Introduction

Suicide may be considered one of the most important international and multifactorial public health problems affecting the general population. The World Health Organization (WHO) estimates that nearly one million people die from suicide each year, with 20 times more attempting it. By 2020, the WHO projected approximately 1.53 million deaths from suicide globally, indicating a significant increase in mortality rates1. A significant increase in suicidality among children and adolescents has been observed, particularly following the onset of the COVID-19 pandemic2. Research indicates that suicide is the second leading cause of death among adolescents aged 15 to 19 years, with alarming statistics emerging from various regions. For instance, a study in Spain reported a dramatic rise in psychiatric emergency visits for suicidal ideation or attempts, with cases increasing from 10 in 2018 to 88 in 2021, predominantly affecting girls and middle adolescents3. Additionally, a systematic review highlighted the global trend of rising suicides among youth, emphasizing the need for targeted interventions to address this urgent public health crisis4. These findings underscore the critical need for comprehensive mental health support for this vulnerable population.

In this context, teachers serve as key agents in identifying suicidal thoughts among children and adolescents5. The educational setting provides an optimal environment for developing prevention programs that leverage the socio-emotional development of young people6. Teachers can act as “gatekeepers” in programs that train individuals in supportive roles, equipping them with strategies to respond effectively to at-risk youth. A study found that 58.8% of teachers reported encountering students who disclosed suicidal thoughts, yet many lacked adequate training, with a median of only 2 h of suicide prevention training received7. There is evidence about the importance of supporting teachers continued engagement in youth suicide prevention and prevention training that targets specific outcomes in teachers’ attitudes and efficacy8. However, the question arises: what provisions exist for addressing teachers’ own suicidal ideations? There is a certain irony in the fact that individuals potentially experiencing suicidal ideation themselves are tasked with mitigating the risk of such ideation among their students. This critical question highlights an often-overlooked aspect of suicide prevention in educational settings. While teachers are frequently positioned as frontline gatekeepers in identifying and supporting students with suicidal ideation, their own mental health and well-being are rarely addressed5. They are in a pivotal position to identify and support students with mental health issues, yet many lack adequate suicide literacy and training, which can hinder their ability to respond effectively8,9.

The acknowledgment of teachers struggling with suicidal ideation is indeed a delicate issue, often considered taboo within sociocultural contexts. Teachers are expected to identify and support students facing mental health challenges10, yet they may themselves experience significant stressors that contribute to suicidal tendencies, such as increased workload and interpersonal relationships11. This duality creates a complex dynamic where societal expectations may prevent open discussions about teachers’ mental health, leading to feelings of shame or inadequacy12. Furthermore, the stigma surrounding mental health issues can hinder teachers from seeking help, despite their critical role as gatekeepers in recognizing and addressing self-harm and suicidal behaviors among students. The need for culturally informed training and support for educators is essential, as it can empower them to address their own mental health while effectively responding to the needs of their students Thus, the interplay of sociocultural factors complicates the discourse surrounding teachers’ mental health and their responsibilities.

Addressing this gap becomes even more pressing when considering that teaching is one of the most stressful professions. Numerous studies highlighting the significant prevalence of stress among educators13. A recent scoping review14 found the prevalence of burnout in teachers ranged from 25.12 to 74%, stress ranged from 8.3 to 87.1%, anxiety ranged from 38 to 41.2% and depression ranged from 4 to 77%. Teachers often experience significant personal and professional stressors, such as increased workloads, lifestyle changes, and interpersonal relationships, which can lead to maladaptation and suicidal tendencies11. The emotional burden of assisting students in distress15,16,17, combined with the high-stress nature of the teaching profession, can lead to burnout, compassion fatigue, and, in some cases, suicidal thoughts. This complex dynamic creates a challenging scenario in which those responsible for safeguarding students’ mental health may be silently struggling with their own psychological distress. Addressing this issue necessitates a holistic approach to school-based suicide prevention that not only equips teachers with the tools to support their students but also provides robust mental health resources and support systems for educators themselves18.

Given the significant mental health challenges teachers face, it becomes crucial to examine the risk and protective factors related to suicidal ideation. The diathesis-stress model, which posits that suicidal behavior arises from the interaction between external stressors and internal vulnerabilities19, provides a relevant framework for this exploration. While our study partially aligns with this model, as it does not account for genetic vulnerability, it focuses on the relationships between anxiety and depression as key risk factors and emotional intelligence, resilience, and positive thoughts about life as protective factors of suicidal ideation. The selection of anxiety and depression as risk factors for suicidal ideation in teachers is grounded in literature. Research shows that teachers experience high levels of anxiety and depression, which can exacerbate feelings of hopelessness and lead to suicidal thoughts11. Moreover, the emotional toll of the teaching profession, characterized by constant demands and limited resources, can further intensify these conditions, highlighting the urgent need for intervention programs that specifically target the reduction of anxiety and depression among educators.

