Associations between perceived stress profiles, social connection and work engagement in clinical registered nurses: a mediation analysis and generalized additive models

Abstract

Background

Investigating the links between individual perceived stress, social connection, and work involvement plays a crucial role in enhancing the psychological health and overall well-being of clinical nurses, as well as elevating the quality of nursing care in clinical settings. This study incorporates the concepts of social connection and work engagement. By considering the potential heterogeneity of variables and applying a mediation model, we identified the influence pathways through which subgroups of perceived stress affect social connection and work engagement. This study offers a valuable reference for understanding the nurses’ perceived stress profiles and improving their mental health.

Method

The research employed a cross-sectional study methodology. To select 600 clinical nurses from 3 hospitals in Guangzhou, a convenience sampling technique was implemented. Participants’ general demographics, levels of work engagement, perceived stress, and social connections were gathered through questionnaires. Statistical analyses were performed utilizing latent profile examination, mediation analysis and generalized additive models.

Results

(1) The analysis revealed heterogeneity in stress levels among nurses, resulting in the identification of three distinct groups: low stress-high self-demand group (23.4%), high tension-low out-of-control group (57.5%), and high stress-low efficiency group (18.2%). (2) Clinical registered nurses that obtained support from their families were more inclined to be placed in the Low stress-high self-demand group. (3) Social connection significantly mediated the relationship between nurses’ work engagement and perceived stress. (4) Work engagement demonstrated a non-linear relationship with both perceived stress and social connection.

Conclusion

The social connections and work engagement of clinical nurses were notably affected by the stress on an individual. Nursing leaders should promptly identify the stress patterns of nurses, implement appropriate stress management strategies, enhance group cohesion through social networks, and promote the nurses’ mental health and occupational well-being, which in turn can decrease nurse turnover rates and enhance the quality of clinical nursing care.

Clinical trial number

Not applicable.

Background

Clinical nurses form the core component of medical facilities and are the most extensive professional group within the healthcare system, making up roughly 59% of all healthcare professionals [1]. The dynamic yet challenging professional environment, coupled with high-intensity workloads and significant job risks, has led to a decline in clinical nurses’ willingness to remain in their positions [2]. As reported from State of the World’s Nursing 2025, the shortage of nursing staff is continued to worsen. In 2023, the global nursing workforce reached 29.8 million, reflecting an increase of approximately 1.9 million since 2018. Nevertheless, persistent challenges, including the uneven geographical distribution and density of nurses worldwide, as well as the substantial disparity between the growth rate of the nursing workforce and overall population growth, have resulted in an ongoing global deficit of 5.8 million nurses [1]. This critical situation threatens to undermine the core role of nurses in achieving universal health coverage and advancing global sustainable development goals. Retaining the current nursing workforce and ensuring the sustained engagement of newly recruited nurses in healthcare services will be a paramount challenge for the global healthcare system.

Owing to the unique characteristics of the clinical nursing profession, the presence of stress is a common phenomenon. During clinical practice, nurses are frequently exposed to multiple stressors, such as the care of critically ill patients [3], frequent night shifts [4], increased risks of occupational exposure and violence [5, 6], and the management of complex interpersonal relationships in clinical settings [7]. The high demands and intense nature of clinical work environments directly contribute to the escalation of occupational stress among nurses. As stress levels rise, nurses’ physical and psychological well-being becomes increasingly compromised, leading to a range of adverse outcomes, including job burnout [8], acute stress disorder [9], anxiety and depression [10], and even suicidal tendencies [11]. Furthermore, unexpected public health events have significantly intensified nurses’ stress, often prompting them to leave their positions as a form of self-protection. Existing research on nursing stress has predominantly focused on general trends, with limited attention given to individual differences in stress experiences [12]. Roy’s Adaptation Model [13] states that nurses use consciousness and self-selection in coping with stress to build a coherent integration between the person and the environment, ultimately respond adaptively or ineffectively, but individuals often exhibit diverse responses when confronted with stressors. This indicates that individuals exhibit varying stress responses when confronted with stressful events. Those who adopt positive coping strategies can stimulate their work vitality, achieving a positive transformation of stress, in contrast, individuals who respond negatively or exhibit high stress tolerance may experience excessive loss of control or heightened tension [14]. Therefore, this study aims to identify latent stress profiles among clinical nurses and uncover distinct patterns of perceived stress, that is, the heterogeneity of perceived stress.

Hypothesis 1

Clinical registered nurses will exhibit heterogeneity in their perceived stress.

