Anxiety-depression symptom networks across emotion regulation profiles in Chinese nurses: a latent profile and network analysis Authors: Ping-Zhen Lin, Xu Wang, Yong-Sen Lin, Yan-Yan Lin, Lan-Lan Chen, Bi-Yu Wu Published: 2026 ABSTRACT BackgroundDepression and anxiety are highly prevalent among clinical nurses. This study employed a person-centered analytical approach to identify distinct emotion regulation (ER) profiles among Chinese clinical nurses and examined the structure of depressive and anxiety symptoms within each profile using latent profile analysis (LPA) complemented by network analysis (NA).MethodsA cross-sectional survey was conducted among clinical nurses in China. A total of 1,436 eligible nurses (mean age = 34.16, SD = 7.49; 93.5% female) participated. Participants completed the 10-item Emotion Regulation Questionnaire (ERQ), the 9-item Patient Health Questionnaire (PHQ-9), and the 7-item Generalized Anxiety Disorder scale (GAD-7). LPA was conducted using the two ERQ subscales, cognitive reappraisal and expressive suppression, to identify distinct ER profiles. NA was then applied to model the relationships among depressive and anxiety symptoms for the total sample and within each ER profile. Centrality (Expected Influence) and bridge centrality (Bridge Expected Influence) indices were calculated to identify core and bridging symptoms. Network accuracy and stability were assessed via bootstrap methods, and network structures were compared across profiles using the Network Comparison Test (NCT).ResultsLPA identified five distinct ER profiles: Low Regulation (7.2%), Below-Average Regulation (44.9%), Flexible Regulation (8.6%), Above-Average Regulation (32.0%), and High Regulation (7.2%). The Flexible Regulation group, characterized by high cognitive reappraisal and low expressive suppression, exhibited the lowest levels of depression and anxiety, whereas the Below-Average and Above-Average groups reported the highest symptom severity. NA revealed “Uncontrollable worry” (GAD6) and “Restlessness” (GAD5) as core and bridge symptoms across multiple profiles, while “Sad mood” (PHQ2), “Motor disturbance” (PHQ8), and “Suicidal ideation” (PHQ9) emerged as profile-specific central and bridge symptoms in the Low and Flexible Regulation groups. NCT indicated significant differences in global network structure across several profiles, yet global strength did not differ.ConclusionsThis study reveals heterogeneous ER patterns among clinical nurses and demonstrates that the Flexible Regulation group is the most adaptive configuration. Key central and bridge symptoms differ across profiles, offering empirically grounded targets for profile-informed interventions. Regular mental health screening with attention to profile-specific symptom patterns is recommended to support tailored prevention strategies and improve overall well-being among clinical nurses.Clinical trial numberNot applicable.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s12912-026-04903-6. Introduction Nurses are particularly susceptible to mental health disorders due to heavy workloads, night shifts, and work-family conflicts [1]. Anxiety and depression are prevalent psychological issues among clinical nurses [2–4] adversely affecting not only the nurses’ own health [5] but also patient safety and the quality of care provided [6, 7]. A comprehensive meta-analysis encompassing 401 studies and 458,754 participants from 58 countries established the pooled prevalence of anxiety and depression among nurses at 31.5% and 28.0%, respectively [8]. Similarly, a study [9] conducted in Hong Kong, China, reported that 35.8% and 37.3% of nurses suffered from depression and anxiety, respectively. During the COVID-19 pandemic, nurses endured intensified pressures and negative emotions [10, 11]. In a recent survey covering clinical nurses from 30 provinces in China, 55.5% reported symptoms of depression and 41.8% reported symptoms of anxiety [12]. Emotion regulation strategies are crucial and commonly employed by clinical nurses [13] to manage stress and mitigate negative emotions in both their professional and personal lives [14]. Emotion regulation (ER) [15–17] is defined as “the processes by which individuals influence which emotions they have, when they have them, and how they experience and express these emotions”. It also pertains to an individual’s ability to manage their emotional experiences and expressions. ER are broadly categorized into two types: cognitive reappraisal, an adaptive approach involving the reinterpretation of emotion-eliciting situations, and expressive suppression, a maladaptive strategy focused on inhibiting emotional responses [17]. Research indicates that maladaptive ER strategies exacerbate psychological and physical burdens while diminishing well-being, leading to heightened anxiety, depression, and increased risk of suicidal behaviors [18]. Conversely, adaptive ER strategies effectively mitigate negative emotions and reduce self-harm risk, conferring protective effects on physical health. However, subsequent studies suggest that the benefits of adaptive strategies in alleviating symptoms of anxiety and depression may be more pronounced when individuals concurrently employ high levels of maladaptive strategies [19]. This suggests adaptive strategies may exert less significant effects when maladaptive strategies are minimally employed [20]. Importantly, this interaction pattern is not universally applicable; individuals with severe symptoms [20], for instance, may derive limited benefit. Nevertheless, this inconsistency does not negate the established phenomenon of strategy interaction. Therefore, a comprehensive understanding of ER’s psychosocial outcomes cannot be achieved through regression analyses examining single strategies in isolation, nor through survey methods treating strategies as independent processes. The person-centered approach addresses this limitation by modeling combinations of variables operating as integrated systems within individuals. To capture the complex, individualized nature of ER strategy deployment [21], this methodology identifies distinct subpopulations characterized by unique configurations of adaptive and maladaptive strategies. Latent Profile Analysis (LPA) [22] is a model-based person-centered method that identifies homogeneous latent profiles within a sample, each defined by a specific pattern of indicator variables. Crucially, unlike traditional non‑latent clustering methods (e.g., cluster analysis), LPA offers greater flexibility through statistical fit indices, enabling evidence-based selection of the optimal subgroup solution [23]. Previous research has employed diverse methodologies to characterize ER profiles. Eftekhari et al. [24], applying cluster analysis to Emotion Regulation Questionnaire (ERQ) data, identified four patterns: High Regulators, High Reappraisers/Low Suppressors, Moderate Reappraisers/Low Suppressors, and Low Regulators. Individuals using fewer ER strategies reported significantly higher anxiety and depression levels compared to High Regulators or High Reappraisers/Low Suppressors, demonstrating an association between ER profiles and psychopathology. Similarly, Zeng et al. [25] identified three profiles among nursing undergraduates: Low Suppression/Moderate Reappraisal (6.6%); Moderate Suppression/High Reappraisal (89.1%), and High Suppression/High Reappraisal (4.3%). However, studies utilizing LPA to identify ER profiles within clinical nurse populations remain scarce. To address this gap, the first exploratory study aims to identify latent ER profiles among Chinese clinical nurses based on the two ERQ subscales, cognitive reappraisal and expressive suppression, and to preliminarily examine differences in anxiety and depression levels across these distinct profiles. Nevertheless, comparing group differences based on total scale scores fails to capture the complex interactions between individual symptoms [26]. The onset and persistence of psychiatric disorders are increasingly understood as being driven by reciprocal causal interactions among symptoms [27]. Network