Conversely, the identification of positive thoughts about life, emotional intelligence, and resilience as protective factors is equally important. Positive thoughts can foster an optimistic outlook, helping teachers cope better with stressors20. Emotional intelligence enables educators to recognize and manage their emotions, enhancing their ability to navigate the challenges of the profession and build supportive relationships with colleagues and students21,22,23. Additionally, resilience equips teachers with the capacity to bounce back from adversity, facilitating effective coping strategies that can mitigate the impact of stress. The capacity to recover from adversity fosters a productive learning environment, benefiting both teachers and students alike. Resilient educators are more capable of managing stress, which enhances their overall effectiveness. Teacher resilience is crucial not only for individual well-being but also for positively influencing student outcomes. When educators demonstrate resilience, they inspire similar traits in their students24,25, nurturing a culture of perseverance. This resilience helps teachers navigate challenging circumstances while remaining dedicated to educational excellence.

This study aims to employ an exploratory approach using psychometric network analysis to investigate the interrelationships between risk factors (anxiety and depression), protective factors (positive thoughts, emotional intelligence, and resilience), and suicidal ideation among teachers in Spain, within the framework of the diathesis-stress model. There are two hypotheses to be tested: (1) suicidal ideation, taken to be the expression of thoughts or desires to end one’s own life, will be positively connected with risk factors and negatively with protective factors; while (2) positive ideation, taken to mean expression of positive thoughts about life, will be positively with protective factors and negatively with risk factors.

The exploratory nature of network models allows for greater flexibility compared to traditional covariance structure models, enabling a nuanced assessment of the connections between risk and protective factors26. This approach can provide valuable insights to guide the development of intervention programs aimed at minimizing risk factors and maximizing protective factors. The estimated strength of interconnections and centrality measures can be highly informative in establishing priority criteria for selecting variables amenable to intervention. To our knowledge, this is the first study to examine teachers’ suicidal ideation in Spain using rigorous psychometric measures and network analysis techniques. By minimizing suicidal ideation in teachers through targeted interventions based on our findings, we may indirectly enhance their capacity to detect and respond to similar suicidal tendencies in their students, thus potentially creating a positive ripple effect throughout the educational community.

Methods

Participants

The sample consists of a total of 1270 teachers from different educational stages. Descriptive statistics are displayed in Table 1.

Table 1 Descriptive statistics of the study’s teacher sample (n = 1270).
Full size table

Measures

The main variable, suicidal ideation, was measured using the Spanish-adapted version of the Positive and Negative Suicide Ideation Inventory (PANSI)27. The PANSI offers a comprehensive approach, measuring both negative risk factors and positive protective factors related to suicidal behaviors, which provides a nuanced understanding of an individual’s suicidal thoughts28,29. Its strong psychometric properties, including high internal consistency and good construct validity across diverse populations, further support its use. Moreover, the reasons behind this selection are its comprehensive assessment of both risk and protective factors related to suicidal ideation, adaptability to teachers’ population, brevity for practical use, and focus on the frequency of thoughts. The inventory comprises 14 items with a 5-point Likert-type response format, forming two factors: (1) negative suicidal ideation (8 items indicating risk factors), and (2) positive thoughts about life (6 items assessing positive thoughts about life as protective factors). Original reliability, assessed with Cronbach’s alpha, was 0.93 for negative ideation and 0.84 for positive thougths.

Risk factors for negative suicidal ideation, specifically anxiety and depression, were measured using the Spanish-adapted version of the General Health Questionnaire (GHQ-12)30. The GHQ-12 has been validated and used in diverse populations worldwide31, including teacher population32. The GHQ-12 comprises twelve Likert-type items assessing psychological well-being based on anxious and depressive symptoms in the last month. The anxiety subscale contains 4 items, and the depression subscale includes the remaining 8 items, with higher scores indicating more negative symptomatology. Original validation studies report Cronbach’s alpha values of approximately 0.75 for both subscales.

In the context of exploring potentially protective variables related to suicidal ideation within our model, we assessed participants’ emotional intelligence using the Brief Emotional Intelligence Scale (BEIS-10) in its validated Spanish version33. While the five-factor model of the BEIS-10 initially showed better fit34, subsequent research has revealed some inconsistent results across different populations, with problematic reliability for individual subscales35. Given these inconsistencies, the scale’s brevity limiting its ability to fully capture five distinct factors, and support for a unidimensional structure in some contexts (such as the Spanish validation), a one-factor model of the BEIS-10 may offer a robust and practical approach to measuring emotional intelligence. This scale comprises 10 Likert-type items with 5 response options, where higher scores indicate greater emotional intelligence. The instrument validation study yielded a reliability coefficient of 0.75, assessed using Cronbach’s alpha.

Finally, resilience was gauged using the Brief Resilient Coping Scale (BRCS), validated in Spanish36. International validations support the BRCS as a robust tool for measuring resilient coping across different cultural contexts37. This unidimensional scale comprises four Likert-type items with five response options, where higher scores signify greater resilient coping. The initial validation study reported a Cronbach’s alpha of 0.86. The BRCS captures resilience through optimism, perseverance, creativity, and positive growth in adversity, depicting resilient coping as an active problem-solving approach.