High perceived stress is often associated with low levels of work engagement, whereas low perceived stress is more likely to lead to moderate or high work engagement [14, 15]. According to Fox’s stress-emotion model [16], when nurses face excessive workloads (e.g., a sudden surge in patient numbers), emotional demands (e.g., managing critically ill patients or conflicts with families), time constraints, or occupational risks, they become vulnerable to emotional exhaustion and depersonalization. Consequently, individuals may resort to mechanically completing tasks, which ultimately results in reduced work engagement. A strong work engagement experience is a key protective factor for nurses’ willingness to stay on the job, and an important predictor of work quality and satisfaction [17, 18]. Work engagement refers to a transient state of pleasure that an individual experiences while engaging in work activities, characterized by heightened concentration, enjoyment derived from work, and intrinsic motivation towards work [19, 20]. Existing studies indicate that the state of being devoted to work can positively influence those around you, and that elevated levels of work engagement enhance the collaborative spirit within professional teams [21]. This, in turn, boosts professional acknowledgment and fosters a feeling of connection to the nursing profession [22]. This highlights the critical importance of alleviating nurses’ occupational stress in promoting their work engagement and retention within the profession. More importantly, strengthening social connection in the team is critical, particularly fostering relationships among clinical colleagues and between nurses and leaders. A highly cohesive and interconnected medical team facilitates the exchange and flow of clinical information through the establishment of a trust-based network, structured information sharing mechanisms, and a shared sense of responsibility. This approach enhances collaborative efficiency, strengthens the team’s capacity to respond to medical incidents and unexpected crises, and ultimately contributes to improved care quality and patient safety [23, 24].

Social connection refers to the fully perceived self aspect of an individual’s closeness to interpersonal relationships in the world around them [25]. Nurses as individuals engage in group-based care activities in order to realize their professional values and to seek connection with others or groups. social connection as an important stress coping resource for individuals or groups has been shown to be an positive influence on nurses’ work engagement [26, 27]. Research indicates that nurses who have stronger social connections are more likely to exhibit increased enthusiasm and dedication in their work [28]. However, nurses with a lower level of social connection may experience negative emotions such as job aversion, which is ultimately detrimental to the development of clinical nursing work [29]. Prior research has also shown that social connection is impacted by individual perceived stress, and that high levels of perceived stress reduce clinical nurses’ sense of group closeness and belonging, which is not conducive to the extension of nurses’ social relations and the enhancement of care quality [30–32]. At the same time, an increased social connection promotes a feeling of belonging to the organization, further enhancing the contagious nature of nursing work engagement and contributing to the nursing quality and their willingness to stay in their jobs [33, 34]. Therefore, building upon the relationship among perceived stress, social connection, and work engagement in clinical nurses, this study further investigates the differential characteristics of social connection and work engagement across distinct stress patterns. By integrating the general mediation model, it conducts a more refined analysis of the relative mediation pathways through the latent profiles of perceived stress.

Hypothesis 2

Perceived stress will be negatively associated with feelings of social connection and work engagement.

Hypothesis 3

Perceived stress profiles will differ in their levels of social connection and work engagement.

Hypothesis 4

Social connection will act as a mediator in the association between perceived stress profiles and work engagement.

In addition, a non-linear dynamic relationship may exist among perceived stress, social connection, and work engagement in clinical nurses. The Job Demands-Resources (JD-R) model [35] posits that both job demands and resources simultaneously influence employees’ psychological states and levels of work engagement. The high-intensity workload and emotional exhaustion commonly experienced by clinical nurses in their daily practice deplete their physical and mental resources [4]. In contrast, collaboration within medical teams and peer support serve as critical resources that enhance nurses’ work motivation and facilitate the achievement of nursing objectives [28]. Variations in the levels or intensities of these three factors may lead to threshold effects or marginal changes. These non-linear patterns suggest the presence of dynamic interactions and offer a valuable entry point for investigating longitudinal trends in variable relationships. Therefore, this study will conclude by exploring the potential non-linear associations among perceived stress, social connection, and work engagement in clinical nurses from an exploratory research perspective.

Hypothesis 5

There will be a non-linear relationship between perceived stress, social connection, and work engagement.

In short, a crucial aspect of our study is to investigate the heterogeneity of nurses’ perceived stress levels. This aims to distinguish from the prior approach of assessing individual stress levels based solely on the total score of the self-assessment scale. Additionally, social connection may be a significant predictor to work engagement [36], and perceived stress may be significantly associated with social connection [37]. In other words, perceived stress, social connection, and work engagement can be integrated into a unified mediating model to examine the interrelationships among these variables [14, 26, 38]. By synthesizing the above methodological approaches, this study aims to develop a relative mediation model based on latent subgroups of perceived stress, thereby addressing the existing research gap. Moreover, we plan to utilize the generalized additive models to preliminarily examine the nonlinear change pattern as the variables increase, thereby offering a reference foundation for implementing progressive intervention strategies. Thus, this study seeks to analyze the heterogeneity in nurses’ perceived stress levels, explore how social connection mediates the link between stress profiles and work engagement, and reveal the complex nonlinear interactions among these factors. Figure 1 illustrates the conceptual model.