Analysis (NA) provides a powerful framework for modeling these symptoms as interconnected systems. It represents symptoms as nodes and their statistical relationships as edges, mapping intricate associations onto a structural plane while quantifying variable importance through centrality indices (e.g., strength, closeness, betweenness) [28]. This approach offers greater granularity than traditional methods [29] by identifying core symptoms, revealing key pathways, and uncovering symptom-activation cycles that sustain psychological distress [30]. Therefore, the second part of our study applies NA to identify highly central symptoms, detect strong symptom-symptom connections, and identify potential intervention targets. Furthermore, the Network Comparison Test (NCT) enables quantitative assessment of differences in global connectivity strength (i.e., overall density) within anxiety-depression networks across distinct ER profiles. Stronger network connectivity suggests greater potential for mutual symptom activation, potentially leading to more stable and treatment-resistant psychopathological states. We hypothesize that clinical nurses with different ER profiles will exhibit profile-specific variations in their anxiety-depression networks, specifically in: (a) the identity of core symptoms (high-centrality nodes), (b) the nature of bridge symptoms connecting anxiety and depression, and (c) the overall network structure and global strength. In summary, the objectives of this study are fourfold: (1) to delineate distinct ER profiles among clinical nurses using LPA; (2) to characterize the demographic and psychological differences across these profiles; (3) to employ NA to map the anxiety-depression symptom network within the overall sample and each ER profile, identifying core and bridge symptoms; and (4) to use NCT to compare the network structures and global strength across profiles. By elucidating how anxiety-depression networks manifest within specific ER configurations, this study aims to provide a foundation for designing highly individualized, symptom-targeted interventions. Based on the literature reviewed, we formulated the following hypotheses: Hypothesis 1 Distinct ER profiles will emerge among clinical nurses based on the two subscales of the ERQ (cognitive reappraisal and expressive suppression). Given that these two strategies are theoretically and functionally distinct [17] and do not always co-occur at congruent levels, we expect to identify profiles characterized by various configurations of the two dimensions, including non-congruent patterns such as high cognitive reappraisal combined with low expressive suppression. Hypothesis 2 The identified ER profiles will differ significantly in levels of depression and anxiety symptoms. Drawing on the process model of ER [31], we hypothesize that profiles characterized by maladaptive strategy configurations (e.g., high expressive suppression relative to cognitive reappraisal) will exhibit higher symptom burden, whereas profiles marked by adaptive configurations (e.g., high cognitive reappraisal with low expressive suppression) will demonstrate lower levels of anxiety and depression. Hypothesis 3 The network structures of anxiety-depression symptoms will differ across ER profiles. Specifically, we hypothesize that: Core symptoms will vary across profiles, as ER capacity may influence which symptoms play a central role in maintaining the symptom network.Bridge symptoms connecting anxiety and depression will also differ across profiles, reflecting potential differences in how anxiety and depression become interconnected depending on individuals’ regulation patterns.The global network structure will differ across profiles, suggesting that ER capacity shapes not only symptom severity but also the organization of symptom interactions. Core symptoms will vary across profiles, as ER capacity may influence which symptoms play a central role in maintaining the symptom network. Core symptoms will vary across profiles, as ER capacity may influence which symptoms play a central role in maintaining the symptom network. Bridge symptoms connecting anxiety and depression will also differ across profiles, reflecting potential differences in how anxiety and depression become interconnected depending on individuals’ regulation patterns. Bridge symptoms connecting anxiety and depression will also differ across profiles, reflecting potential differences in how anxiety and depression become interconnected depending on individuals’ regulation patterns. The global network structure will differ across profiles, suggesting that ER capacity shapes not only symptom severity but also the organization of symptom interactions. The global network structure will differ across profiles, suggesting that ER capacity shapes not only symptom severity but also the organization of symptom interactions. Methods Participants and procedure This cross-sectional study was conducted from February to March 2025. A convenience sample of full-time clinical nurses was recruited from hospitals in Quanzhou City, Fujian Province, China. The research team contacted head nurses or nurse leaders from participating hospitals and invited them to distribute the questionnaire link to the WeChat groups specifically established for nursing staff, ranging in size from 20 to 50 nurses per group. Participation was voluntary and uncompensated. The survey was administered online via the WeChat application. The process began with an informed consent page detailing the study’s purpose, procedures, and anonymity assurance. Participants indicated their voluntary agreement by selecting “I agree” before proceeding to the questionnaire. All questionnaires were administered through the Wenjuanxing online platform, which required completion of all items before submission, ensuring no missing data in the final dataset. To ensure data quality, two attention-check items were embedded in the questionnaire (e.g., “Please select ‘strongly agree’ for this item”). Participants who failed to correctly answer both attention-check items were excluded from the final analysis. A total of 63 participants were excluded based on this criterion. Participants were eligible if they: (1) held a valid registered nurse license; (2) had been employed for at least six months at their hospitals; (3) had no diagnosed dyslexia and were proficient in Mandarin communication; and (4) provided informed consent. Exclusion criteria included: (1) a diagnosed mental disorder; (2) current receipt of psychotherapy or antipsychotic medication; or (3) being a nursing student or visiting scholar. The sample size was planned to meet the requirements of the primary analytical methods. For LPA, a minimum of 300-500 participants was targeted, as recommended by previous research [32]. A total of 1,436 completed responses were obtained. All data were stored securely and were accessible only to the principal investigators to protect participant confidentiality. Measurement Demographic questionnaire Demographic characteristics included age, gender, marital status, education level, location, years of working, professional title. A self-designed demographic questionnaire was used to collect information on participants’ age, gender, marital status, education level, location, years of working experience, and professional title. Anxiety symptoms Anxiety symptoms were assessed using the Chinese version of the 7-item Generalized Anxiety Disorder scale (GAD-7) [33, 34]. Respondents rate the frequency of their symptoms over the past two weeks on a 4-point scale from 0 (not at all) to 3 (nearly every day) (e.g., “Feeling nervous, anxious, or on edge”). The total score ranges from 0 to 21, with higher scores indicating greater anxiety severity. A cutoff score of 10 or higher was used to identify individuals with a probable anxiety disorder [33]. In the present study, the scale demonstrated excellent internal consistency, with a Cronbach’s α of 0.942. Depressive symptoms Depressive symptoms were measured with the Chinese version of the 9-item Patient Health Questionnaire (PHQ-9) [35, 