Procedure

A cross-sectional selective design was used to study the suicide ideation network in teachers. Intentional sampling considered strata within the teacher population for better representativeness. Scales and sociodemographic variables were combined into a LimeSurvey online questionnaire (30 min). The process involved visiting educational centers with scheduled appointments with principals. Principals gathered teaching staff in a room with internet coverage, providing electronic devices. Participants gave informed consent, understanding research objectives, data handling per data protection laws, and school benefits (a technical report with aggregated results). Subsequently, each participant was individually provided with a QR code that directly linked to the questionnaire, and ample time was allocated for them to complete it. A member of the research team was present to address any questions, difficulties, or concerns that may have arisen and to ensure that the conditions for the proper administration of the questionnaires were met. Out of a total of 1,820 teachers contacted, 1,270 ultimately completed the protocol, resulting in a participation rate of 69.78%. All participants completed the questionnaire between the months of October and November 2023.

Ethical aproval

In ensuring compliance with ethical regulations, the protocol for this study has received approval from the Ethics Committee of the University of the Balearic Islands (Approval Number: 365CER23). All methods were conducted in accordance with Declaration of Helsinki and its later amendments or comparable ethical standards. Participant data has been anonymized in adherence to Article 9 of Organic Law 3/2018, dated December 5, pertaining to the protection of personal data and the guarantee of digital rights in Spain, as well as Article 9 of Regulation (EU) 2016/679 of the European Parliament and of the Council, dated April 27, 2016, concerning the protection of individuals regarding the processing of personal data and the free movement of such information. The protocol also included an alert system for disseminating information and providing assistance in the event individuals with a high level of suicidal ideation were identified.

Statistical analysis

The first step of the analytical plan consisted of a preliminary data examination aimed to check quality and verify statistical assumptions, including univariate (Kolmogorov-Smirnov) and multivariate (Mardia) normality tests. The dataset had no missing values, and no imputation methods were used. The second one was to estimate a fit assessment of each constructs’ latent structure through Confirmatory Factor Analysis (CFA). A set of one-factor confirmatory model using Structural Equation Modeling (SEM) for suicidal ideation, positive thoughts about life, anxiety, depression, emotional intelligence, were estimated with Maximum Likelihood method and robust standard errors (MLR), and computed using the ‘lavaan’38 and ‘semTools’39 R program packages40. We subsequently obtained Omega coefficient as a reliability indicator. Overall model fit followed criteria by Hu and Bentler41: a non-significant chi-square, CFI ≥ .95, RMSEA ≤ .05 (with 90% CI), and SRMR ≤ .08, indicating excellent fit. Specifically, for RMSEA, widely used in the literature for assessing overall fit, values lower than .08 were deemed acceptable.

After assessing the latent structure and reliability of each scale of the study, in the third step we estimated network models about suicidal ideation, the risk and protective factors. Although network models initially debuted in the field of psychology with a focus on how symptoms could configure psychopathologies and syndromes, their application is evolving and cannot be limited solely to clinical objectives. The use of network analysis in psychology is expanding to examine relationships between psychological constructs, moving beyond individual symptoms. For instance, these models are being applied in social and organizational psychology to study group dynamics and interpersonal relationships42. We leverage this broader application of network models to understand how risk factors like anxiety and depression interact with protective factors such as emotional intelligence, resilience, and positive thoughts in relation to suicidal ideation among teachers. This approach allows us to examine interactions at a broader level, identify potential direct and indirect relationships among these factors, and visualize the structure of these relationships in a way that is accessible to education professionals and policymakers. Network analysis allows us to examine complex interplay and relationships among variables, and identify central factors that may not be apparent from correlations alone. A similar approach has been recently applied by De Beurs et al.26 using a constructs network approach focusing on suicidal ideation. By employing network analysis in this manner, we aim to contribute to the growing body of research that applies these models beyond their original focus, highlighting their versatility and potential in various areas of psychological research.

We implemented the network analysis using two approaches: The first approach considered a Global network model including total scores of suicidal ideation, positive thoughts about life, anxiety, depression, emotional intelligence, and resilience and was estimated using the ‘bootnet’43 R package. For Global network estimation, we used a Gaussian graphical model with graphical LASSO (EBICglasso) and chose the regularization parameter via the ‘walktrap’ algorithm. The second approach, the Specific network model, was identical to the previous global model but involving the 14 items’ scores from the two dimensions of suicidal ideation according to the PANSI structure. The Specific network model was estimated using the ‘EGAnet’44 R package using Exploratory Graph Analysis (EGA), akin to factor analysis, identifying relationships and variable groupings. Factor loadings facilitated a comprehensive analysis of suicidal ideation dimensions, exploring item interrelationships and associations with the other variables in the model, that in turn, can potentially forming the risk and protective factors.