Fig. 1.

The conceptual model of perceived stress groups, social connection and work engagement in clinical nurses

Materials and methods

Participants

This study adopted a cross-sectional survey approach, utilizing a single-source design. A convenience sampling method was applied to recruit a total of 621 clinical nurses from three hospitals in Guangzhou between September 2024 and November 2024. After eliminating invalid responses, 21 participants were excluded, resulting in an effective response rate of 96.6%. The inclusion criteria were: (1) nurses holding a valid clinical practice certificate; (2) individuals capable of communicating fluently in Chinese; and (3) individuals who gave informed consent and took part in the study voluntarily. The exclusion criteria included: (1) nursing trainees; and (2) advanced practice nurses.

Sample size

The sample size required for this study was calculated using G*Power 3.1 software. The effect size value was determined based on Ferguson’s established criteria for social sciences [39], in conjunction with findings from prior relevant research [14]. Given an effect size (f) of 0.25, an alpha level of 0.05, and a desired statistical power (1-β) of 0.80, the analysis indicated that a total of 120 participants would be necessary to be needed. And, we relied on Yang’s finding that a minimum of 50 participants in each latent profile is crucial for obtaining reliable model fit information in the context of LPA or LCA [40]. Given that this research identified 3 distinct subgroups, it is preferable for the overall sample size to be at least 150. Additionally, a dropout rate of 20% needs to be considered. Taking these factors into account, a sample size of 600 participants was deemed appropriate.

Demographic information

General demographic information are outlined as follows: gender, age, whether or not the nurse is an only child, education level, labour relationship, nursing role models, family support.

Work engagement

The Work Related Flow Inventory (WOLF), which was developed by Bakker [41], encompasses three essential dimensions: concentration, enjoyment, and intrinsic motivation. This inventory comprises a total of 13 items aimed at gauging these dimensions effectively. Notably, the Chinese version of the WOLF has demonstrated a high level of reliability [42], indicating that it can be consistently used for research or practical applications in Chinese-speaking populations. In the current study, the reliability of the WOLF was quantified using Cronbach’s alpha, which yielded a robust value of 0.893, further reinforcing the instrument’s validity in measuring the flow experience related to work activities.

Perceived stress

Perceived Stress Scale (PSS), which was created by Cohen [43], assesses two primary dimensions related to psychological stress: the sense of loss of control and the feeling of tension. This scale comprises a total of 14 items that respondents evaluate using a 5-point Likert scale. Higher scores on this scale indicate that individuals experience greater levels of perceived psychological stress. Importantly, the Chinese version of the PSS has shown reliable psychometric properties [44]. In the context of this study, the internal consistency of the PSS was evaluated, yielding a Cronbach’s alpha of 0.865, suggesting a strong level of reliability for this assessment tool.

Social connection

Social Connectedness Scale (SCS), initially developed by Lee [45], has undergone a process of continuous refinement, leading to the creation of its revised version. This updated scale comprises a total of 20 items and utilizes a 6-point Likert scale, where responses range from 1 to 6. Higher scores on this scale are indicative of greater levels of social connection among individuals. The Chinese adaptation of the SCS-R has also proven to be reliable, as evidenced by its strong psychometric properties [36]. In the context of this study, the reliability was measured using Cronbach’s alpha, which yielded a value of 0.928, suggesting an excellent level of internal consistency for the scale.

Data analysis

First, general demographic and occupational profiles were described to identify potential factors of work engagement using univariate analysis. Bayesian correlation examination was employed to examine the relationship among variables. Harman’s univariate model was employed to assess the potential common method variance (CMV) issue [46].

Second, latent profile analysis was employed to detect hidden subgroups related to perceived stress. A stepwise profiling procedure was implemented using 1 to 5 categories of the model, and various indices were assessed to determine the best profile. These indices included Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), Lo-Mendell-Rubin (LMR), Bootstrapped Likelihood Ratio (BLRT), adjusted Bayesian Information Criterion (aBIC), and Entropy value. Reduced values for AIC, BIC, and aBIC indicate a more optimal fit for the model. Additionally, entropy values approaching 1 reflect greater classification accuracy, and notable p-values from the LMR and BLRT tests suggest that the model with k categories is preferable compared to the model with k-1 categories [47]. In addition, univariate logistic regression was used to identify potential indicators of LPA-based perceptual stress [48]. A one-way ANOVA was used to compare work engagement and social connection among clinical nurses with different perceived stress types.

