Introduction

The ability to overcome adversity and thrive has inspired many storytellers and songwriters over the past several decades; however, it took some time until it started to catch the attention of psychiatrists and psychologists. In the 1970s and 1980s, while interviewing Holocaust survivors, Aaron Antonovsky recognized these individuals’ outstanding resilience (Mittelmark, 2022).

Based on his observations, Antonovsky developed a “salutogenic” view of mental health that has gained attraction over the past decades and has contributed to the understanding of psychological functioning. Instead of trying to understand the origin of diseases (as in pathogenesis), the salutogenic model focuses on factors that initiate and sustain health and well-being (Mittelmark, 2022). One essential part of the salutogenic health model is what he describes as a “sense of coherence,” with the notion that the effective management of stress and finding of meaning maintain health and promote overall well-being and resilience (Mittelmark, 2022; Antonovsky & Sagy, 1986).

Research on and definitions of the sense of coherence have evolved through extensive investigations of stress and coping mechanisms and have been refined through cross-cultural studies (Mittelmark, 2022; Antonovsky & Sagy, 1986). Antonovsky finally defined the sense of coherence as “a global orientation that expresses the extent to which one has a pervasive, enduring though dynamic feeling of confidence that the stimuli from one’s internal and external environments in the course of living are structured, predictable, and explicable the resources are available to one to meet the demands posed by these stimuli; and these demands are challenges, worthy of investment and engagement” (Mittelmark, 48,2,3,; Antonovsky, 1993).

In Antonovsky’s definition, we find the three core elements of the sense of coherence: comprehensibility, manageability and meaningfulness (Antonovsky, 1993). Comprehensibility allows one to appreciate reality without becoming overwhelmed, enabling one to predict coming events and plan accordingly. Manageability refers to the ability to use available resources and means and to build and maintain relationships. Meaningfulness refers to the ability to see a life purpose and discern positivity in adversity (Mittelmark, 2022; Antonovsky, 1987). The sense of coherence can be conceptualized as a complex system of cognitive (comprehensibility), behavioral (manageability), and motivational (meaningfulness) dimensions shaped by interactions (Portoghese, 2024).

Depression is a recurrent debilitating disorder characterized by a persistent feeling of sadness or loss of interest; it also impairs education, relationships, and employment and is associated with obesity, cardiac disease, and early death, including suicide (Marwaha et al., 2023). Depression is one of the leading causes of disability, both in terms of years lost and excess mortality (Herrman et al., 2019). The biomedical models conceptualize depression as a disorder of widely distributed neural networks with abnormalities in neurotransmitters (Marwaha et al., 2023; Moncrieff et al., 2023). Although a biopsychosocial approach to treating depression (including psychological interventions and social support) is recommended, medication is often considered essential for more severe cases (Marwaha et al., 2023).

A salutogenic approach and understanding of depressive disorders might help individuals identify strategies for promoting mental health and preventing depressive disorders. In particular, the sense of coherence seems to decrease in persons with mental health problems (Schafer et al., 8,10,; Mattisson et al., 2014), with an increasing magnitude from childhood to young adulthood (Schafer et al., 2023). A poor sense of coherence is particularly strongly related to internalizing disorders, such as anxiety and depression (Schafer et al., 2023; Ristkari et al., 2006; Konttinen et al., 2008). Furthermore, a low sense of coherence seems to predict the onset of depression (Sairenchi et al., 2011). However, the exact underlying mechanisms are unclear (Sairenchi 2011), and some evidence suggests a mediating role (between personality structure and depression) of the different factors of the sense of coherence (Kövi et al., 2017; Pallant & Lae 2002).

The current study aimed to analyse the relationship between a sense of coherence and depression to gain insight into how these two psychological constructs might interact with each other, especially how well-being, mental health, and particularly depression are affected. The study sets itself apart from other studies by implementing a network analytical approach. By visualizing and statistically modelling the relationship, we expect to gain detailed insight into the structure and dynamics of the relationship between the sense of coherence and depression, thus allowing us to identify its nuances and mechanisms.

Methods

Study design and population

We designed a prospective, randomized, controlled study to evaluate the effect of Internet Cognitive Behavioural Therapy (ICBT) (Rauen, 2020). The local ethics committee of the Canton of Zurich in Switzerland approved the study (BASEC-Nr. 2013-0542); it was registered after receiving approval (clinicaltrials.gov NCT02112266). The study was conducted in accordance with the Declaration of Helsinki, and all the subjects provided their electronic informed consent before participation.