In network graphs, nodes represent variables connected by lines, their thickness indicating association strength (thicker lines denote stronger relationships). Line color signifies the relationship’s sign; red lines typically indicate a negative relationship, while positive relationships can be represented by green or blue lines. To assess the robustness of the two networks, we performed a non-parametric bootstrap procedure (n = 1000) involving the calculation of the correlation stability coefficient. Additionally, we conducted bootstrapped difference tests to examine the statistical significance of the estimates derived from the global bootnet and specific EGA networks obtained through the ‘bootnet’ and ‘EGAnet’ R packages, respectively. The Correlation Stability (CS) coefficient served as a metric to evaluate whether the expected influence centrality index exhibited significant variation across successive case-dropping subsets. A CS value exceeding 0.5 indicated adequate stability.

In the concluding phase, we examined the metric invariance, specifically the equality of loadings, within the EGA structure that was derived for the specific network comprising items. This analysis was conducted across two subsamples formed based on the chromosomal sex reported by the participants (male/female). These techniques do not currently support the assessment of more stringent levels of invariance. Figure 1 displays the flow diagram including the steps of the analytical plan.

Fig. 1
figure 1

Flow diagram including the steps of the study’s analytical plan.

Full size image

Results

Preliminary results

The data matrix underwent thorough error-checking; no imputation was needed due to no missing values. Univariate normality tests (Kolmogorov-Smirnov with Lilliefors correction) revealed non-normal distributions for PANSI items and criterion variables (p < .001). Multivariate normality, dependent on univariate normality, was not met (Mardia’s skewness = 5108.30, p < .001; Mardia’s kurtosis = 87.53, p < .001). Robust methods, using non-parametric bootstrap for network estimation, addressed violations in latent structure models. See Table 2 for PANSI items and criterion variables correlation matrix in secondary analyses.

Table 2 Correlation matrix and descriptive statistics of the PANSI items, the risk factors (anxiety and depression), the protective factors (emotional intelligence and resilience), and the overall score for negative and positive ideation.
Full size table

Before network estimation, we evaluated unifactorial latent structure suitability for the global network’s six dimensions: negative suicidal ideation (PANSI-), positive thougths about life (PANSI+), anxiety, depression, emotional intelligence, and resilience. Negative suicidal ideation showed a good one-dimensional fit (Chi-square = 25.14, df = 20, p = .196; RMSEA = 0.014; CFI = 0.995; SRMR = 0.019); the positive subscale had slightly worse but appropriate results (Chi-square = 55.05, df = 9, p < .001; RMSEA = 0.064; CFI = 0.961; SRMR = 0.035). Criterion variables demonstrated reasonable unidimensional fit: Anxiety (Chi-square = 25.08, df = 2, p < .001; RMSEA = 0.070; CFI = 0.956; SRMR = 0.035), Depression (Chi-square = 75.39, df = 14, p < .001; RMSEA = 0.061; CFI = 0.920; SRMR = 0.046), Emotional Intelligence (Chi-square = 881.59, df = 35, p < .001; RMSEA = 0.067; CFI = 0.951; SRMR = 0.056), and Resilience (Chi-square = 4.53, df = 6, p = .104; RMSEA = 0.032; CFI = 0.997; SRMR = 0.013). With satisfactory measurement scores confirmed, we proceeded to estimate the network models (global and specific). Reliability of the scores was adequate for the six scales using the Omega coefficient: 0.94 (PANSI-), 0.81 (PANSI+), 0.70 (GHQ-Anxiety), 0.80 (GHQ-Depression), 0.77 (BEIS-10), and 0.75 (BRCS).

Global network

Global network (Fig. 2) explores relationships among total scores for suicidal ideation, risk factors (anxiety and depression), and protective factors (positive thoughts, emotional intelligence and resilience). Suicidal ideation is mainly influenced by depression (0.18), with a minor impact from anxiety (0.14), showing negative associations with positive thoughts (-0.21) and emotional intelligence (-0.02), serving as protective factors; resilience shows no significant relationships. Positive thougths is negatively associated with depression (-0.21) and, to a lesser extent, anxiety (-0.1). Positive thoguths is positively influenced by emotional intelligence (0.21) and resilience (0.14), both carrying similar weight. Depression strongly associates with comorbid anxiety (0.43) and negatively links with resilience (-0.07) but not emotional intelligence. Anxiety exhibits weak negative associations with resilience (-0.06) and emotional intelligence (-0.05). Emotional intelligence and resilience display a robust and positive interrelationship (0.48).

Fig. 2
figure 2

Global network model including overall positive and negative suicidal ideation variables, risk and protective factors with strength edges’ values (bootstrapped solution, n = 1000).

