disorders, such as post-traumatic stress disorder [29], depression [30], and addictions [31,32], including IGD [33-35]. In this study, network analysis was employed to identify the core symptoms of the Chinese SSI and the bridge symptoms linking the SSI to IGD. While stress sensitivity is a well-documented risk factor for numerous mental disorders and is broadly associated with IGD symptomatology [11,23], a systematic network analysis of its core structure and the specific pathways that connect it to IGD is lacking. Elucidating this network could deepen our theoretical understanding of the stress sensitivity construct and reveal its underlying mechanisms in the development of addictive disorders.
This study aimed to evaluate the psychometric properties of the Chinese version of the SSI and to examine the relationship between the SSI and symptoms of IGD using psychological network analysis. We hypothesize that stress sensitivity is significantly associated with symptoms of IGD and that this association is bridged by specific items within the stress sensitivity framework.
METHODS
Procedure and participants
After receiving permission from the developers, we translated the SSI into simplified Chinese. The Chinese SSI was then back-translated into English, and we verified the accuracy of the translation with the developers. After the authors reached a consensus on the translation, the Chinese SSI, along with additional scales, was administered to a group of Chinese young adults through an online platform (www.wjx.cn).
The questionnaires were administered to 1,235 respondents by sharing a QR code in student chat groups at Southwest Medical University, resulting in a sample consisting largely of university students and individuals from the surrounding local area. After data cleaning, valid data from 1,085 participants were retained for analysis, yielding a response rate of 87.85%. A subset of 347 respondents completed the Chinese SSI again four weeks later for the test-retest reliability assessment.
Additional scales administered
Additional scales were selected according to research findings related to stress sensitivity. The Generalized Anxiety Disorder-7 (GAD-7) [36] and the Patient Health Questionnaire-9 (PHQ-9) [37] were used to assess anxiety and depression symptoms in respondents. The Perceived Stress Scale (PSS) [38] was used to evaluate the level of perceived stress during the last month. The PSS is a commonly used self-report instrument for perceived stress with 10 items. The Neuroticism subscale of the NEO Five-Factor Inventory [39] was used to measure the Neuroticism dimension of the five-factor model of personality. To explore whether there is a link between signs of stress sensitivity and IGD symptoms, the Chinese version of the 9-item dichotomous Internet Gaming Disorder Scale (IGDS) [40,41] was also administered to assess the IGD symptoms of respondents. The IGDS does not provide a predefined clinical cutoff for IGD diagnosis. However, because its items closely align with the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition criteria for IGD, participants endorsing five or more items were considered to be at high clinical risk for the disorder in this study. The Cronbach’s α of the IGDS was 0.89 in the current study.
Ethics
This study was approved by the Institutional Review Board of The Affiliated Hospital of the Southwest Medical University and was performed in accordance with the ethical guidelines of the Declaration of Helsinki (ethical approval number: KY2024181). Informed consent was obtained online via the first item, which allowed participants to voluntarily opt out or proceed. Participants were also informed that they were free to discontinue the survey midway without consequences. Upon completion, participants were compensated with a random bonus of 1 to 10 yuan RMB.
Data analysis
Data were analyzed using SPSS 22.0 (IBM Corp.) and R version 4.4.2 (R Core Team, 2015).
The psychometric property analysis of the Chinese version SSI
For item analysis, we calculated corrected item-total correlations for each item. The reliability of the SSI was assessed via Cronbach’s α and test-retest correlations, respectively. The validity analysis consisted of two parts: construct validity and criterion validity analysis. The construct validity was assessed using an exploratory factor analysis (EFA) and a confirmatory factor analysis (CFA). Criterion validity was evaluated by computing Pearson’s correlations between SSI scores and scores on the additional scales. The statistical threshold was set at p<0.05.