36]. The instrument assesses symptoms based on the DSM criteria, with each item scored from 0 (not at all) to 3 (nearly every day) (e.g., “Feeling down, depressed, or hopeless”). Total scores range from 0 to 27, and higher scores reflect more severe depression. A cutoff score of 10 or higher is was used to identify individuals with a probable depressive disorder [35]. In this study, the PHQ-9 showed excellent reliability, with a Cronbach’s α of 0.92. Emotion regulation strategies Emotion regulation strategies were assessed using the Chinese version of the Emotion Regulation Questionnaire (ERQ) [17, 37]. This 10-item scale comprises two distinct subscales: cognitive reappraisal (6 items) and expressive suppression (4 items). Items are rated on a 7-point Likert scale, ranging from 1 (strongly disagree) to 7 (strongly agree) (e.g., “I control my emotions by changing the way I think about the situation I’m in” for cognitive reappraisal; “I keep my emotions to myself” for expressive suppression). Subscale scores are calculated by summing the respective items, with higher scores indicating a greater tendency to use that specific strategy. In the present study, the internal consistency was acceptable for both cognitive reappraisal (α = 0.80) and expressive suppression (α = 0.74). Ethical consideration This study was approved by the Ethics Committee of Quanzhou First Hospital (Approval No.: 2024-K168) and was conducted in accordance with the ethical standards of the Declaration of Helsinki. Statistical analysis Prior to analysis, we confirmed that all questionnaires were completed in full, as the Wenjuanxing platform required responses to all items before submission; consequently, no missing data were present in the final dataset. Descriptive statistics were used to characterize the sample’s demographic and psychological profiles. Because GAD-7 and PHQ-9 total scores exhibited positive skewness across profiles (see Supplementary Table S1), Spearman’s rank correlations were used to describe bivariate associations, as Spearman’s ρ is more robust to non-normality and outliers [38]. All bivariate correlations were conducted in SPSS 26.0 with a two-tailed significance threshold of α = 0.05. Latent profile analysis LPA was conducted in Mplus 8.3 to identify subgroups based on the two subscale scores of ERQ. Following the recommendation of Spurk et al. [23] that LPA indicators should be theoretically driven and factorially distinct, we used the mean scores of the two well-established ERQ subscales, cognitive reappraisal and expressive suppression [17], as continuous indicators. Prior to the analysis, both subscale scores were standardized (z‑scores) to ensure equal weighting and to facilitate interpretation relative to the sample mean. We estimated models specifying 1 to 8 latent profiles, assuming local independence. Model fit was evaluated using the following criteria: lower values on the Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), and sample-size-adjusted BIC (aBIC) indicated better fit. The statistical significance of the Lo-Mendell-Rubin adjusted likelihood ratio test (LMRT) and the bootstrapped likelihood ratio test (BLRT) was used to determine if the k-profile model provided a superior fit to the k-1 profile model. Classification accuracy was assessed with entropy, where values closer to 1 indicate clearer delineation of profiles [39]. In addition to these statistical fit indices, model selection was also guided by theoretical interpretability and clinical meaningfulness of the identified profiles. Once the optimal model was identified, the Bolck-Croon-Hagenaars (BCH) approach was used to examine differences in demographic and clinical variables across profiles [40], accounting for classification uncertainty. Network estimation and analysis Symptom network analysis was conducted using R software (version 4.4.2). Gaussian Graphical Models (GGM) were employed to estimate the networks for the overall sample and within each ER profile [41]. The Extended Bayesian Information Criterion (EBIC) [42] combined with the Least Absolute Shrinkage and Selection Operator (LASSO) [43–45] regularization was applied to shrink spurious edges to zero, ensuring sparse and interpretable networks [46]. Nodes in the network represented items from the GAD-7 and PHQ-9 scales, which were predefined into two communities: the anxiety community and the depression community. Edges indicated the associations between variables. Red edges represent the negative associations, and blue edges represent the positive associations. Stronger relationships were indicated by wider edges [41, 46]. Since the network contained negative correlations, Expected Influence (EI) index, the sum of all edges extending from a given node, was adopted, as it outperforms the Strength index in such contexts [47]. Bridge Expected Influence (BEI), the sum of edge weights connecting a node to all nodes in the other community, captures cross-community interactions. Both EI and BEI were computed to identify central and bridge symptoms within the depression-anxiety network [48]. Nodes with higher BEI values were considered bridge symptoms [49], connecting the anxiety and depression communities. Accuracy and stability were evaluated using the bootnet package. Edge weight accuracy was assessed with 2,000 nonparametric bootstrap samples, and centrality stability was examined via case-dropping subset bootstrapping [46]. The Correlation Stability Coefficient (CS-C) greater than 0.25 [46] was deemed acceptable. Significance tests for differences in edge weights and in node EI or BEI values were performed using 2,000 bootstrap samples (α = 0.05) [46]. Finally, NCT was used to statistically compare global network structure and global strength (the absolute sum of all edge weights in a network) across the different ER profiles. Results Participant characteristics The final sample comprised 1,436 participants with a mean age of 34.16 years (SD = 7.49, range: 21-55). The sample was predominantly female (93.5%), married or cohabiting (68.3%), and residing in urban areas (68.5%). Half of the participants (50.2%) had an educational level of junior college or below. In terms of professional experience, 60.1% had worked for more than ten years, and 45.4% held a professional title above the intermediate level. Based on the clinical cutoffs for probable disorder, 7.24% of participants screened positive for probable anxiety disorder (GAD-7 ≥ 10) and 13.86% for probable depressive disorder (PHQ-9 ≥ 10). A comprehensive summary of the sample’s demographic characteristics is provided in Table 1. Table 1Demographic and job characteristics of the sample (N = 1,436)Itemsn(%)/ M(SD)Itemsn(%)/ M(SD)Ages, years34.16 ± 7.49 Years of working 21–25219(15.25)0.5-6350(24.37)26–35687(47.84)6–10223(15.53)36–45396(27.58)10–33863(60.10)46–55134(9.33) Professional title Gender registered nurse286(19.92)Male94(6.55)primary470(32.73)Female1342(93.45)intermediate561(39.06) Marital Status senior/deputy senior90(6.27)Other455(31.69)others29(2.02)Married/live with a partner981(68.31) Location Education level rural206(14.35)Junior college or below721(50.21)suburban247(17.20)Bachelor degree or above715(49.79)urban983(68.45)Note: M: mean; SD: Standard Deviation Table 1Demographic and job characteristics of the sample (N = 1,436)Itemsn(%)/ M(SD)Itemsn(%)/ M(SD)Ages, years34.16 ± 7.49 Years of working 21–25219(15.25)0.5-6350(24.37)26–35687(47.84)6–10223(15.53)36–45396(27.58)10–33863(60.10)46–55134(9.33) Professional title Gender registered nurse286(19.92)Male94(6.55)primary470(32.73)Female1342(93.45)intermediate561(39.06) Marital Status senior/deputy senior90(6.27)Other455(31.69)others29(2.02)Married/live with a partner981(68.31) Location Education level rural206(14.35)Junior college or below721(50.21)suburban247(17.20)Bachelor degree or above715(49.79)urban983(68.45)Note: M: mean; SD: Standard Deviation Demographic and job characteristics of the sample (N = 1,436) Note: M: mean; SD: Standard Deviation Correlational analysis of study variables The descriptive statistics and correlation matrix for anxiety, depression, cognitive