Full size image

In network metrics for psychological relationships, we analyze the expected influence or “node expected influence”, indicating the cumulative impact of a node’s connections, assessing its overall importance by considering both positive and negative relationships. This measure provides insight into how changes in one variable might influence the entire network, taking into account the sign of connections. Expected influence is particularly useful in psychological networks where variables can have both facilitating and inhibiting effects on each other, offering a more nuanced view of a node’s role in the network structure. Figure 3 represents the centrality plot of the expected influence statistic of our global network.

Fig. 3
figure 3

Centrality plot of the Global network (Expected influence).

Full size image

The expected influence values indicate that depression (0.33) and anxiety (0.26) are positively associated with suicidal ideation, while emotional intelligence (0.62) and resilience (0.50) are positively related to positive thoughts and negatively associated with suicidal ideation. Positive thoughts show a negative expected influence (-0.17), suggesting they may inhibit suicidal ideation. Suicidal ideation itself has a nearly zero expected influence (-0.01), which might reflect a balance between the positive influences of risk factors and the negative influences of protective factors. This interplay in the network suggests that enhancing emotional intelligence and resilience could support positive thoughts, potentially offering some protection against suicidal ideation, while also highlighting the importance of addressing depression and anxiety due to their risk-enhancing roles.

To assess global network stability, we used non-parametric bootstrap, resampling the dataset 1000 times. This created a model distribution, addressing data variations. Edge weights in each model were examined for connection strength, and confidence intervals were calculated to show potential value ranges. Wider intervals signaled more uncertainty, narrower ones indicated greater reliability. Bootstrap results not only boosted confidence in network estimates but also provided insights into variable relationships. Figure 4 shows mean stability plot values for 15 connections across bootstrapped global network samples, with confidence intervals.

Fig. 4
figure 4

Network stability plot: Mean edge values (with confidence intervals) across 1000 bootstrapped estimated Networks.

Full size image

Examining the graph’s horizontal axis at zero reveals two non-significant relationships where the network did not draw lines. The confidence intervals for the mean values of the resilience and suicidal ideation, and emotional intelligence and depression relationships, include zero. In the lower-left and upper-right corners, representing negative and positive relationships, respectively, the highest magnitude relationships are located. The line for the initial network values and the mean of the 1000 resamples for each edge show substantial overlap. Simultaneously, the gray band representing confidence intervals displays minimal variations, indicating remarkable stability across edges. This evidence points to a robust and stable estimated relational structure model. Conducting a second Bootstrap analysis involved gradually reducing sample sizes from 100 to 25%, evaluating the network’s stability and its ability to retain identified connections. The average correlation for the expected influence statistic was then calculated in relation to the complete sample (Fig. 5).

Fig. 5
figure 5

Global network stability analysis. Plot of average correlation between expected influence centrality index of networks resampled with persons dropped and the original sample for the Global network.

Full size image

The stability of each node expected influence is robust, with the trend reaching an average correlation exceeding 0.75. The CS-coefficient measures the proportion of our sample that can be removed while maintaining a correlation of r = .70 between the original expected influence centrality index and that derived from bootstrapped samples. A CS-coefficient above 0.50 indicates strong evidence of node stability within the population, independent of sample-specific characteristics. Ideally, the CS-coefficient should exceed 0.25, with 0.50 or higher being preferable. In our global network, the CS-coefficient for expected influence is 0.75, substantiating the model’s structural stability.

Specific network (EGA)

We estimated a specific EGA model incorporating all PANSI ítems (suicidal ideation and positive thougths). This facilitated examining factor loading behavior and item content’s impact on risk (anxiety and depression) and protective factors (emotional intelligence and resilience). Table 3 shows the items’ contents of the PANSI to facilitate the understanding of the relationships. Figure 6 displays the EGA model using the glasso procedure and walktrap algorithm, post-bootstrap (n = 1000) of the initial sample solution.

Table 3 Contents of the items of the PANSI by dimensions (negative and positive).
Full size table

The EGA model’s bootstrap analysis found four different latent structures, with the 4-factor structure being most prevalent (76.6%), followed by the 3-factor structure (21.7%). The 2-factor structure (1.3%) and the 5-factor structure (0.4%) were practically residual. The four-factor model categorizes dimensions of negative and positive suicidal ideation, grouping risk and protective variables into separate factors. All PANSI items are accurately categorized based on their relevance to suicidal ideation or positive thougths dimensions. Analyzing bootstrap resampling, items from the suicidal ideation dimension were correctly assigned as follows: items 3, 7, 9, and 11 at 100%, item 10 at 99.7%, item 1 at 94.4%, item 4 at 92.7%, and item 5 at 71.6%.

Fig. 6
figure 6

Specific network model. EGA model bootstrapped solution (n = 1000) for the relationships among the PANSI items conforming positive and negative dimensions, and the risk and protective factors.

Full size image

Concerning positive thougths, items 2, 6, 8, 12, and 14 consistently belong to their dimension in 100% of samples, while item 13 is correctly assigned in 97.2% of cases. For risk factors, anxiety saturates the dimension in 81.8% of samples, and depression does so in 82.3%. In the protective factors dimension, both emotional intelligence and resilience are correctly attributed in 93% of samples. Overall structural consistency for each of the four dimensions stands at 68.6% for suicidal ideation, 98.3% for positive thougths, 93% for protective factors, and 82.3% for risk factors.