Network analysis
Network estimation was carried out using the “estimateNetwork” function in the “bootnet” package [42] with the “mgm” [43] method, and visualized using “qgraph” [44] and “ggplot2” packages [45]. The Extend Bayesian Information Criterion (EBIC) was applied for model selection [46]. The hyperparameter gamma in the EBIC was set to 0.25 (the default value in mgm). Nodes represented individual items and edges represented correlations between items. The centrality of each node was evaluated using indices of expected influence (EI) and bridge expected influence (BEI), respectively [47]. EI refers to the sum of all connections of a given node, reflecting the relative importance of a node in the network [48]. BEI, on the other hand, indicates a node’s total connectivity with other communities (scales) [28]. Unlike the strength centrality and bridge centrality, EI and BEI do not take the absolute value of edges before summing them, which is useful for networks that have both negative and positive edges [48].
The stability and accuracy of the network model were evaluated using the “bootnet” package. Specifically, a correlation stability coefficient (CS-C) was used to evaluate network stability. All bootstrapping processes were conducted 1,000 times as recommended [42].
Network comparisons
Finally, the overall connectivity and network structure of networks based on man versus woman samples were compared using Network Comparison Tests (NCT) via the R package “NetworkComparisonTest.” [49] Specifically, the global strength (S value) and the maximum difference in edge weights (M value) of the two networks were compared to describe differences in global and local characteristics. Global strength is defined as the weighted absolute sum of all edges in the network. Higher global strength indicates stronger connectivity among symptoms, which could reflect the vulnerability of the disorder.
RESULTS
Participant characteristics
The final sample consisted of 1,085 Chinese young adults (mean age, 21.76±2.81 years, range: 16-41; 481 women, 604 men), with the majority being college students from Southwest Medical University or individuals from the surrounding local area. Using a tentative cut-off of ≥5 in IGDS scores, 137 respondents (12.6%) were classified as high risk for IGD; the remaining (n=948) were classified as low risk. The high-risk group demonstrated significantly higher scores on SSI, PSS, GAD-7, PHQ-9, and Neuroticism (Table 1).
A subsample of 347 respondents (mean age, 21.18±1.08 years, range: 18-30; 215 women, 132 men) completed the Chinese SSI 4 weeks later.
Item analysis and reliability
The item-total correlation of item #5 was 0.27, which is below the traditional cutoff value of 0.3,50 indicating the need for it to be revised or eliminated. The remaining eleven items all showed acceptable item-total correlations (0.57 to 0.78).
When item #5 was retained, Cronbach’s α was 0.896. After its removal, the internal consistency (Cronbach’s α) of the 11-item version of the SSI was 0.90. The test-retest reliability of the 11-item version of the SSI was 0.75 (p<0.001).
Validation analysis
To investigate the construct validity of the Chinese SSI, the inventory was subjected to EFA. EFA without item #5 produced a unidimensional structure, which accounted for 50.27% of the variance. The factor loadings of the items were all above 0.50 (0.54-0.78) (Table 2). For this model, Bartlett’s test of sphericity showed p<0.001, and the KMO was 0.93, suggesting that factor analysis was appropriate. This unidimensional structure was subsequently confirmed through CFA, which yielded an excellent model fit based on established standards (Root Mean Square Error of Approximation=0.077, Comparative Fit Index=0.991, Incremental Fit Index=0.991, and Goodness of Fit Index=0.993) [51,52].
In contrast, when item #5 was retained, EFA produced a two-dimensional structure that explained 55.73% of the total variance (Bartlett’s test p<0.001, KMO=0.935), with item #5 alone constituting the second factor. Further EFA restricted to a unidimensional structure (Bartlett’s test p<0.001, KMO=0.935, variance explained 47.24%) showed that item #5 had a loading of 0.404 on the single factor, which is marginally above the common cutoff of 0.40 [53], while the other items had loadings >0.54.
Moreover, exploratory item response theory analysis (assuming a unidimensional structure) revealed that item #5 exhibited the lowest discrimination and item information, along with a negative difficulty parameter (b=-1.2). In contrast, all other items displayed positive difficulty parameters (b ≥0.369), further indicating the low specificity of item #5 (Supplementary Table 1). These findings indicate that item #5 measures a construct distinct from the rest of the scale, thereby compromising its theoretical coherence. Consequently, item #5 was excluded from the Chinese version of the SSI. The final 11-item version was used in all subsequent validation and network analyses. The final version of the Chinese SSI can be found in Supplementary Table 2.