reappraisal, and expressive suppression are presented in Table 2. Table 2Descriptive statistics and spearman correlations among study variables (N = 1,436)ItemsM (SD)12341. PHQ-94.03 (4.19)12. GAD-75.32 (5.18)0.84**13. CR28.36 (8.67)-0.21**-0.18**14. ES17.27 (5.41)-0.05*-0.02-0.72**1Note: Spearman correlation coefficients are reported. M: mean; SD: Standard Deviation; PHQ-9: 9-items Patients Health Questionnaire; GAD-7: 7-item Generalized Anxiety Disorder scale; CR: cognitive reappraisal; ES: expressive suppression. **p<0.01,*p<0.05 Table 2Descriptive statistics and spearman correlations among study variables (N = 1,436)ItemsM (SD)12341. PHQ-94.03 (4.19)12. GAD-75.32 (5.18)0.84**13. CR28.36 (8.67)-0.21**-0.18**14. ES17.27 (5.41)-0.05*-0.02-0.72**1Note: Spearman correlation coefficients are reported. M: mean; SD: Standard Deviation; PHQ-9: 9-items Patients Health Questionnaire; GAD-7: 7-item Generalized Anxiety Disorder scale; CR: cognitive reappraisal; ES: expressive suppression. **p<0.01,*p<0.05 Descriptive statistics and spearman correlations among study variables (N = 1,436) Note: Spearman correlation coefficients are reported. M: mean; SD: Standard Deviation; PHQ-9: 9-items Patients Health Questionnaire; GAD-7: 7-item Generalized Anxiety Disorder scale; CR: cognitive reappraisal; ES: expressive suppression. **p<0.01,*p<0.05 Identification of emotion regulation profiles via latent profile analysis Fit indices for models with 1 through 8 classes are presented in Table 3. The LMRT remained significant up to the 7-class solution, and the BIC continued to decline. However, the 8-class solution yielded a non-significant LMR (p = 0.128) and a smallest class comprising only 1.8% of the sample (n = 26), which falls below the recommended 5% minimum class size [50]. Consequently, solutions beyond seven classes were deemed inadmissible. Among the statistically tenable models (2–7 classes), we retained the 5-class solution for the following reasons. First, the 5-class model uniquely revealed a theoretically critical profile characterized by low expressive suppression and high cognitive reappraisal (Class 3, 8.6% of the sample), which directly aligns with the conceptual distinction between the two ER strategies outlined in our introduction. Second, entropy increased from 0.841 in the 4-class solution to 0.876 in the 5-class solution, indicating improved classification certainty. Third, each profile in the 5-class model contained at least 104 participants, ensuring adequate statistical power for subsequent profile-specific network analyses. Although the 6- and 7-class models exhibited slightly lower BIC values, the additional profiles represented largely quantitative subdivisions of existing profiles (e.g., splitting the “Average Regulators” group) without introducing qualitatively distinct configurations. Thus, the 5-class model was selected as the final solution, offering an optimal balance between statistical fit, theoretical interpretability, and parsimony. Table 3Latent profile analysis of emotion regulation with model fit results (N = 1,436)LPA ModelsAICBICaBICEntropyLMRT(p)BLRT(p)Number of each subgroup1-class8156.3828177.4618164.7541--14362-class7358.4267395.3137373.0760.951< 0.001< 0.001111/13253-class6481.4916534.1876502.4210.885< 0.001< 0.001106/933/3974-class6231.5226300.0276258.7300.841< 0.001< 0.001103/247/479/6075-class6057.5666141.8806091.0530.876< 0.001< 0.001645/104/124/104/4596-class5793.3615893.4835833.1270.903< 0.001< 0.00176/101/596/110/133/4207-class5487.8075603.7385533.8520.938< 0.01< 0.001107/484/243/103/76/117/3068-class5286.4395418.1795338.7630.9540.1283< 0.00176/26/83/477/104/304/243/123Note: AIC: Akaike information criterion; BIC: Bayesian information criterion; aBIC: adjusted Bayesian information criterion; LMRT: Lo-Mendell-Rubin adjusted likelihood ratio test; BLRT: Bootstrapped likelihood ratio test Table 3Latent profile analysis of emotion regulation with model fit results (N = 1,436)LPA ModelsAICBICaBICEntropyLMRT(p)BLRT(p)Number of each subgroup1-class8156.3828177.4618164.7541--14362-class7358.4267395.3137373.0760.951< 0.001< 0.001111/13253-class6481.4916534.1876502.4210.885< 0.001< 0.001106/933/3974-class6231.5226300.0276258.7300.841< 0.001< 0.001103/247/479/6075-class6057.5666141.8806091.0530.876< 0.001< 0.001645/104/124/104/4596-class5793.3615893.4835833.1270.903< 0.001< 0.00176/101/596/110/133/4207-class5487.8075603.7385533.8520.938< 0.01< 0.001107/484/243/103/76/117/3068-class5286.4395418.1795338.7630.9540.1283< 0.00176/26/83/477/104/304/243/123Note: AIC: Akaike information criterion; BIC: Bayesian information criterion; aBIC: adjusted Bayesian information criterion; LMRT: Lo-Mendell-Rubin adjusted likelihood ratio test; BLRT: Bootstrapped likelihood ratio test Latent profile analysis of emotion regulation with model fit results (N = 1,436) Note: AIC: Akaike information criterion; BIC: Bayesian information criterion; aBIC: adjusted Bayesian information criterion; LMRT: Lo-Mendell-Rubin adjusted likelihood ratio test; BLRT: Bootstrapped likelihood ratio test Figure 1 displays the 5-class solution. Clinical nurses were categorized into the following profiles: Class 1 (n = 104, 7.2%; “Low Regulation” group) exhibited markedly low scores on both strategies, indicating the least frequent use of ER. Class 2 (n = 645, 44.9%; “Below-Average Regulation” group) was characterized by slightly below-average scores on both cognitive reappraisal and expressive suppression. Class 3 (n = 124, 8.6%; “Flexible Regulation” group) was defined by above-average cognitive reappraisal combined with below-average expressive suppression. Class 4 (n = 459, 32.0%; “Above-Average Regulation” group) was characterized by moderately above-average scores on both strategies. Class 5 (n = 104, 7.2%; “High Regulation” group) displayed the highest scores on both strategies, reflecting intensive use of ER. Fig. 1Emotion regulation profiles identified by latent profile analysis (N = 1,436). Note. The x-axis represents the two subscales of the Emotion Regulation Questionnaire; the y-axis represents the z-score for each subscales across the three identified profiles. ES: expressive suppression; CR: cognitive reappraisal Emotion regulation profiles identified by latent profile analysis (N = 1,436). Note. The x-axis represents the two subscales of the Emotion Regulation Questionnaire; the y-axis represents the z-score for each subscales across the three identified profiles. ES: expressive suppression; CR: cognitive reappraisal Comparisons of demographics and clinical symptoms across profiles After deciding on a 5-class model according to the LPA results, the differences between the profiles in depression, anxiety, and basic characteristics were tested. As seen in Table 4, there were significant differences among the three latent profiles in depression, anxiety, and demographics characteristics. The Class 2 (Below-Average Regulation) and Class 4 (Above-Average Regulation) demonstrated significantly higher scores on PHQ-9 and GAD-7 relative to the other profiles (For PHQ-9: χ2 = 100.509, for GAD-7: χ2 = 110.989, ps < 0.001). Table 4Comparison of basic and psychological characteristics among various profiles (N = 1,436)ItemsClass 1(n = 104)Class 2(n = 645)Class 3(n = 124)Class 4(n = 459)Class 5(n = 104)Test statistic (χ2)Pairwise comparisons (BCH)Ages, years M(SE)33.96 (0.79)32.85 (0.29)36.10 (0.84)35.59 (0.41)34.37 (0.79)36.594***C2 < C4***=C3***Gender, n(%)7.281C1 < C3*Male12(11.54)45(7.00)4(3.13)26(5.66)7(6.73)Female92(88.46)600(93.00)120(96.77)433(94.34)97(93.27)Marital Status, n(%)15.867**C2 < C4**, C2 < C3**Other35(33.65)235(36.34)29(23.39)125(27.23)31(29.81)Married/live with a partner69(66.35)410(63.66)95(76.61)334(72.77)73(70.19)Education level, n(%)12.733*C2 < C3*, C2 < C4*, C1 < C3*, C5 < C3*Junior college or below57(54.81)344(53.33)50(40.32)212(46.19)58(55.77)Bachelor degree or above47(45.19)301(46.67)74(59.68)247(53.81)46(44.23)Location, n(%)19.987**C2 > C3***, C2 > C4**, C5 > C3*, C1 > C3*Rural13(12.50)105(16.28)10(8.07)57(12.42)21(20.19)Suburban22(21.15)135(20.93)16(12.90)66(14.38)8(7.69)Urban69(66.35)405(62.79)98(79.03)336(73.20)75(72.12)Years