In terms of average connection strength for each variable in the model (Fig. 7), the highest values are seen for item 14 (positive thougths), item 7 (suicidal ideation), depression (risk factors), and resilience and emotional intelligence (protective factors). Node strength quantifies the combined weight and significance of a node’s connections in a network, considering both quantity and weight. Regarding stability indices (CS) of the EGA network, considering maximum drop proportions to maintain a correlation of 0.7 in at least 95% of samples, the values for strength was 0.75. Strength surpasses the 0.50 cutoff for structural stability.

Fig. 7
figure 7

Centrality plot of the Specific network model (strength).

Full size image

After confirming structural stability, we interpret identified relationships. Regarding risk variables’ influence on items, a strong, positive relationship is observed between depression and items 5 (Did you think about ending your life because you couldn’t accomplish something that was very important in your life?) and 1 (Have you seriously considered taking your own life because you couldn’t meet the expectations others had of you?). This connection links depressive states to suicidal thoughts arising from unmet goals and others’ expectations.

In contrast, anxiety doesn’t directly impact negative ideation items but influences them solely through depression mediation. Nevertheless, anxiety maintains direct negative associations with items 14 (Did you have confidence in achieving your goals in the future?) and 12 (Did you have confidence in your abilities to deal with most of life’s problems?) in the positive thougths dimension. Its connection with other items in this dimension is mediated. In this scenario, anxiety substantially affects confidence in achieving goals and the ability to cope with life’s challenges.

Regarding protective variables, emotional intelligence positively affects items 6 (Did you have hope for the future because things were going the way you wanted them to?) and 12. Once again, item 14 serves as a conduit for this effect on the remaining items. Resilience also displays direct positive relationships with items 12 and 14. The model suggests that emotional intelligence and resilience can mitigate anxiety’s adverse effects by enhancing emotional management skills and coping resources for personal and professional life events. Consequently, this improvement in personal resources can heighten the sense that life is worth living (item 13, Did you feel that life was worth living?), reducing the probability of perceiving life as a failure and generating suicidal ideation (item 9, Did you think about ending your life because you saw your life as a failure?).

Within positive thougths and suicidal ideation dimensions, the most significant impact is mediated through item 13, especially towards item 9, and items 3 (Did you contemplate suicide because you had no hope for the future?), 7 (Did you think about suicide because you couldn’t find a solution to a personal problema?), 10 (Did you believe your problems were so severe that the only option you had was to take your own life?), and 11 (Have you felt so alone or so sad that you wanted to end your life to put an end to that suffering?). Furthermore, item 6 also exhibits direct negative associations with respect to items 9 and 10. The model suggests that transitioning between positive thougths and suicidal ideation hinges on perceiving life as meaningful and fostering hope for the future, driven by events aligning with our desires. Within each PANSI dimensions, all items are interconnected, although items in the suicidal ideation dimension create a more closely connected relationship pattern than the positive thougths dimension. Nevertheless, none of the items in either dimension is isolated.

Finally, we conducted a network invariance analysis for the EGA model, taking into account the chromosomal sex (male/female) self-reported by participating teachers. This analysis aimed to determine if the models for both men and women exhibit the same underlying structure (configural level) and whether the relationships between variables have the same weights (metric level). The results demonstrate configural and metric invariance, as the four-factor structure is consistently present in both subsamples, and all network loadings remain invariant (p < .001).

Discussion

The present study aimed to investigate the complex interplay between suicidal ideation, risk factors, and protective factors among teachers using network analysis in the framework of the diathesis-stress model. To the best of our knowledge, this is one of the few studies conducted in Spain on suicidal ideation among teachers utilizing validated psychometric measures. Our findings provide valuable insights into the relationships between these variables and their potential implications for suicide prevention strategies in educational settings. The global network analysis revealed that depression has the strongest direct relationship with suicidal ideation, followed by anxiety. Suicidal ideation’s link with depression may manifest in emotional exhaustion, disinterest in teaching tasks, and negative self-perception, while anxiety can contribute to sustained preoccupation with performance and uncertainty in the educational environment. This aligns with previous research highlighting the significant role of depressive symptoms in suicidal thoughts and behaviors among teachers45,46.

The strong positive relationship between depression and anxiety in our network supports the well-established comorbidity between these two conditions, which has been observed in various occupational groups, including educators47. The implications of the association between depression, anxiety, and negative suicidal ideation are significant, with a potential relationship not only teachers’ personal well-being but also the quality of teaching and the educational environment. Given teachers’ daily high work and emotional demands, there’s a vulnerability to chronic stress, supporting the rationale of the diathesis-stress model48. The absence of intervention and prior assessment of anxiety and depression levels, potentially leading to suicidal ideation, could negatively affect students’ training and development.