Correlation analysis showed that the SSI scores were positively correlated with scores on GAD-7 (r=0.68, p<0.001), PHQ-9 (r=0.69, p<0.001), PSS (r=0.71, p<0.001), Neuroticism (r=0.59, p<0.001), and IGDS (r=0.42, p<0.001) (Table 3), suggesting good criterion validity of the Chinese SSI. The correspondence between the Chinese SSI and the original SSI is presented in Table 2, along with summarizations.
Network analysis results
As shown in Figure 1, the SSI items formed a network with 32 out of 55 possible edges, resulting in a sparsity of 58.2%. The three nodes with the highest EI in this network were items #8 (functional impairment), #6 (cognitive decline), and #3 (mental freeze).
The combined SSI-IGDS network contained 78 out of 190 possible edges, with a sparsity of 41.1%. As shown in Figure 2, SSI items #8 (functional impairment) and #9 (substance use) were the nodes with the strongest BEI associations toward the IGD symptom cluster. Conversely, IGDS items #1 (preoccupation) and #3 (withdrawal) exhibited the highest BEI values linking the SSI cluster.
The CS-C was 0.75 for EI in the SSI network, and 0.75 and 0.36 for EI and BEI in the SSI-IGDS network, indicating acceptable stability for both centrality measurements in both networks (Supplementary Figure 1) [42]. Likewise, for both networks, bootstrapped 95% confidence intervals for estimated edge weights with a narrow range suggested that the edge weights were reliable (Supplementary Figure 1). For the SSI network, the NCT analysis revealed no significant gender differences in global strength (S=0.10; p=0.832) or maximum edge weight differences (M=0.16; p=0.950).
DISCUSSION
This study aimed to evaluate the psychometric properties of the Chinese version of the SSI and to examine the relationship between the SSI and symptoms of IGD using psychological network analysis. The main findings are as follows: 1) Item #5 was removed due to low coherence with other items. 2) The Chinese SSI has a unidimensional structure, demonstrating strong reliability and validity. 3) There is a positive correlation between SSI scores and IGDS scores (r=0.42, p<0.001). 4) In the SSI network, the highest EI was found in item #8 (functional impairment), item #6 (cognitive decline), and item #3 (mental freeze). 5) In the SSI-IGDS network, item #8 (functional impairment) and item #9 (substance use) exhibited the strongest BEI associations with IGD symptoms.
This study concluded that item #5 (“I have tried to find something positive that can come out of it.”) is not a valid item within the Chinese version of the SSI. This decision is based on the following evidence: 1) Item #5 showed a relatively low item-total correlation, which was below the commonly used 0.3 exclusion criterion and clearly lower than that of the other 11 items. Furthermore, removing item #5 slightly increased the scale’s overall internal consistency (Cronbach’s α), indicating it contributes little to the measurement. 2) EFA suggested that item #5 likely does not measure the same latent trait as the other 11 items. Although stress sensitivity could potentially be a multidimensional structure, it is psychometrically unsound for a single item to define a distinct factor [54]. 3) Item #5 showed much lower discrimination and item information, and a negative difficulty parameter, compared to the other 11 items, indicating it lacks specificity. These quantitative findings are underpinned by a fundamental conceptual discrepancy. Item #5 measures an active, adaptive coping strategy (positive reappraisal), while the other items predominantly measure negative stress experiences and responses. While items #8 (social withdrawal), #9 (functional impairment), and #10 (substance use) also partly involve coping, they describe passive or avoidant reactions, not proactive efforts. Theoretically, stress sensitivity reflects psychological and physical reactivity to stressors [4], which is distinct from coping. For instance, the transactional stress appraisal model [2] proposes reactivity to stress as a two-step process: a primary cognitive appraisal of the extent to which a situation is potentially harmful or threatening (a stressor), and a secondary appraisal to evaluate accessible coping resources to determine if the stressor can be controlled. Stress coping is then triggered when an individual perceives a stressor that taxes or exceeds one’s resources. Based on the above considerations, we decided to remove item #5 from the Chinese SSI.