Patients with depressive symptoms were recruited online and assessed for demographic and clinical baseline characteristics. The inclusion criteria were age between 18 and 65 years and at least two weeks of moderate to severe depressive episodes. The exclusion criteria were low German proficiency, suicidal ideation, alcohol or substance use, history of psychotic symptoms or bipolarity, and current treatment (Rauen, 2020).

We used demographic and clinical baseline characteristics for the current secondary analysis. A total of 839 people started the questionnaire; 319 reported having a current depressive disorder. Of those, 181 participants who completed the questionnaires without missing items were included in the current analysis.

Clinical assessments

Beck Depression Inventory-I (BDI-I). The BDI-I was developed to measure the severity of depression in adults (Beck et al., 1961). The BDI-I is a self-report questionnaire with 21 items, each with a choice of four statements. The statements describe symptom severity along an ordinal continuum from absent (0) to very severe (3): with a range from 0 to 63, higher scores are considered indicative of more severe depression. The BDI-I has good psychometric properties and is widely used in research and clinical practice (Beck et al., 1988). The BDI-I scale used in the present study is the validated German version; it has similar psychometric properties to the original scale, with an alpha of 0.89 (Kammer, 1983).

Sense of Coherence Scale (SOC-13). The SOC-13 was developed to measure a person’s ability to deal with stress and adversity (Antonovsky, 1987). It is a self-report questionnaire with 13 items, each evaluating a series of statements regarding resources and dispositions. Each item is rated on a seven-point Likert scale ranging from absent (seldom) to marked (frequent); five items (“1,” “2,” “3,” “7,” and “10”) are inversely rated. The SOC-13 concentration ranges from 0 to 78, with higher values indicating a greater sense of coherence (Antonovsky, 1993). This scale has shown good psychometric properties and is widely used in research practice (Eriksson & Contu, 2022). We used the validated German version of the SOC-13 scale, which also has good psychometric properties, with an alpha of 0.85 (Singer & Brähler, 2007; Schumacher et al., 2000).

Statistical analysis

Descriptive statistics (proportion, mean, and standard deviation) are presented for patient demographics and baseline and outcome characteristics. The BDI-I and the SOC-13 internal consistency were examined using Cronbach’s alpha coefficient. We calculated the Pearson correlation coefficient between the BDI-I score and the SOC-13 score. Considering the differences in the scales’ ratings, we calculated the z scores for both scales before examining the level of accuracy and precision between the BDI-I and the SOC-13 with the concordance correlation coefficient (Lin, 1989; King & Chinchilli, 2001). To evaluate the agreement between the two scales, we used the Bland‒Altman plot. The difference between the two scales was plotted on the y-axis, while the mean was plotted on the x-axis. The confidence intervals and limits of agreement for both scales were calculated (Bland & Altman, 1986; Carkeet, 2015).

In network models, variables (i.e., items) are presented as nodes connected via edges, representing undirected regularized partial correlations (Borsboom, 2021). The network models of the scales (BDI-I and SOC-13) were calculated using the “extended Bayesian information criterion” (EBIC) and the “least absolute shrinkage and selection operator” (LASSO) regularization methods implemented within a single Gaussian random field network. For the degree of shrinkage, we used a low hyperparameter (gamma = 0.0) to maximize the stability of the network and balance sensitivity and specificity (Epskamp et al., 2018). To test the accuracy and stability of the network parameters, we estimated confidence intervals for the edge weights and the correlation stability coefficient using nonparametric bootstrapping (Epskamp & Fried, 2018).

The topological properties of the networks were described using centrality measures: strength, closeness, betweenness, expected influence, and bridge influence. The strength sums the absolute edge weights of the edges per node. Closeness quantifies the distance between one node and all other nodes by averaging the shortest path lengths. Betweenness, on the other hand, quantifies how frequently a node lies on the shortest path connecting two other nodes. The expected influence, in contrast to strength, considers the sign of the edge weight; it quantifies the variance influenced by the surrounding nodes. Bridge influence quantifies the variance of a node accountable for another construct or dimension, in our case, how sense of coherence determines depressive symptoms (Borsboom, 2021; Jones et al., 2021; McNally, 2021). To identify the outstanding nodes (i.e., items), we normalized the centrality measures and identified the nodes above the 95th percentile.

Statistical analyses and figures were generated using RStudio (2023.12.1 + 402); the statistical software R (4.3.2); and the R packages tidyverse (2.0.0), ltm (1.2-0), blandr (0.5.1), pwr (1.3-0), qgraph (1.9.8), bootnet (1.6), and networktools (1.5.1).