of working, n(%)20.420***C2 < C3**, C2 < C4***, C1 < C3*0.5-631(29.81)183(28.37)19(15.32)93(20.26)24(23.08)6–1012(11.54)108(16.74)20(16.13)67(14.60)16(15.38)10–3361(58.65)354(54.88)85(68.55)299(65.14)64(61.54)Professional title, n(%)26.032***C1 < C3*,C2 < C3***, C2 < C4**, C5 < C3*Registered nurse27(25.96)149(23.10)15(12.10)74(16.12)21(20.19)Primary35(33.65)226(35.04)35(28.23)139(30.28)35(33.65)Intermediate30(28.85)233(36.12)59(47.58)198(43.14)59(56.73)Senior/deputy senior8(7.69)23(53.57)14(11.29)40(8.71)14(13.46)Others4(3.85)14(2.17)1(0.81)8(1.74)1(0.96)GAD-7, M(SE)1.96 (0.33)5.00 (0.17)1.98 (0.35)4.08 (0.23)2.45(0.48)110.989***C1 = C3=C5 < C4**< C2***PHQ-9, M(SE)3.22 (0.48)6.46 (0.20)2.66 (0.44)5.54 (0.29)2.94 (0.56)100.509***C3 = C5=C1 < C4***< C2*ES, M(SE)5.22 (0.20)15.42 (0.09)12.53 (0.43)21.25 (0.13)27.54 (0.13)11369.523***C1 < C3< C2 < C4< C5***CR, M(SE)7.82 (0.31)23.72 (0.13)37.11 (0.40)34.03 (0.19)41.70 (0.15)14745.174***C1 < C2< C4 < C3< C5***Note: - Not applicable; Class 1: “Low Regulation” group; Class 2: “Below-Average Regulation” group; Class 3: “Flexible Regulation” group; Class 4: “Above-Average Regulation” group; Class 5: “High Regulation” group. M: mean; SE: standard error; PHQ-9: 9-items Patients Health Questionnaire; GAD-7: 7-item Generalized Anxiety Disorder; ES: expressive suppression; CR: cognitive reappraisal; BCH: Bolck-Croon-Hagenaars method; C: class. ***p<0.001, **p<0.01,*p<0.05 Table 4Comparison of basic and psychological characteristics among various profiles (N = 1,436)ItemsClass 1(n = 104)Class 2(n = 645)Class 3(n = 124)Class 4(n = 459)Class 5(n = 104)Test statistic (χ2)Pairwise comparisons (BCH)Ages, years M(SE)33.96 (0.79)32.85 (0.29)36.10 (0.84)35.59 (0.41)34.37 (0.79)36.594***C2 < C4***=C3***Gender, n(%)7.281C1 < C3*Male12(11.54)45(7.00)4(3.13)26(5.66)7(6.73)Female92(88.46)600(93.00)120(96.77)433(94.34)97(93.27)Marital Status, n(%)15.867**C2 < C4**, C2 < C3**Other35(33.65)235(36.34)29(23.39)125(27.23)31(29.81)Married/live with a partner69(66.35)410(63.66)95(76.61)334(72.77)73(70.19)Education level, n(%)12.733*C2 < C3*, C2 < C4*, C1 < C3*, C5 < C3*Junior college or below57(54.81)344(53.33)50(40.32)212(46.19)58(55.77)Bachelor degree or above47(45.19)301(46.67)74(59.68)247(53.81)46(44.23)Location, n(%)19.987**C2 > C3***, C2 > C4**, C5 > C3*, C1 > C3*Rural13(12.50)105(16.28)10(8.07)57(12.42)21(20.19)Suburban22(21.15)135(20.93)16(12.90)66(14.38)8(7.69)Urban69(66.35)405(62.79)98(79.03)336(73.20)75(72.12)Years of working, n(%)20.420***C2 < C3**, C2 < C4***, C1 < C3*0.5-631(29.81)183(28.37)19(15.32)93(20.26)24(23.08)6–1012(11.54)108(16.74)20(16.13)67(14.60)16(15.38)10–3361(58.65)354(54.88)85(68.55)299(65.14)64(61.54)Professional title, n(%)26.032***C1 < C3*,C2 < C3***, C2 < C4**, C5 < C3*Registered nurse27(25.96)149(23.10)15(12.10)74(16.12)21(20.19)Primary35(33.65)226(35.04)35(28.23)139(30.28)35(33.65)Intermediate30(28.85)233(36.12)59(47.58)198(43.14)59(56.73)Senior/deputy senior8(7.69)23(53.57)14(11.29)40(8.71)14(13.46)Others4(3.85)14(2.17)1(0.81)8(1.74)1(0.96)GAD-7, M(SE)1.96 (0.33)5.00 (0.17)1.98 (0.35)4.08 (0.23)2.45(0.48)110.989***C1 = C3=C5 < C4**< C2***PHQ-9, M(SE)3.22 (0.48)6.46 (0.20)2.66 (0.44)5.54 (0.29)2.94 (0.56)100.509***C3 = C5=C1 < C4***< C2*ES, M(SE)5.22 (0.20)15.42 (0.09)12.53 (0.43)21.25 (0.13)27.54 (0.13)11369.523***C1 < C3< C2 < C4< C5***CR, M(SE)7.82 (0.31)23.72 (0.13)37.11 (0.40)34.03 (0.19)41.70 (0.15)14745.174***C1 < C2< C4 < C3< C5***Note: - Not applicable; Class 1: “Low Regulation” group; Class 2: “Below-Average Regulation” group; Class 3: “Flexible Regulation” group; Class 4: “Above-Average Regulation” group; Class 5: “High Regulation” group. M: mean; SE: standard error; PHQ-9: 9-items Patients Health Questionnaire; GAD-7: 7-item Generalized Anxiety Disorder; ES: expressive suppression; CR: cognitive reappraisal; BCH: Bolck-Croon-Hagenaars method; C: class. ***p<0.001, **p<0.01,*p<0.05 Comparison of basic and psychological characteristics among various profiles (N = 1,436) Note: - Not applicable; Class 1: “Low Regulation” group; Class 2: “Below-Average Regulation” group; Class 3: “Flexible Regulation” group; Class 4: “Above-Average Regulation” group; Class 5: “High Regulation” group. M: mean; SE: standard error; PHQ-9: 9-items Patients Health Questionnaire; GAD-7: 7-item Generalized Anxiety Disorder; ES: expressive suppression; CR: cognitive reappraisal; BCH: Bolck-Croon-Hagenaars method; C: class. ***p<0.001, **p<0.01,*p<0.05 Network structure of depression and anxiety symptoms Network estimation and visualization The estimated network structures for the full sample and for each of the five ER profiles are depicted in Fig. 2. The network for the full sample was the most densely connected, with 73% of possible edges (88/120) present. At the profile level, the Class 1 (“Low Regulation” group) network was the sparsest, with a density of 45% (54/120). The Class 5 (“High Regulation” group) network was the second sparsest, with a density of 51.7% (62/120). The Class 3 (“Flexible Regulation” group) network exhibited an intermediate density of 57.5% (69/120). The Class 2 (“Below-Average Regulation” group) and Class 4 (“Above-Average Regulation” group) networks showed the highest profile-level densities, both at 61.7% (74/120). Despite these differences in density, the average absolute edge weight remained consistent at 0.06 across all networks. Fig. 2Networks of anxiety and depression in the total sample and across emotion regulation profiles. Note. Network structures are shown for (A) Class 1 (n = 104; “Low Regulation” group), (B) Class 2 (n = 645; “Below-Average Regulation” group), (C) Class 3 (n = 124; “Flexible Regulation” group), (D) Class 4 (n = 459; “Above-Average Regulation” group), (E) Class 5 (n = 104; “High Regulation” group), (F) the total sample (N = 1,436). Blue edges represent positive associations; red edges represent negative associations. Thicker edges indicate stronger associations Networks of anxiety and depression in the total sample and across emotion regulation profiles. Note. Network structures are shown for (A) Class 1 (n = 104; “Low Regulation” group), (B) Class 2 (n = 645; “Below-Average Regulation” group), (C) Class 3 (n = 124; “Flexible Regulation” group), (D) Class 4 (n = 459; “Above-Average Regulation” group), (E) Class 5 (n = 104; “High Regulation” group), (F) the total sample (N = 1,436). Blue edges represent positive associations; red edges represent negative associations. Thicker edges indicate stronger associations Centrality and bridge centrality indices Figure 3 depicts the results of the centrality analysis, with the outcome of the EI centrality difference test provided in Supplementary Figure S1. In the full sample, GAD6 (Uncontrollable worry) exhibited the highest EI, followed by PHQ2 (Sad mood) and GAD7 (Inability to relax). Across the five ER profiles, the patterns of symptom centrality differed. For Low Regulation group, the highest EI values were observed for PHQ2 (Sad mood), GAD2 (Irritability), and PHQ8 (Motor). In Below-Average Regulation group, the most central symptoms were GAD6 (Uncontrollable worry), GAD7 (Inability to relax) and PHQ2 (Sad mood). For Flexible Regulation group, PHQ2 (Sad mood) and GAD4 (Excessive worry) had the highest EI, with GAD6 (Uncontrollable worry) also showing notably high influence. In Above-Average Regulation group, the most central symptoms were GAD6 (Uncontrollable worry), PHQ8 (Motor) and PHQ4 (Fatigue). For High Regulation group, the most central symptoms were PHQ5 (Appetite changes), GAD6 (Uncontrollable worry) and PHQ4 (Fatigue). EI centrality demonstrated acceptable stability across all networks, with CS-C values of 0.75 for the full sample and 0.28–0.67 across the five profiles, all exceeding the recommended threshold of 0.25 (Supplementary Figure S2). Stability was relatively lower for the Low Regulation (0.28), Flexible Regulation (0.36), and High Regulation (0.37) groups, though still within acceptable limits. BEI analysis identified key symptoms connecting the anxiety and depression communities. In the full sample, GAD5 (Restlessness) had the highest BEI, followed by GAD3 (Dread or foreboding) and PHQ7 (Concentration). Profile-specific bridge symptoms varied notably, with GAD5 (Restlessness) emerging as a primary bridge symptom in the Below-Average, Above-Average, and High Regulation groups, whereas PHQ8 (Motor) and PHQ9 (Suicidal ideation) were the leading bridge symptoms in the Low Regulation group, and PHQ2 (Sad mood) and GAD4 (Excessive worry) served this role in the Flexible Regulation group (see Fig. 3 for full results). Node stability analysis demonstrated low stability for BEI, with CS-C values of 0.13 (full sample), 0.05 (Low Regulation), 0 (Below-Average Regulation), 0.05 (Flexible Regulation), 0.13 (Above-Average Regulation), 0 (High Regulation), as detailed in Supplementary Figure S3. Fig. 3Expected influence and bridge expected influence across networks. Note. (A) Standardized expected influence (z-scores) for each node in the total sample and across profiles. Nodes with higher values exert stronger overall influence within their respective networks. (B) Standardized bridge expected influence (z-scores). Nodes with higher values are identified as bridge symptoms connecting depression and anxiety communities. All centrality indices are presented as standardized z-scores to facilitate cross-network comparison. PHQ1, Anhedonia; PHQ2, Sad mood; PHQ3, Sleep problems; PHQ4, Fatigue; PHQ5, Appetite changes; PHQ6, Guilt; PHQ7, Concentration problems; PHQ8, Motor disturbances; PHQ9, Suicidal ideation; GAD1, Nervousness; GAD2, Irritability; GAD3, Dread or foreboding; GAD4, Excessive worry; GAD5, Restlessness; GAD6, Uncontrollable worry; GAD7, Inability to relax Expected influence and bridge expected influence across networks. Note. (A) Standardized expected influence (z-scores) for each node in the total sample and across profiles. Nodes with higher values exert stronger overall influence within their respective networks. (B) Standardized bridge expected influence (z-scores). Nodes with higher values are identified as bridge symptoms connecting depression and anxiety communities. All centrality indices are presented as standardized z-scores to facilitate cross-network comparison. PHQ1, Anhedonia; PHQ2, Sad mood; PHQ3, Sleep problems; PHQ4, Fatigue; PHQ5, Appetite changes; PHQ6, Guilt; PHQ7, Concentration problems; PHQ8, Motor disturbances; PHQ9, Suicidal ideation; GAD1, Nervousness; GAD2, Irritability; GAD3, Dread or foreboding; GAD4, Excessive worry; GAD5, Restlessness; GAD6, Uncontrollable worry; GAD7, Inability to relax Edges weights The strongest edge weights varied across the different networks. In the full sample, the top five edges were Fatigue-Sleep problems (PHQ4-PHQ3; 0.346), Anhedonia-Fatigue (PHQ1-PHQ4; 0.345), Motor disturbances-Suicidal ideation (PHQ8-PHQ9; 0.343), Anhedonia-Sad mood (PHQ1-PHQ2; 0.330), and Uncontrollable worry-Inability to relax (GAD6-GAD7; 0.326). Within the profile-specific networks, the strongest connections differed. For Low Regulation group, they were Nervousness-Dread or foreboding (GAD1-GAD3; 0.538), Anhedonia-Sad mood (PHQ1-PHQ2; 0.520), Fatigue-Sleep problems (PHQ3-PHQ4; 0.471), Sad mood-Fatigue (PHQ2-PHQ4; 0.355), and Guilt-Concentration problems (PHQ6-PHQ7; 0.342). In Below-Average Regulation group, the strongest edges were Anhedonia-Fatigue (PHQ1-PHQ4; 0.378), Motor disturbances-Suicidal ideation (PHQ8-PHQ9; 0.359), Dread or foreboding-Uncontrollable worry (GAD3-GAD6; 0.333), Anhedonia-Sad mood (PHQ1-PHQ2; 0.332), and Nervousness-Irritability (GAD1-GAD2; 0.303). In Flexible Regulation group, the strongest edges were Motor disturbances-Suicidal ideation (PHQ8-PHQ9; 0.414), Anhedonia-Sad mood (PHQ1-PHQ2; 0.271), Guilt-Motor disturbances (PHQ6-PHQ8; 0.270), Concentration problems-Motor disturbances (PHQ7-PHQ8; 0.244), and Nervousness-Irritability (GAD1-GAD2; 0.223). In Above-Average Regulation group, the strongest edges were Uncontrollable worry-Inability to relax (GAD6-GAD7; 0.386), Anhedonia-Fatigue (PHQ1-PHQ4; 0.357), Fatigue-Sleep problems (PHQ3-PHQ4; 0.355), and Concentration problems-Motor disturbances (PHQ7-PHQ8; 0.291). For High Regulation group, the top five edges were Nervousness-Irritability (GAD1-GAD2; 0.485), Irritability-Excessive worry (GAD2-GAD4; 0.391), Dread or foreboding-Uncontrollable worry (GAD3-GAD6; 0.384), Fatigue-Sleep problems (PHQ3-PHQ4; 0.380), and Uncontrollable worry-Inability to relax (GAD6-GAD7; 0.376). The results of the edge weight difference tests are presented in Supplementary Figure S4. Edge weight estimation demonstrated high precision, with CS-C values of 0.75 (full sample), 0.44 (Low Regulation), 0.75 (Below-Average Regulation), 0.44 (Flexible Regulation), 0.75 (Above-Average Regulation), 0.37 (High Regulation), as detailed in Supplementary Figures S5-S6. Network comparison across profiles The NCT revealed significant differences in global network structure across several profile pairs (see Supplementary Table S2 for full statistics). Specifically, the Low Regulation group network differed significantly from the Below-Average Regulation (M = 0.538, p < 0.001), Flexible Regulation (M = 0.469, p = 0.030), and Above-Average Regulation (M = 0.409, p = 0.022) groups, but not from the High Regulation group (M = 0.422, p = 0.538). The Flexible Regulation group network also differed from the Below-Average Regulation (M = 0.368, p = 0.023) and Above-Average Regulation (M = 0.368, p = 0.019) groups. No other pairwise comparisons were significant, and global strength did not differ across profiles (all ps > 0.05). Discussion This study provides valuable insights into the complex relationship between depression and anxiety among Chinese clinical nurses by integrating LPA and NA to examine how this relationship varies across distinct ER profiles. Using the two ERQ subscales, cognitive reappraisal and expressive suppression, LPA identified five distinct ER profiles: a “Low Regulation” group (7.2%), a “Below-Average Regulation” group (44.9%), a “Flexible Regulation” group (8.6%), an “Above-Average Regulation” group (32.0%), and a “High Regulation” group (7.2%). NA revealed that GAD6 (Uncontrollable worry) served as a central symptom across the full sample as well as in the Below-Average, Above-Average, and High Regulation groups, whereas PHQ2 (Sad mood) emerged as the central symptom within the Low and Flexible Regulation groups. Notably, bridge symptoms connecting anxiety and depression exhibited profile-specific patterns: GAD5 (Restlessness) functioned as the primary bridge symptom in the full sample and in the Below-Average, Above-Average, and High Regulation groups, while PHQ8 (Motor disturbances) and PHQ9 (Suicidal ideation) served as the dominant bridge symptoms within the Low Regulation group, and PHQ2 (Sad mood) and GAD4 (Excessive worry) within the Flexible Regulation group. These findings enhance the understanding of heterogeneous ER patterns and the core structure of depression and anxiety symptoms among clinical nurses. They also provide hospital administrators with empirically supported targets for developing tailored mental health interventions specific to different ER profiles. Latent profiles of emotion regulation The Flexible Regulation group, characterized by high cognitive reappraisal and low expressive suppression, is theoretically important. Consistent with the process model of ER, this profile was associated with the lowest depression and anxiety scores. This aligns closely with the “High Reappraisers/Low Suppressors” group identified by Eftekhari et al. [24] and the “Low Suppression/Moderate Reappraisal” group described by Zeng et al. [25] in nursing undergraduates, both of which reported the lowest symptom levels. The replication of this adaptive profile in a clinical nurse sample not only validates its generalizability but also underscores that the combination of high reappraisal and low suppression, rather than the absolute level of either strategy alone, constitutes the most beneficial regulation configuration. In contrast, the Below-Average Regulation and Above-Average Regulation groups, which together accounted for over 75% of the