After item #5 was removed, the Chinese version of the SSI has 11 valid items. The item-total correlations of the Chinese SSI ranged from 0.57 to 0.78, suggesting sound item-total consistency for the scale. The internal consistency (Cronbach’s α=0.90) and the test-retest reliability (r=0.75) of the scale were good. The EFA and CFA confirmed a unidimensional structure of the scale. Together, the Chinese SSI showed good reliability and construct validity.
Moreover, our findings revealed a strong association between the SSI scores and PSS scores, with a large effect size (r=0.71). Although as a measure of state-level stress perception, PSS scores have been repeatedly associated with measures of stable psychological traits, such as neuroticism, trait anxiety, and locus of control [55]. For example, a survey showed that perceived stress in adolescents was determined by high neuroticism, low self-efficacy and self-esteem, among others [56]. The strong correlation between SSI and PSS scores suggests that the SSI effectively captures individual differences in stress perception. Similarly, this study also revealed significant correlations between the SSI scores and severity of depression (PHQ-9) and anxiety (GAD-7) symptoms. These results were consistent with prior findings showing that higher stress reactivity is associated with increased negative emotional experiences and heightened emotional symptoms [5-7,57]. Importantly, this study revealed a positive correlation between the SSI scores and Neuroticism with a large effect size (r=0.59). This association aligns with established empirical and theoretical literature on the relationship between personality and stress vulnerability. According to the integrative models of stress process [58], personality is posited to influence an individual’s predisposition to both exposure to stressors and psychological/physiological reactivity to stress. In line with our results, a meta-analysis showed that Neuroticism was positively correlated with the Sensory Processing Sensitivity (a trait that captures sensitivity to internal and external stimuli [59]) in both children (r=0.42) and adults (r=0.40) [60]. The close association of the SSI with Neuroticism further supported the concept that it captures a relatively stable trait that is related to reactivity to stress.
In the SSI network, the three nodes with the highest EI were functional impairment (#8), cognitive decline (#6), and mental freeze (#3). These results suggest that functional impairment is a critical aspect of trait stress sensitivity, consistent with clinical models that emphasize functional deficits as central outcomes of heightened stress reactivity [1]. In support of this, longitudinal studies of 9/11 survivors suggested that greater reactivity to daily stressors and sleep disturbances were associated with poorer interpersonal and social functioning [61], underscoring how high stress sensitivity may disrupt real-world adaptive capacities in individuals. The strong influence of cognitive decline indicates that cognitive manifestations of stress—such as impairments in attention and memory—are essential in understanding stress sensitivity. Multiple lines of evidence support this concept. For instance, attentional control deficits in early adolescence have been shown to heighten stress sensitivity and increase vulnerability to depressive symptoms [62]. Preclinical research further underscores the neurobiological underpinnings of this relationship; in a mouse model of affective disorders, increased stress reactivity was associated with cognitive deficits and decreased hippocampal brain-derived neurotrophic factor [63]. Additionally, studies of children with autonomic dysregulation—characterized by parasympathetic withdrawal and sympathetic activation in response to cognitive and emotional challenges—reveal a pattern of mutual connections among heightened vulnerability to stress, functional impairment, and physical injury, suggesting a dynamic interplay between physiological stress reactivity, cognitive decline, and functional outcomes [64]. Finally, mental freeze is frequently reported in high-stress situations, often occurring when a circumstance is perceived as threatening and no effective coping strategy appears viable (such as fight or flight). This reaction can be understood as an extreme manifestation of cognitive decline [65]. Taken together, our network analysis suggests that functional impairment, cognitive decline, and mental freeze constitute central, interconnected elements of trait stress sensitivity. These findings point toward an integrated profile of stress vulnerability that spans functional, cognitive, and behavioral domains.