Furthermore, the mediation role of social connection between LPA-based perceived stress (categorical variable) and work engagement was analyzed [49]. Finally, generalized additive models (GAM) were employed to detect potential non-linear relationships between the variables, and a threshold and saturation effect test was conducted to determine the optimal fold in the non-linear curve. 95% confidence intervals did not contain 0 was judged to be statistically significant [50].

Data analysis was performed using SPSS 26.0, JASP 0.17.1, Mplus 8.3, Empower Stats 4.1.

Results

Population features

Initially, 621 clinical nurses were selected to participate in this survey. However, 21 participants were subsequently excluded because of invalid responses. The average of the clinical nurses was 34.49 ± 7.72 years old. The survey results indicated that 82.5% of clinical nurses held a bachelor’s degree, while 71.0% were employed on contract. Significant differences in work engagement were observed across factors such as age and family support (all P < 0.001), as outlined in Table 1.

Table 1.

Demographic and professional characteristic differences in scores of work engagement

Characteristics	N(%)	work engagement (M ± SD)	P value
Gender			0.817
Male	19(3.2)	42.05 ± 10.39
Female	581(96.8)	42.50 ± 8.26
Age			< 0.001
20–30	216(36.0)	40.48 ± 8.50
31–40	262(43.7)	42.79 ± 8.08
41–50	105(17.5)	45.27 ± 7.70
Over 51 years old	17(2.8)	46.18 ± 7.18
Only Children			0.156
Yes	74(12.3)	41.20 ± 8.80
No	526(87.7)	42.67 ± 8.25
Education level			0.515
College degree	95(15.8)	41.65 ± 8.75
Bachelor degree	495(82.5)	42.62 ± 8.19
Master degree or above	10(1.7)	43.80 ± 10.76
Labor relationship with the hospital			0.374
Authorized strength	13(2.2)	41.23 ± 4.94
Contract employee	426(71.0)	42.24 ± 8.53
Service Dispatching(Third Party)	161(26.8)	43.24 ± 7.94
Is there a nursing role model you admire?			0.128
Yes	533(88.8)	42.64 ± 8.48
No	67(11.2)	41.24 ± 6.86
Whether your family supports your clinical nursing work?			< 0.001
Yes	521(86.8)	43.26 ± 8.18
No	79(13.2)	37.35 ± 7.39

Common method variance test

The results indicate that an initial unrotated factor analysis yielded eight factors with eigenvalues exceeding 1. The first factor accounted for 28.33% of the total variance, which is below the commonly accepted threshold of 40%. Therefore, it can be concluded that common method variance has not exerted a significant influence on the findings.

The perceived stress profiles

Potential subgroups 1–5 were examined based on fit indices, and a 3-class model was determined to be the best choice based on (1) comparatively small AIC, BIC, and aBIC, and (2) a sample size ratio exceeding 10% for each category, (3) relatively high Entropy values, and (4) significant P-values for LMR. Therefore, three profiles of perceived stress were identified and named low stress-high self-demand group (23.4%, C1), high tension-low out-of-control group (57.5%, C2), and high stress-low efficiency group (18.4%, C3). Refer to Table 2; Fig. 2. Therefore, the results of the LPA support Hypothesis 1.

Table 2.

Fitting index and group size of latent profile analysis models

Indices	Unconditonal model
	1-profile	2-profile	3-profile	4-profile	5-profile
Fit statistics
LL	-11813.397	-11027.887	-10601.113	-10375.998	-10198.508
AIC	23682.794	22141.775	21318.226	20897.996	20573.016
BIC	23805.908	22330.843	21573.248	21218.972	20959.946
aBIC	23717.016	22194.330	21389.114	20987.217	20680.570
Entropy	—	0.886	0.890	0.886	0.893
BLRT	—	0.0000	0.0000	0.0000	0.0000
LMR (P)	—	0.0844	0.0462	0.5067	0.6857
Group-sizes(%)
C1	600(100%)	191(31.8%)	146(24.3%)	126(21.0%)	41(6.8%)
C2	—	409(68.2%)	345(57.5%)	59(9.8%)	160(26.7%)
C3	—	—	109(18.2%)	283(47.2%)	168(28.0%)
C4	—	—	—	132(22.0%)	200(33.3%)
C5	—	—	—	—	31(5.2%)

Notes: Bold figures highlight the selected class solution. C1-C5: 1–5 categories

Abbreviations: LL (Log-likelihood); AIC (Akaibe Information Criterion); BIC (Bayesian Information Criterion); aBIC (Adjusted BIC); LMR (Lo-Mendell-Rubin); BLRT (the Bootstrapped Likelihood Ratio Test)

Fig. 2.