sample, reported the highest symptom severity. Although these two profiles differed in the absolute frequency of strategy use, they shared a common feature: cognitive reappraisal and expressive suppression were used at congruent levels. Their elevated symptoms suggest that using the two strategies at similar levels, whether at the low or high end, may reflect a form of regulatory rigidity. This pattern parallels Brandão et al. [51], who identified a “suppressive regulation profile” (high suppression, moderate reappraisal) that was also associated with the greatest symptom burden. One explanation may be that undifferentiated strategy use reflects a process of “struggling yet failing” to achieve adaptive regulation. Although these individuals attempt to regulate their emotions, their efforts may be inflexible and poorly matched to situational demands, which may in itself aggravate emotional distress [52]. In the Below-Average group, insufficient reappraisal may fail to counter distress, whereas in the Above-Average group, the concomitant elevation of suppression, a strategy consistently linked to poorer psychological outcomes [17, 51], likely offsets any benefits of heightened reappraisal. Notably, the Low Regulation group, despite reporting the lowest absolute levels of both strategies, did not show elevated symptoms compared to the Flexible or High Regulation groups. While some prior studies have linked low overall strategy use to higher distress [24, 53], in high-stress contexts such as nursing, a minimalist regulation style may partly reflect lower emotional reactivity, greater reliance on automatic regulatory processes, or the use of unmeasured adaptive strategies (e.g., acceptance, problem-solving) that effectively buffer distress [54]. Alternatively, the very low endorsement of both reappraisal and suppression might represent a distinct, resilient phenotype in this population, though the small sample size (n = 104) warrants cautious interpretation. The High Regulation group, which scored extremely high on both strategies, also reported relatively low symptom levels, comparable to the Flexible and Low Regulation groups. This suggests that intensive use of cognitive reappraisal may counteract the detrimental effects of high expressive suppression, at least with respect to symptom severity. However, as the cross-sectional design prevents causal inference, it remains possible that these individuals engage in high levels of both strategies precisely because they experience greater emotional intensity, and their regulatory efforts, while subjectively high, may not translate into corresponding psychological benefits. Furthermore, the effectiveness of specific ER strategies may be shaped by cultural context; in certain settings, an overreliance on cognitive reappraisal at the expense of contextually appropriate expressive suppression could itself become a source of stress [55]. In summary, the five profiles reveal two distinct pathways to lower symptom burden: a flexible pattern marked by high reappraisal and low suppression, and a minimalist pattern with low use of both strategies. The majority of nurses, however, showed congruent strategy levels and correspondingly higher distress, suggesting that regulatory inflexibility, rather than absolute strategy frequency, is the stronger correlate of psychological symptoms in this population. Network structure of anxiety and depression Central symptoms of anxiety and depression Centrality analysis of the depression-anxiety network revealed both shared and distinct central symptoms across the three ER profiles. High EI symptoms play a key role in maintaining and activating the entire symptom network, making them potential targets for intervention [56]. As demonstrated by Fried et al. [57], central symptoms increase the likelihood of developing additional symptoms compared to peripheral ones, which may facilitate the transition of the entire network to a healthier state [58]. Symptoms with relatively consistent centrality. “Uncontrollable worry” (GAD6) emerged as a central symptom across multiple profiles, ranking among the top three in the full sample, Below-Average, Above-Average, and High Regulation groups. Similarly, “Sad mood” (PHQ2) ranked among the top three central symptoms in the full sample, Low Regulation, Below-Average, and Flexible Regulation groups. The broad prominence of these two symptoms across multiple profiles reinforces their transdiagnostic importance: “Uncontrollable worry” as a core feature of generalized anxiety [59, 60], and “Sad mood” as a cardinal symptom of depression recognized by both DSM-5 [61] and ICD-10 [62]. Research further indicates that sad mood may outperform other depressive symptoms, and in some cases even total depression scores, in predicting depression diagnoses [63]. However, the centrality of these symptoms was not universal. “Uncontrollable worry” was notably less dominant in the Low and Flexible Regulation groups, where “Sad mood” instead occupied the highest centrality. Conversely, “Sad mood” lost its central position in the Above-Average and High Regulation groups, where somatic symptoms (e.g., “fatigue”, “appetite changes”) took precedence. This divergence suggests that reliance on expressive suppression may shift the network’s focus away from core mood symptoms and toward somatic manifestations, consistent with the notion that chronic suppression diverts psychological distress into physical complaints [17]. In profiles where suppression is low, whether because individuals use few strategies overall (Low Regulation) or preferentially engage reappraisal (Flexible Regulation), “Sad mood” remains the primary network hub. Symptoms varying across profiles. “Irritability” (GAD2) exhibited high EI exclusively in the Low Regulation group, alongside “Sad mood” and “Motor disturbance” (PHQ8). Given this group’s minimal use of both reappraisal and suppression, the centrality of irritability may signal a fundamental ER deficit. Neuroimaging studies have linked irritability to dysfunction in fronto-striatal-amygdala circuitry [64, 65], supporting the interpretation that low regulators may benefit from interventions targeting basic emotion recognition and regulation skills. Other profile-specific patterns were noteworthy. In the Above-Average and High Regulation groups, somatic symptoms gained centrality: “Motor disturbance” (PHQ8) and “Fatigue” (PHQ4) were among the top symptoms in the Above-Average group, while “Appetite changes” (PHQ5) and “Fatigue” appeared in the High Regulation group. The emergence of these somatic symptoms as central in profiles with elevated suppression aligns with the notion that chronic expressive suppression may divert psychological distress into physical complaints [17]. Meanwhile, in the Flexible Regulation group, “Excessive worry” (GAD4) was among the leading central symptoms, alongside “Sad mood” and “Uncontrollable worry”, suggesting that even in an adaptive profile, worry-related processes remain active, albeit perhaps more circumscribed. Bridge symptoms between anxiety and depression Bridge centrality analysis revealed profile-specific patterns in how anxiety and depression symptoms interconnect, shedding light on potential pathways that may maintain comorbidity. In the full sample and in the Below-Average, Above-Average, and High Regulation groups, “Restlessness” (GAD-5) emerged as the primary bridge symptom, consistent with prior network studies of Chinese nurses [59] and across multiple populations [66–69]. This pattern suggests that restlessness represents a common transdiagnostic pathway linking anxiety and depression, particularly in profiles characterized by moderate-to-high use of both regulation strategies. Its transdiagnostic role is further supported by longitudinal evidence linking restlessness to relapse in major depressive disorder [70] and early improvement predicting remission [71]. The Low Regulation and Flexible Regulation groups, however, exhibited distinct bridge configurations. In the Low Regulation group, “Motor disturbance” (PHQ8), “Uncontrollable worry” (GAD6), and “Suicidal