In this study, the SSI scores were closely correlated with IGD symptoms (r=0.42). This finding is in line with previous findings of close link between stress (and stress reactivity) and drug craving and relapse susceptibility [66], as well as the development of addiction [67], suggesting an association between trait stress sensitivity and IGD symptoms. Furthermore, network analysis identified specific components of stress sensitivity that may contribute to this association. In the SSI-IGDS network, the strongest edges emerged between SSI items #8 (functional impairment) and #9 (substance use) and IGD symptoms. This highlights the potential vulnerability to addictive gaming behaviors among individuals with high stress sensitivity, particularly when they also report elevated functional impairment or substance use. Functional impairment refers to difficulties in daily functioning—such as in work or academic life—which can result from either stress [68] or IGD [69]. The majority of individuals with IGD experience impaired functioning, particularly in their health, career, and social life [70]. On the other hand, individuals with high functional impairment may struggle to balance gaming with other responsibilities, potentially exacerbating both stress and IGD symptoms. Notably, although functional impairment is also a diagnostic criterion for IGD, this link cannot be accounted for by mere tautology: if it were, IGDS item #6 (“functional problems due to gaming”) would have appeared as a bridging symptom. Instead, the key bridging symptoms on the IGDS side were #3 (withdrawal) and #1 (preoccupation).
Additionally, substance use (item #9) also demonstrated a strong connection with IGD symptoms. Systematic reviews indicate that IGD frequently cooccurs with other addictive behaviors, such as alcohol use disorder [71]. Meta-analysis further suggests that behavioral addictions like IGD and substance use disorders share overlapping neural alterations, such as hyperconnectivity in the putamen [72] and between the frontoparietal network and the default mode, affective, and salience networks [73]. These common neurofunctional substrates may underlie the link between substance use and IGD symptoms.
Clinically, clarifying these associations can inform treatment for individuals presenting with elevated stress sensitivity and IGD. For example, addressing functional impairment could involve time-management training and value-based activity scheduling to help patients rebuild daily structure and reconnect with meaningful offline goals [74]. Concurrently, for those exhibiting substance use, integrated interventions that target shared mechanisms of addiction—such as craving regulation, impulse control, and alternative stress-coping strategies—may be particularly beneficial [72]. By targeting such bridge symptoms, interventions may disrupt the reinforcing loop between stress sensitivity and IGD, thereby reducing gaming-related symptoms and improving overall psychosocial functioning.
The present study has several limitations. First, the CS-C for BEI in the SSI-IGDS network was reported as 0.361, which is considered acceptable but falls below the ideal threshold of >0.50.42 This suggests that the identified order of node importance could vary across different samples. Consequently, the findings regarding bridge symptoms should be validated in future studies with more diverse populations. Second, this study adopted a cross-sectional design, which only captures the characteristics of a population at a specific point in time and cannot establish the temporal sequence between variables. Future longitudinal studies could offer valuable insights into how changes in stress sensitivity influence IGD symptoms over time, and vice versa. Third, this study did not examine potential differences in stress sensitivity across various demographic groups, which limits the generalizability of our findings. Future research incorporating participants from diverse backgrounds, such as different age groups, residential areas, and other relevant demographic factors, is needed to expand upon our findings. Finally, all constructs were measured using self-report instruments. Although the measures used are well-validated, this methodological approach may introduce common method variance, potentially inflating the observed relationships between variables.
Conclusions
In summary, this study demonstrated robust psychometric properties of the Chinese version of the SSI, making it a suitable instrument for assessing stress sensitivity in both clinical and research settings. The analysis revealed significant interconnections between stress sensitivity and IGD symptoms, with functional impairment (item #8) emerging as a particularly influential node both within the SSI cluster and in linking SSI to IGD. These findings advance our understanding of the bidirectional relationship between trait stress sensitivity and gaming-related behaviors. Future research should further investigate these connections to inform integrated interventions targeting both stress vulnerability and problematic gaming patterns.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation
Supplement1
Print