Probability of scoring on PSS for 3 latent profiles of nurses’ perceived stress. Note: C1 = Low stress-high self-demand group; C2 = High tension-low out of control group; C3 = High stress-low efficiency group

Bayesian correlation examination of work engagement, social connection and perceived stress

Perceived stress (40.20 ± 8.44) was significantly correlated with work engagement (42.49 ± 8.32) and social connection (88.38 ± 16.72). Perceived stress was associated both negatively with work engagement (r=-0.414, P < 0.001) and social connection (r=-0.530, P < 0.001), while social connection showed a positive association with work engagement (r = 0.362, P < 0.001). The Bayesian estimation method demonstrated that, as the sample size grew, the correlation between variables became statistically significant, and the sample exhibited excellent stability and representativeness. Therefore, the Bayesian correlation analysis verified Hypothesis 2. For more details, see Table 3; Fig. 3. In Table 4, logistic regression indicated that family support for nurses was a significant indicator for the perceived stress subgroup. In Table 5, the LPA-based perceived stress subgroups were significantly different in both work engagement and social connection (P < 0.001). The aforementioned results provide strong support for the validity of Hypotheses 3.

Table 3.

Pearson correlation among perceived stress, work engagement and social connection

Variables	Correlation matrix
	Mean	SD	1	2	3
1.Perceived stress	40.20	8.44	1
2.Work engagement	42.49	8.32	-0.414**	1
3.Social connection	88.38	16.72	-0.530**	0.362**	1

Note: ^**correlation is significant at the 0.01 level (2-tailed)

Abbreviations: SD, Standard Deviation

Fig. 3.

Bayesian correlation of nurses’ perceived stress with social connection and work engagement

Table 4.

Logistic regression for predicting external characteristics of three profiles of perceived stress

Variables	C2 vs. C1			C3 vs. C1
	OR (95% CI)	P		OR (95% CI)	P
Gender (female as ref)
Male	2.595 (0.573–11.741)	0.216		3.462 (0.659–18.189)	0.142
Age (over 51 years old as ref)
20–30	0.913 (0.277–3.004)	0.881		1.796 (0.313–10.288)	0.511
31–40	0.834 (0.256–2.715)	0.763		1.477 (0.260–8.397)	0.660
41–50	0.805 (0.236–2.750)	0.730		1.071 (0.175–6.544)	0.940
Only children (no as ref)
Yes	1.345 (0.738–2.453)	0.334		0.731 (0.310–1.723)	0.474
Education level (Master degree or above as ref)
College degree	5.446 (1.294–22.927)	0.021		7.143 (0.751–67.976)	0.087
Bachelor degree	2.756 (0.728–10.435)	0.136		3.465 (0.398–30.167)	0.260
Labor relationship (third party as ref)
Authorized strength	1.567 (0.318–7.717)	0.581		2.192 (0.342–14.047)	0.408
Contract employee	0.887 (0.572–1.376)	0.593		1.103 (0.619–1.964)	0.739
Nursing role model (no as ref)
Yes	0.951 (0.522–1.732)	0.869		1.664 (0.690–4.010)	0.257
Family support (no as ref)
Yes	0.430 (0.196–0.941)	0.035		0.153 (0.067–0.349)	<0.001

Note: C1 = Low stress-high self-demand group; C2 = High tension-low out of control group; C3 = High stress-low efficiency group

Table 5.

Comparison of work engagement and social connection scores under three categories

Categories	Work engagement	Social connection	Range of the PSS
Low stress-high self-demand group (C1)	47.20 ± 8.78	100.28 ± 14.12	14 ~ 40
High tension-low out of control group (C2)	41.78 ± 7.37	86.37 ± 14.80	34 ~ 55
High stress-low efficiency group (C3)	38.42 ± 7.70	78.79 ± 16.97	38 ~ 70
F	42.942	70.797	—
P	<0.001	<0.001	—
η2	0.126	0.192	—
Post-hoc	C3< C2< C1	C3< C2< C1	—

Notes: Post-hoc comparisons were conducted using the Bonferroni multiple comparison test; η² = eta squared represents variance of a dependent variable by three LPA-based categories

Abbreviations: PSS, Perceived Stress Scale

Social connection as a mediator between perceived stress (continuous and categories variable) and work engagement