Results

Demographic characteristics of the sample

The mean age of the individuals in the sample was 35.59 (11.50) years (range 18–63), and three-quarters of the participants were females (76.8%, n = 139). The participants had an average education of 15.18 (5.23) years; approximately two-thirds had a professional education (i.e., either a completed apprenticeship or a higher education college/university degree) (62.4%, n = 113); most participants were either employed (48.0%, n = 87) or were in training/formation (24.3%, n = 44). All participants had at least one previous depressive episode; the duration of the current episode was 30.30 (77.30) days (for further details, see Table 1).

Table 1 Demographic and clinical characteristics of the sample (n = 181)
Full size table

Psychometric characteristics of the sample

The mean BDI-I sum score was 49.90 (9.26) points, reflecting moderate to severe depression. The internal consistency of the BDI-I is excellent, with a Cronbach’s alpha of 0.84. The mean SOC-13 sum score was 40.56 (9.99) points. The internal consistency of the SOC-13 is good, with a Cronbach’s alpha of 0.72. The correlation index between the BDI-I score and SOC-13 score was − 0.65 (95% CI: − 0.72 to − 0.55). The Bland‒Altman plot showed good overlap between the two scales, with only two (1.05%) outliers. A post hoc analysis yielded a statistical power of 0.99, indicating a high probability of producing accurate results (Cohen 2013) (Fig.1).

Fig. 1
figure 1

Correlations between the BDI-I and the SOC-13 scores (A). Bland‒Altman plot for the BDI-I and the SOC-13 scale (B). BDI-I Beck Depression Inventory, SOC-13 Sense of Coherence Scale. In the Bland‒Altman plot, the means of both scales (BDI-I and SOC-13) are plotted on the x-axis, while the difference between the two scales is plotted on the y-axis

Full size image

Network models of the scales (BDI-I and SOC-13)

The BDI-I scale network (see Fig. 2A) has an excellent stability index of 0.82. Due to the LASSO approach, low threshold associations were reduced to zero. Thus, the network presented is restricted to the salient associations. In the BDI-I, we can identify several overlapping communities of nodes: one related to exhaustion and coping (items L, M, O; Q); another related to self-contempt (items C, E; F, K; G, H; N); one related to hopelessness (items A; B; D and I); and one related to bodily sensations (items J; N P; R; S). The nodes had a mean strength of 0.74 (0.17); the node with the highest strength was node D (dissatisfaction). The nodes had a mean closeness of 0.88 (0.07); node D (dissatisfaction) had the highest closeness. The nodes had a mean betweenness of 0.45 (0.23), with node N (body image) having the highest betweenness. The nodes had an expected influence of 0.65 (0.23), with nodes B (pessimism) and D (dissatisfaction) having the highest influence within the network. The BDI-I nodes had a bridge influence of 0.22 (0.13), with node A (Sadness) having the highest influence on the SOC-13 network (for further details, see Fig. 2B).

Fig. 2
figure 2

Network analysis model (A) and centrality indices (B) of the Beck Depression Inventory I (BDI-I). Within the graphical representation, edges are the lines between the nodes (items) representing regularized partial correlations, which help estimate the relationship between two variables while controlling for all other variables. A line (edge) indicates an association between variables; the absence of a line indicates no association. The blue lines represent positive associations, while the red lines represent negative associations. The wider and more saturated an edge is represented, the stronger the association (Color figure online)

Full size image

The SOC-13 scale network (see Fig. 3A) has an excellent stability index of 0.79. Due to the LASSO approach, low threshold associations were reduced to zero. Thus, the network presented is restricted to the salient associations. Network, we could see four communities of nodes: one regarding trust in social relationships (Items 2 and 3); one concerning meaningfulness (Items 1, 4, 7, and 12); one concerning comprehensiveness (Items 5, 6, and 10); and one concerning manageability (Items 8, 9, 11, and 13). The nodes had a mean strength of 0.83 (0.16); the nodes with the highest strength were 06 (helplessness), 10 (bad luck), 12 (meaningless), and 13 (emotional control). The nodes had a mean closeness of 0.93 (0.07); again, nodes 06 (helplessness), 10 (bad luck), 12 (meaningless), and 13 (emotional control) had the highest closeness. The nodes had a mean betweenness of 0.56 (0.22), with node 12 (meaningless) having the highest betweenness. The nodes had an expected influence of 0.78 (0.21), with nodes 06 (helplessness), 10 (bad luck), 12 (meaningless), and 13 (emotional control) within the network. The SOC-13 nodes had a bridge influence of 0.35 (0.22), with node 01 (unattached) having the highest influence on the BDI-I network (for further details, see Fig. 3B).