ideation” (PHQ9) served as the dominant bridges, suggesting that when ER capacity is minimal, behavioral and somatic pathways, including thoughts of self-harm, mediate the anxiety-depression link. In the Flexible Regulation group, “Sad mood” (PHQ2) and “Excessive worry” (GAD4) were the primary bridges, indicating that even in the most adaptive profile, emotional and cognitive symptoms remain interlinked, albeit at lower intensity. Together, these findings suggest that ER configurations shape not only which symptoms are central but also which pathways maintain comorbidity, offering preliminary targets for profile‑tailored interventions. Comparative network structure across profiles NCT revealed that the global network structure of the Low Regulation group differed significantly from the Below-Average, Flexible, and Above-Average Regulation groups, while the Flexible Regulation group also differed from the Below-Average and Above-Average Regulation groups. In contrast, the High Regulation group did not differ structurally from any other profile. Notably, global strength (overall connectivity) did not differ significantly across any profiles. This pattern is noteworthy: the elevated symptom burden observed in the Below-Average and Above-Average Regulation groups was not accompanied by greater global strength, suggesting that their clinical vulnerability may stem from distinct configurations of symptom relationships rather than simply higher overall connectivity. The structural similarity between the Low and High Regulation groups is also notable, despite differing in symptom severity, they may share underlying network features. This interpretation aligns with the concept of regulatory rigidity [52], where a mismatch between strategy use and situational demands, whether reflecting insufficient or excessive regulation, may contribute to maladaptive symptom networks. Crucially, these results extend existing network models of psychopathology by demonstrating that the configuration of cognitive reappraisal and expressive suppression is associated not only with symptom severity but also with the structural organization of the symptom network itself. The finding that network architecture differs across profiles, while global strength remains comparable, suggests that interventions may need to target profile-specific edges rather than solely focusing on highly central symptoms. However, given the cross‑sectional design, these interpretations remain exploratory and warrant replication in longitudinal studies. Implication for practice These findings reveal considerable heterogeneity in ER profiles among clinical nurses, highlighting the need for profile-informed, tiered intervention strategies. NA further identified distinct core and bridge symptoms across profiles, providing specific targets for prevention. For the Below-Average and Above-Average Regulation groups, which together represent over 75% of nurses and reported the highest symptom severity, interventions targeting “Uncontrollable worry” and “Restlessness” may offer broad benefits, with additional attention to somatic complaints in the Above-Average group. The Low Regulation group would likely benefit from foundational ER skills, while the Flexible Regulation group may need support in consolidating reappraisal and managing residual worry. These insights can assist hospital administrators in moving beyond uniform mental health programs toward tailored strategies that address the specific symptom patterns and regulation needs of different nurse subgroups. Limitations This study has several limitations. First, the cross-sectional design precludes causal inferences regarding the relationship between ER profiles and anxiety–depression networks. Although network analysis serves as a valuable exploratory tool to identify symptom-level targets, the directionality of the identified associations remains uncertain. Future longitudinal research should examine the temporal dynamics of these associations and whether changes in ER profiles precede changes in symptom network structure. Second, convenience sampling may limit sample representativeness. Participants were recruited from urban tertiary hospitals, which may not fully represent clinical nurses in primary care, community health centers, or non-tertiary institutions. Additionally, the proportion of male nurses was relatively low, consistent with the general nursing workforce but potentially limiting insights into gender-specific patterns. Future studies should adopt more rigorous sampling strategies to enhance demographic and institutional diversity. Third, while we focused on ER profiles’ relationships with depression and anxiety, other unexamined factors (e.g., coping styles, workplace environment, or personality traits) may contribute to symptom development. Investigating these factors would provide a more comprehensive understanding of how individual and contextual variables interact with ER configurations to shape psychological outcomes. Fourth, several profiles had relatively small sample sizes: the Low Regulation, Flexible Regulation, and High Regulation groups each comprised only 104 nurses (7.2%), and the Below-Average Regulation group had 124 nurses (8.6%). These limited sample sizes may affect the precision and stability of network estimates for these subgroups. Although stability analyses indicated acceptable centrality stability across all profiles (CS-C = 0.28–0.67), the lower values observed in the Low Regulation group (CS-C = 0.28) warrant cautious interpretation, and findings for these smaller profiles should be replicated in larger samples. Fifth, the GAD-7 and PHQ-9 items exhibited positive skewness within some profiles, which may affect the assumption of multivariate normality underlying Gaussian Graphical Models. Although we employed regularization and non-parametric bootstrapping to mitigate this concern, future research could employ non-parametric network estimation methods to further validate our findings. Conclusions This study identified five distinct ER profiles among Chinese clinical nurses. The Flexible Regulation group, high cognitive reappraisal with low expressive suppression, was the most adaptive, associated with the lowest depression and anxiety. In contrast, the Below-Average and Above-Average Regulation groups, comprising over 75% of the sample, reported the highest symptom severity, suggesting that undifferentiated strategy use is more strongly linked to distress than absolute strategy frequency. NA revealed that “Uncontrollable worry” and “Restlessness” served as transdiagnostic core and bridge symptoms, while “Sad mood”, “Motor disturbance”, and “Suicidal ideation” emerged as profile-specific features. Global network structure differed significantly across profiles, yet global strength did not, indicating that emotion regulation shapes how symptoms interconnect rather than overall connectivity levels. These findings highlight the heterogeneity of ER among clinical nurses and offer empirically grounded targets for profile-informed interventions. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1: Additional supporting information can be found online in the Supplementary Material section. Fig. S1. Node Expected Influence Difference Test. Fig. S2. Stability of Expected Influence Indices. Fig. S3. Stability of Bridge Expected Influence Indices. Fig. S4. Edge Weight Difference Test. Fig. S5. Edge Weight Accuracy. Fig. S6. Edge Stability via Case-Dropping Bootstrap. Table S1. Descriptive statistics and normality indices for GAD-7 and PHQ-9 total scores within each latent profile Supplementary Material 1: Additional supporting information can be found online in the Supplementary Material section. Fig. S1. Node Expected Influence Difference Test. Fig. S2. Stability of Expected Influence Indices. Fig. S3. Stability of Bridge Expected Influence Indices. Fig. S4. Edge Weight Difference Test. Fig. S5. Edge Weight Accuracy. Fig. S6. Edge Stability via Case-Dropping Bootstrap. Table S1. Descriptive statistics and normality indices for GAD-7 and PHQ-9 total scores within each latent profile