General factors that showed statistical significance were incorporated as controlled variables in the mediation model. The results indicate that, after introducing social connection as a mediating variable, perceived stress negatively predicts both work engagement (β=-0.291, P < 0.001) and social connection itself (β=-0.510, P < 0.001), while social connection positively predicts work engagement (β = 0.162, P < 0.001). These findings suggest that social connection partially mediates the relationship between perceived stress and work engagement, accounting for 22.2% of the mediation effect. For detailed results, please refer to Table 6; Fig. 4. Low stress-high self-demand group as a reference, social connection mediated the effect between high stress-low out of control group and work engagement, with a mediating effect of 28.8%, the 95% Bootstrap CI for the three types of effect estimates do not include 0, indicating statistical significance. In addition, social connection mediated the association between the high stress-low efficiency group and work engagement, with a mediating effect of 28.5% and 95% Bootstrap CI for the three types of effect estimates do not include 0, indicating statistical significance. For further information, refer to Tables 7 and 8, and Fig. 5. Therefore, the results of the relative mediation support Hypotheses 4.

Table 6.

The mediating effect of social connection between perceived stress and work engagement

Variables	β	SE	t	P	LLCI	ULCI	R 2
	Mediating variable model (Outcome variable: social connection)						0.301
Constant	-0.432	0.218	-1.984	0.048	-0.859	-0.004
Perceived stress	-0.510	0.035	-14.511	< 0.001	-0.579	-0.441
	Dependent variable model (Outcome variable: work engagement)						0.236
Constant	-0.189	0.228	-0.830	0.407	-0.638	0.259
Perceived stress	-0.291	0.043	-6.797	< 0.001	-0.375	-0.207
Social connection	0.162	0.043	3.786	< 0.001	0.078	0.246

Note: Age and family support as control variables

Abbreviations: SE, standard error; LLCI, lower level of confidence interval; ULCI, upper level of a confidence interval

Fig. 4.

The mediation of social connection between nurses’ perceived stress and work engagement. Note: ^**P < 0.001; Age and family support as control variables, and the control variables are not presented in the figure for brevity

Table 7.

The mediating effect of social connection between perceived stress categories and work engagement

Variables	β	SE	t	P	LLCI	ULCI	R 2
	Mediating variable model (Outcome variable: social connection)						0.217
High tension-low out-of-control group	-0.801	0.088	-9.107	< 0.001	-0.974	-0.629
High stress-low efficiency group	-1.197	0.115	-10.395	< 0.001	-1.423	-0.971
	Dependent variable model (Outcome variable: work engagement)						0.216
High tension-low out-of-control group	-0.430	0.094	-4.566	< 0.001	-0.614	-0.245
High stress-low efficiency group	-0.651	0.125	-5.192	< 0.001	-0.897	-0.405
Social connection	0.218	0.041	5.298	< 0.001	0.137	0.298

Note: Age and family support as control variables; C1 (Low stress-high self-demand group) is used as the reference

Abbreviations: SE, standard error; LLCI, lower level of confidence interval; ULCI, upper level of a confidence interval

Table 8.

Direct and indirect effect of perceived stress profiles (categorical variables) on work engagement

Effects	Variables	Estimate	SE	t	LLCI	ULCI
Indirect effect	High tension-low out of control group	-0.174	0.038	4.087	-0.252	-0.105
	High stress-low efficiency group	-0.260	0.057	4.656	-0.379	-0.155
Direct effect	High tension-low out of control group	-0.430	0.094	-4.566	-0.614	-0.245
	High stress-low efficiency group	-0.651	0.125	-5.192	-0.897	-0.405
Total effect	High tension-low out of control group	-0.604	0.090	-6.701	-0.781	-0.427
	High stress-low efficiency group	-0.911	0.118	-7.728	-1.143	-0.679

Note: Age and family support as control variables; C1 (Low stress-high self-demand group) is used as the reference

Abbreviations: SE, standard error; LLCI, lower level of confidence interval; ULCI, upper level of a confidence interval

Fig. 5.