Fig. 3
figure 3

Network analysis model (A) and centrality indices (B) of the Sense of Coherence Scale (SOC-13). Within the graphical representation, edges are the lines between the nodes (items) representing regularized partial correlations, which help estimate the relationship between two variables while controlling for all other variables. A line (edge) indicates an association between variables; the absence of a line indicates no association. The blue lines represent positive associations, while the red lines represent negative associations. The wider and more saturated an edge is represented, the stronger the association (Color figure online)

Full size image

Figure 4 displays the final network model of the BDI-I and SOC-13 scales with all items (i.e., 34 nodes). The network stability index was 0.78 (i.e., the maximum proportion of patients who could be dropped and still retained a correlation over 0.70 with the original estimate in 95% of the samples). In the joint BDI-I and SOC-13 network, depressive symptoms clustered around the SOC communities (see Fig. 4).

Fig. 4
figure 4

Network analysis model of the Beck Depression Inventory (BDI-I) and the Sense of Coherence Scale (SOC-13). BDI-I Items: A: Sadness; B: Pessimism, C: Failure; D: Dissatisfaction; E: Guilt; F: Punishment; G: Self-Dislike; H: Self-Accusations; I: Suicidal Ideas; J: Crying; K: Irritability; L: Social Withdrawal; M: Indecisiveness; N: Body Image; O: Work Difficulty; P: Insomnia; Q: Fatigability; R: Appetite; S: Weight, T: Somatic; U: Libido. SOC-13 Items: 01: Unattached; 02: Betrayal; 03: Disappointment; 04: Purposeless; 05: Injustice; 06: Helplessness; 07: Tediousness; 08: Confusion; 09: Suppression; 10: Bad Luck; 11: Appraisal; 12: Meaningless; 13: Emotional Control. Within the graphical representation, edges are the lines between the nodes (items) representing regularized partial correlations, which help estimate the relationship between two variables while controlling for all other variables. A line (edge) indicates an association between variables; the absence of a line indicates no association. The blue lines represent positive associations, while the red lines represent negative associations. The wider and more saturated an edge is represented, the stronger the association (Color figure online)

Full size image

Discussion

Our analysis revealed a strong (inverse) correlation between the strength of the sense of coherence measured by the SOC and symptoms/severity of depression measured by the BDI-I. Our results confirm previous findings reporting that clinically depressed persons tend to have a lower sense of coherence (Carstens & Spangenberg, 1997; Valimaki et al., 2009). The novel network analysis approach identified detailed relationship dynamics between symptoms, communities, and domains (Borsboom & Cramer, 2013).

According to our findings, relationships (i.e., social and personal) are pivotal in the SOC-13 network, and previously known factors (comprehensiveness, manageability, and meaningfulness) are involved (Antonovsky, 1993). The notion that supporting and reliable social relationships play a significant role in maintaining mental health is largely known (Cohen & Wills, 1985). It can be regarded as a determining factor for health and well-being (House et al., 1988). Our analysis’s results overlap to some extent with previous results and strengthen the notion that the sense of coherence is a complex system of cognitive, behavioral, and motivational dimensions that is structurally divisible but functionally indivisible (Portoghese, 2024).

Our results emphasize the importance of relationships in the salutogenic model in general and the sense of coherence in particular (Ejlertsson et al., 2013). Furthermore, in the context of overcoming depression, the therapeutic relationship is the most important single factor determining the efficacy and outcome of a psychological intervention for depression (Lambert & Barley, 2001; Steger & Kashdan, 2009). Nevertheless, the ability to commit to such a relationship might be severely impaired in a major depressive episode (Marshall & Harper-Jaques, 2008). Similarly, the ability to predict others’ behavior could be largely impaired due to symptoms of depression (Nestor et al., 2022).

By graphically representing the network model of SOC-13, we observed the nuclear position of social relationships in the network. Although the SOC-13 items reflecting relationships have relatively weak centrality indices, this apparent weakness might allow them to interact freely with the other nodes (Granovetter, 1973). This discrepancy from previous findings could be explained by the fact that, in contrast to our sample, the Sense of Coherence Scale was developed in nonclinical populations (Eriksson & Lindstrom, 2005), where an intact ability to relate can be expected. Moreover, the ability to maintain close relationships seems to be a critical factor in holocaust survivors (Armour, 2010). Furthermore, collective adversity brings up special emotional bonds (as beautifully shown in the Movie: “Marek Edelman… And There Was Love in the Ghetto”) (2024).