Our results also underscore the protective nature of positive thoughts and emotional intelligence, which showed negative associations with suicidal ideation49. Interestingly, resilience demonstrated weak negative associations with both depression and anxiety, suggesting its potential role in buffering against these risk factors. This finding is consistent with previous studies that have identified resilience as a crucial factor in maintaining teachers’ mental health and well-being49. In terms of intervention, the model highlights a positive relationship between emotional intelligence and resilience, acting as protective factors on positive ideation, expressing positive thoughts about life50. This positive ideation serves as a robust protective factor against negative ideation. Similar to the strong comorbidity among risk factors, the model indicates a robust positive association between emotional intelligence and resilience. These results support the hypothesis that higher emotional intelligence levels, encompassing the ability to recognize and manage emotions, enhance teachers’ stress management and ability to navigate challenges in the classroom. Emotional intelligence channels positive emotions, fosters positive thoughts about life, and reduces suicidal thoughts.

In relation to resilience, defined as the ability to recover from difficult situations and considered a protective factor, the model suggests that more resilient teachers approach challenges with a mindset that promotes overcoming difficulties, positively influencing their life outlook. However, contrary to expectations, the model indicates a non-significant relationship between resilience and negative suicidal ideation. Consequently, the impact of resilience on negative ideation could be entirely mediated by positive ideation. Similarly, resilience, operating through positive cognition and emotional intelligence, could moderate the relationship between negative emotions and suicide risk51. It’s important to note that the network’s estimation alone do not provide sufficient evidence to claim mediation effects. So, specific mediation tests for network models52 should be implemented for future research to explore potential mediating relationships. Despite the acknowledged value of emotional intelligence, the lack of resources and institutional support in educational settings may outweigh its benefits for preventing depression. In essence, the link between emotional intelligence and depression in teachers may not be as straightforward due to the specific demands of their work environment. Exploring the relationship between emotional intelligence and mental health in various populations necessitates careful consideration of these contextual factors.

The specific EGA network provided a deeper understanding of the relationships between individual PANSI items and the risk and protective factors. Depression showed strong positive connections with items related to suicidal thoughts arising from unmet goals and others’ expectations. This highlights the importance of addressing performance-related stress and external pressures in suicide prevention efforts for teachers. The teaching profession is often characterized by high expectations and performance demands, which can contribute to feelings of inadequacy and depression when these expectations are not met53. Our findings suggest that interventions targeting these specific aspects of teacher stress may be particularly effective in reducing suicidal ideation.

Although anxiety is not directly linked to negative ideation items, it may be associated with them through depression and can affect confidence in achieving goals and coping with life’s challenges. This suggests that anxiety management strategies could indirectly reduce suicidal ideation among teachers. Previous research has shown that anxiety is a common issue among educators, often stemming from workload pressures, classroom management challenges, and administrative demands54. Implementing school-based programs that focus on anxiety reduction techniques, such as mindfulness practices or cognitive-behavioral strategies, may help alleviate these symptoms and, in turn, decrease the risk of suicidal ideation55,56.

Emotional intelligence and resilience demonstrated positive relationships with items related to hope for the future and confidence in problem-solving abilities. These findings support the potential of interventions aimed at enhancing emotional management skills and coping resources for teachers57. Emotional intelligence has been identified as a crucial factor in teacher effectiveness and well-being58, while resilience has been associated with reduced burnout and increased job satisfaction among educators59. Developing professional development programs that focus on building these skills could have a significant impact on reducing suicidal ideation and improving overall mental health in the teaching population.

Our results indicate that perceiving life as meaningful and maintaining hope for the future play crucial roles in the transition between positive thoughts and suicidal ideation. This underscores the importance of fostering a sense of purpose and optimism among teachers. The teaching profession often attracts individuals who are driven by a sense of purpose and a desire to make a positive impact on students’ lives60. However, the challenges and stressors associated with the job can sometimes overshadow this sense of meaning. Interventions that help teachers reconnect with their core values and the intrinsic rewards of their profession may be particularly effective in promoting positive thoughts and reducing the risk of suicidal ideation. This underscores the need for preventive intervention to boost institutional recognition of teaching staff’s contributions and allocate resources, fostering a sense of fulfillment and social recognition. Importantly, the teaching profession’s gradual loss of social status exacerbates this situation.

Overall, the model suggests that the shift from positive to negative thoughts hinges on perceiving life as meaningful and nurturing hope for the future. To address the transition between negative and positive suicidal thoughts among teaching staff, interventions should focus on strengthening the perception of life’s meaning and fostering hope. Recommended interventions include emotional support programs, coping strategy training, and creating work conditions that align with individual desires. By concentrating on these aspects, the goal is to cultivate an educational environment that enhances mental health, thereby reducing negative suicidal thoughts and promoting positive thinking among teaching staff.