Diagram of a model of the relative mediation of social connection between nurses’ perceived stress categories and work engagement. Note: ^**P < 0.001; Age and family support as control variables, and the control variables are not presented in the figure for brevity; Low stress-high self-demand group is used as the reference

The non-linear relationship between perceived stress, social connection and work engagement

Considering the span of scores of the study variables, GAM was utilized to investigate the nonlinear relationship between the variables. As the scores of nurses’ perceived stress and social connection increased, the scores of work engagement gradually decreased or increased, but showed a non-linear relationship (Fig. 6A and B). The threshold effect and saturation effect examination revealed that the turning point between perceived stress and work engagement was at 27 (P < 0.001). This indicates that for each one-point increase in stress below the threshold, the work engagement score decreased by 0.787. In contrast, for each one-point increase in stress above the threshold, the work engagement score decreased by 0.318. Additionally, the threshold point between social connection and work engagement was found to be 101 (P < 0.01). This suggests that prior to the threshold, work engagement increased by 0.074 for each one-point increase in social connection. After the threshold, work engagement increased by 0.482 points for each one-point rise in social connection. The details are shown in Table 9. In addition, Fig. 6C shows that a linear relationship between perceived stress and sense of social connection. The results of the smooth curve fitting provide support for Hypothesis 5. It is worth noting that the observed linear relationship between perceived stress and social connection will be further discussed in the Discussion section.

Fig. 6.

The non-linear relationship between perceived stress, social connection and work engagement in clinical nurses

Table 9.

Threshold effect analysis of perceived stress and social connection on work engagement using piece-wise linear regression

	Crude			Adjusted
	β	95%CI	P	β	95%CI	P
Perceived stress
K < 27	-0.809	(-1.211, -0.408)	< 0.001	-0.787	(-1.178, -0.396)	< 0.001
K ≥ 27	-0.361	(-0.446, -0.275)	< 0.001	-0.318	(-0.403, -0.232)	< 0.001
Social connection
K < 101	0.126	(0.083, 0.169)	< 0.001	0.074	(0.023, 0.124)	0.004
K ≥ 101	1.017	(0.665, 1.370)	< 0.001	0.482	(0.337, 0.627)	< 0.001

Note: Crude: no adjustment; Adjusted: adjusted for age, family support; K: the optimal threshold point in the non-linear relationship

Abbreviations: CI, confidence interval

Discussion

First, Hypothesis 1 was confirmed. Our study demonstrated that the perceived stress among clinical registered nurses can be categorized into three groups, named low stress-high self-demand group (C1), high stress-low out-of-control group (C2), and high stress-low efficiency group (C3). This also provides empirical support for Roy’s Adaptation Model [13], which explains the various response patterns exhibited by individuals when confronted with stress-inducing stimuli. The finding diverges from the research results of Gu et al. [51] regarding the four types of perceived stress among nurses (low, mild, moderate, and high stress groups). This difference may be attributed to the varying levels of work intensity and environmental conditions that individuals have encountered. More crucially, it depends on the coping resources they have access to when facing pressure, as well as their personal aspirations for professional achievement [14]. This further underscores the importance for nursing managers to dynamically identify the stress profiles of clinical nurses according to distinct occupational environments and implement suitable stress management strategies. The number of people in the C2 and C3 groups together accounted for 75.7%, indicating that clinical registered nurses had an overall medium to high level of perceived stress. Nurses in the high-stress-low-efficiency group (18.2%) should be given more support. This group struggles to cope effectively with stressful events in clinical nursing and is vulnerable to workplace environments or professional interpersonal relationships, which may lead to burnout and reduced productivity [52]. At the same time, this research indicates that nurses who receive strong family support can manage work-related stress in a positive manner and exhibit enhanced self-regulatory efficacy, consistent with previous studies [53]. Family support enhances nurses’ self-esteem and self-confidence, enabling them to confront work-related challenges and stress more effectively. This support may also facilitate nurses’ ability to handle unfamiliar work situations, strengthen their intrinsic resilience to stress, and ultimately enhance both productivity and the quality of care provided [54, 55]. This indicates that nursing leaders can increase the interaction frequency between nurses and family members via multimedia interaction plans, enhance support for family communication, and adjust nurses’ perceived stress levels [56].

Secondly, Hypotheses 2 and 3 were validated. Perceived stress showed a negative relationship with both social connection and work engagement. The low stress-high self-demand group had the highest social connection and work engagement and the high stress-low efficiency group had the lowest. Adopting a scientific approach to stress management can reduce nurses’ burnout and negative emotions and enhance work engagement levels [57]. The rational emotive therapy can break through the inherent thinking affected by stress, improve irrational self-perceptions, reframe positive coping styles, and enhance the quality of life [58]. In addition, Positive Mood Cognitive Therapy (MBCT) and Positive Stress Reduction (MBSR) programmes can help individuals improve the negative effects of negative emotions and high stress states, and strengthen stress coping skills [59, 60]. Therefore, nursing managers should organize regular stress management skills training for clinical nurses, and at the same time, optimize flexible scheduling management and clinical supervision to enhance inter-group understanding and intimacy and improve work engagement levels.