In the network of depressive symptoms, we observed several overlapping communities of symptoms. The first group had increased negative affect, and the second related to reduced positive affect. Together, they encompass the core symptom domains of a depressive disorder. A third community forms around the bodily sensations. A fourth community involved exhaustion and coping with obligations and daily activities. Finally, suicidal thoughts were related to several symptoms: fatigue, pessimism, social withdrawal, and dissatisfaction. Taken together, our results confirm previous results on depressive networks (Ma, 2022).

The cluster related to hopelessness reflects reduced positive affect—furthermore, the appearance of suicidal thoughts is closely related to feelings of hopelessness and despair. At the same time, the cluster related to an increased negative affect shows two cores. The first relates to a sense of rejection, including punishment and irritability; the second is self-contempt with failure and self-dislike. The passive form of self-dislike is closely related to bodily sensations. Thus, we believe that one reflects the agitated form of depression, while the other is the melancholic type (Koukopoulos & Koukopoulos, 1999).

Disability from depression results from fatigue, indecisiveness, and the consequent neglect of obligations (either work, educational, social, or personal) and recreational activities (Bruce, 2001). Furthermore, obligations and daily activities are considered a source of suffering, distress, or frustration, fuelling maladaptive thoughts and beliefs, especially in younger groups (Lee, 2018). It is thus not surprising that these factors are associated with social withdrawal and suicidal thoughts. Interventions that encourage the reconstitution of daily activities and self-appraisal through rewarding achievements effectively alleviate suffering in people with a depressive disorder (Cuijpers et al., 2011).

In general, bridge symptoms on the side of the SOC-13 Scale exerted a more significant influence than did those on the BDI-I side, with concern and neglect exercising the most influence on depressive symptoms. These patients are closely followed by symptoms of tediousness from obligations and daily activities. Thus, both reflect the loss of meaningfulness through depressive symptoms. We found that the feeling of loss and the lack of emotional control reflect the second line: the influence of manageability on depressive symptoms. Regarding items of the BDI-I, feelings of sadness influence the Sense of Coherence score; these symptoms are related to social withdrawal, thus spanning the core element of personal relationships.

Our study has several limitations that must be accounted for. First, we used solely self-assessment questionnaires without an external validation of the symptoms and diagnosis. The participants involved in the study actively exhibited help-seeking behavior and had at least one previous depressive episode; therefore, we can assume that they were familiar with the disorder and its symptoms. Furthermore, both questionnaires have proven validity and reliability (Carstens & Spangenberg, 1997; Välimäki et al., 2009).

In our study, we used the BDI-I questionnaire instead of newer versions. The BDI-I does not include all of the DSM-IV diagnostic criteria for depressive disorder (in contrast to the BDI-II). We consider the BDI-I to be better suited for classifying subtypes of depression, such as melancholia or agitated depression (Beck et al., 1996). The scores on the BDI-I we found in our sample corresponded to a moderate to severe major depressive disorder, thus compensating for the limitations mentioned above for the BDI-I score. Furthermore, the analysis of the seven items corresponding to the BDI-I screening scale yielded similar severity ranges of depression as did the BDI-I sum score.

Another limitation is the high rate of missing values; we consider this to be within the expectation for an online survey. Including only complete datasets strengthens the analysis by reducing imputation bias (Böwing-Schmalenbrock, 2011). Our analysed sample included only 181 individuals; although this sample might seem small compared to that of other studies, it is sufficient to perform a robust network analysis (Guadagnoli & Velicer, 1988). This was confirmed by the bootstrap analysis we performed to test network stability (Epskamp & Fried, 2018). Our study included mainly middle-aged adults, with a few participants older than 60 years. Therefore, we could not determine whether the Sense of Coherence increased with age, as previously reported (Eriksson & Lindström, 2006), and we could not determine whether age has a protective effect (Dezutter et al., 2013).

Our results revealed a strong correlation between the sense of coherence and the severity of reported symptoms of depression. Network analysis confirmed the pivotal role of personal relationships in the presence of a sense of coherence. We observed the central role of dissatisfaction in the depressive network. The relationship between the two networks relies on the meaningfulness and manageability of daily living. We believe that social relationships and social withdrawal constitute the core action targets that might alleviate the severity of depression on the one hand and strengthen the sense of coherence on the other hand. In conclusion, positive relationships are crucial for treating and preventing depression through the provision of support, understanding, and a sense of connection.