In sum, the global network emphasizes that the pathway to reducing the likelihood of negative suicidal ideation involves enhancing positive ideation, which can be amplified by addressing both emotional intelligence and resilience, while simultaneously seeking to decrease the levels of depression, to a greater extent, and anxiety as well. It is worth noting that the relationship of protective factors is stronger with positive suicidal ideation itself than with negative ideation and risk factors. This pattern suggests the existence of a quasi-complete mediation model between protective factors and negative suicidal ideation, mediated through positive ideation. Future studies will incorporate specific network techniques to analyze these mediation relationships more thoroughly.

Finally, the invariance analysis across chromosomal sex in our study supports the generalizability of our findings to both male and female teachers. This suggests that intervention strategies based on our results could be broadly applicable across the teaching population. However, it is important to consider that other demographic factors such as age, years of experience or teaching level may influence the relationships between variables in our network. Future research could explore these potential differences to further refine intervention strategies for specific subgroups of teachers.

Implications for intervention

Based on our findings, we propose several implications for suicide prevention strategies targeting teachers. First, prioritizing depression screening and treatment is crucial, given its strong direct relationship with suicidal ideation. School districts and educational institutions should consider implementing regular mental health check-ins for teachers, similar to those already in place for students in many schools. These screenings could help identify teachers at risk for depression and provide early intervention opportunities. Additionally, ensuring that teachers have access to confidential and affordable mental health services, either through employee assistance programs or partnerships with local mental health providers, is essential.

Second, implementing anxiety management programs can indirectly reduce suicidal thoughts and enhance confidence among teachers. These programs could include stress reduction workshops, time management training, and strategies for dealing with challenging classroom situations. Mindfulness-based interventions have shown promise in reducing teacher stress and anxiety61 and could be incorporated into professional development offerings or made available through online platforms for easy access.

Third, developing interventions aimed at boosting emotional intelligence and resilience can strengthen protective factors against suicidal ideation. These interventions could focus on enhancing self-awareness, emotional regulation, and adaptive coping strategies. For example, programs like the Cultivating Awareness and Resilience in Education (CARE) for Teachers have demonstrated effectiveness in improving teachers’ well-being and classroom climate62. Adapting and implementing such programs more widely could have significant benefits for teacher mental health.

Fourth, fostering a sense of purpose and meaning in teachers’ professional lives is crucial for promoting positive thoughts and reducing suicidal ideation. This could involve creating opportunities for teachers to engage in meaningful professional development, collaborate with colleagues on innovative projects, or participate in mentoring programs63. Additionally, school leaders can play a role in recognizing and celebrating teachers’ contributions to student success and the school community, their teacning talent, reinforcing the value and impact of their work64.

It is important to note that these interventions should not be implemented in isolation but rather as part of a comprehensive approach to teacher well-being. This approach should address both individual and systemic factors that contribute to teacher stress and mental health issues. For example, addressing workload concerns, improving school climate, and providing adequate resources and support for teachers are all critical components of a holistic strategy to reduce suicidal ideation and promote mental health in the education sector65.

Furthermore, our findings highlight the interconnected nature of risk and protective factors in relation to suicidal ideation in the framework of the diathesis-stress model for suicide behavior. This suggests that interventions targeting multiple factors simultaneously may be more effective than those focusing on a single aspect. For instance, a comprehensive program that addresses depression and anxiety while also building positive thougths about life, emotional intelligence and resilience skills could have a synergistic effect in reducing suicidal ideation among teachers.

Limitations of the study

While our study offers valuable insights, it has several limitations. The cross-sectional design of the data restricts our ability to draw causal inferences about the relationships between variables. Longitudinal studies are necessary to better understand the temporal dynamics of these relationships and how they may evolve over time or in response to interventions. Furthermore, although our focus on teachers provides important insights for this specific population, the use of intentional sampling may limit the diversity of our sample and impact the generalizability of our findings. Additional research is required to assess how these insights apply to teachers across different cultural contexts and settings. Another limitation is not having introduced measures of perceived stress and/or burnout, assuming that the profession in general presents high levels of both variables. In this line, vulnerability measures should be also included to cover the scope of the diathesis-stress model.

Conclusions

In conclusion, this study offers a comprehensive network analysis of the risk and protective factors associated with suicidal ideation among teachers. Our findings emphasize the complex relationships between these variables and provide critical insights for developing targeted suicide prevention strategies within educational settings. By addressing both risk and protective factors, interventions can potentially reduce suicidal ideation and enhance the mental well-being of educators as, in turn, gatekeepers for suicide risk in their students.

Future research should focus on developing and evaluating interventions based on these findings, as well as exploring the long-term effects such interventions may have on teacher mental health and retention in the profession. The mental health and well-being of educators are essential not only for the individuals themselves but also for the overall quality of education and student outcomes. Investing in comprehensive strategies to support teacher mental health can reduce the risk of suicidal ideation and create more positive and effective learning environments for both educators and students. As we continue to face challenges in education, including increased stress and burnout among teachers, it is imperative to prioritize the mental health of those who play such a vital role in shaping future generations.