Further, Hypothesis 4 of this study was tested. clinical nurses’ sense of social connection had a significant mediating role between perceived stress and work engagement. High stress-low efficiency group reflected a lower sense of social connection and work engagement. Elevated occupational perceived stress may affect intergroup work interactions, reduce clinical nurses’ group intimacy and sense of belonging, result in a rise in negative emotions, like personal burnout, and eventually impact work engagement [14, 61]. The two-way impact on individuals and groups requires more effective interventions by clinical care decision-makers or managers. According to Fox’s stress-emotion model [16], the different levels and types of stress have an impact on individuals’ perceptions and evaluations, individuals experience a range of emotional responses, and short-term or long-term emotions can influence individuals’ psychology and behaviour. Therefore, nursing managers should actively give psychological counseling to alleviate the burnout of such nurses. At the same time, in order to strengthen nurses’ self-efficacy and relieve occupational stress, managers can establish clearly defined phased goals tailored to the individual competency levels of nurses, thereby enhancing their sense of professional accomplishment [12, 62]. Assistance in maintaining individual social relationships, especially with family members, or initiating new relationships with others through the video conferencing will help the nurse to call on external resources to cope with occupational stress and re-establish a good work status [63]. Employee Assistance Programs (EAP) and the Balint work group format also helped to improve burnout and stress among nurses [64, 65].

It is worth noting that a portion of the content outlined in Hypothesis 5 was validated. Our research findings reveal a non-linear relationship between work engagement and perceived stress, as well as social connection. Specifically, this relationship exhibits either a negative gradual decrease or a positive gradual increase. This suggests that groups with strong social connections are more conducive to enhancing work engagement. Furthermore, perceived stress demonstrates a negative linear relationship with social connection. According to the Job Demands-Resources model [35], interpersonal connections and collaborative efforts within teams serve as a key resource for nurses in managing high workload stress and enhancing work immersion, thereby exerting a direct positive influence on stress alleviation and work immersion improvement. Consequently, nursing managers can identify nurses who possess high organizational competence and strong interpersonal intimacy, enabling them to collaboratively support individuals exhibiting stress-inefficiency tendencies in building robust psychological resilience. This ensures a positive state of work engagement while reinforcing organizational belongingness.

Limitations

First, our study focused on clinically registered nurses in China, whose characteristics might not align with those of nurses in Western countries. Consequently, the findings might not be relevant for nurses from different contexts or backgrounds. Second, due to the limitations of the investigation methods, we are unable to draw causal conclusions. Future investigations should explore longitudinal or clinical intervention approaches to confirm the relationships identified in this study. Third, an imbalance in gender distribution within our sample could have introduced potential selection bias.

Conclusion

Family support is a critical factor influencing subgroups of job engagement and perceived stress among clinical nurses. Perceived stress levels among registered clinical nurses can be classified into three profiles. Furthermore, the influence of perceived stress on work engagement is mediated by social connection. Notably, the relationship between perceived stress, work engagement, and social connection is not linear. Consequently, nursing managers should identify stress patterns among nurses in a timely manner, implement targeted stress management interventions, strengthen group cohesion through social networks, and foster the psychological health and work happiness of nurses, especially guarantee the crucial role of family support for nurses. These efforts can help reduce nurse turnover and enhance the quality of clinical nursing care.

Abbreviations

SD

Standard Deviation

Author contributions

Yuan Liao and Miaochun Huang: Writing - original draft, Data curation, Conceptualization, Methodology, Software; Zhimin Gu: Data curation, Resources, Validation; Chun Li and Yan Yu: Investigation, Resources, Validation; Qimei Zhang, Xiangyu Lai and Jialin Liu: Investigation, Resources; Kang He and Huiyun Chu: Investigation; Yao Zhao, Xinyu Wu and Lihua Wu: Resources; Yu Li and Sujuan Fang: Guidance, Review and Editing.

Funding

The work was funded by Guangdong Undergraduate Colleges and Universities Teaching Quality and Teaching Reform Construction Project (Grant numbers: 2024-30-271) and Guangdong Provincial Traditional Chinese Medicine Bureau Scientific Research Project (Grant numbers: 20252012).

Data availability

The corresponding author can provide the data that supports the conclusions of this study upon a reasonable request.

Declarations

Ethical approval

The research received approval from the Ethics Review Committee of the Second Affiliated Hospital of Guangzhou University of Chinese Medicine (No: YE2022-355-01). Before conducting the survey, all participants provided written informed consent, and the research was executed in accordance with the principles outlined in the Declaration of Helsinki.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.