Is a sense of coherence associated with prolonged grief, depression, and satisfaction with life after bereavement? A longitudinal study

Abstract There is growing interest in psychological factors maintaining healthy functioning following adverse events. One such variable is a sense of coherence (SOC), an orientation to life comprising manageability, comprehensibility, and meaningfulness. Little research has examined the role of SOC in adjustment to bereavement. The present longitudinal study examined the role of SOC in recovery from loss, in a Danish sample (N = 221) of elderly spousally bereaved people. The aim was twofold. First, we aimed to establish the optimal measurement model of SOC, evaluating the fit of different factor solutions for the 29‐item SOC‐29 scale and 13‐item SOC‐13 scale, using confirmatory factor analysis. Second, we sought to examine associations of emerging SOC factors with symptoms levels of prolonged grief disorder (PGD) and depression, and with satisfaction with life, assessed concurrently (at 6 months post‐loss) and at two consecutive time points, 13 and 18 months post‐loss. Results showed that the three‐factor model of the SOC‐13 (with distinct manageability, comprehensibility, and meaningfulness factors) provided a good fit to our data. With respect to our second aim, analyses showed that the three SOC factors were associated with concurrently assessed PGD, depression, and satisfaction with life. In the analyses predicting outcomes at Wave 2 and Wave 3, meaningfulness (but not manageability and comprehensibility) predicted some of the outcomes, above and beyond baseline scores of the outcomes. Findings suggest that meaningfulness may increase healthy and attenuate unhealthy responses to loss. Helping bereaved people to experience life's demands as worthy of investment and engagement is likely an important target for bereavement care.

of elderly spousally bereaved people. The aim was twofold. First, we aimed to establish the optimal measurement model of SOC, evaluating the fit of different factor solutions for the 29-item SOC-29 scale and 13-item SOC-13 scale, using confirmatory factor analysis. Second, we sought to examine associations of emerging SOC factors with symptoms levels of prolonged grief disorder (PGD) and depression, and with satisfaction with life, assessed concurrently (at 6 months post-loss) and at two consecutive time points, 13 and 18 months post-loss. Results showed that the three-factor model of the SOC-13 (with distinct manageability, comprehensibility, and meaningfulness factors) provided a good fit to our data. With respect to our second aim, analyses showed that the three SOC factors were associated with concurrently assessed PGD, depression, and satisfaction with life. In the analyses predicting outcomes at Wave 2 and Wave 3, meaningfulness (but not manageability and comprehensibility) predicted some of the outcomes, above and beyond baseline scores of the outcomes.
Findings suggest that meaningfulness may increase healthy and attenuate unhealthy responses to loss. Helping bereaved people to experience life's demands as worthy of investment and engagement is likely an important target for bereavement care. Research in the field of loss and trauma has strongly focused on identifying risk factors and psychological processes negatively influencing adjustment. Over the last years, there is growing interest in protective factors that maintain healthy functioning and boost adjustment following such events. One such variable is a sense of coherence (SOC), a concept central to the salutogenic theory of health (Antonovsky, 1979(Antonovsky, , 1987. Briefly, this theory postulates that health and disease are not dichotomous constructs. Instead, over the course of life, individuals move back and forth on a continuum between health and illness (or ease and disease), affected by positive and negative life experiences and resources available to deal with these experiences. SOC is one key concept within this framework and one key resource that confers protection to adversity. SOC is an orientation to life encompassing manageability (the sense of having capacities and resources to deal with challenges), comprehensibility (the perception that internal and external experiences and situations are understandable and can be structured), and meaningfulness (the basic sense that life's demands are worthy of investment and engagement). To measure SOC, Antonovsky (1979Antonovsky ( , 1987 developed the 29-item SOC scale  including items for all three theoretically derived SOC dimensions and a shortened version including a selection of 13 items (the SOC-13). Both measures have been widely examined in different languages and samples. Studies have quite consistently supported the reliability and construct validity of the SOC-13 and SOC-29, but have left unclarity about the distinctiveness of the elements of manageability, comprehensibility, and meaningfulness (see, e.g., Antonovsky, 1993;Eriksson & Lindström, 2005;Eriksson & Mittelmark, 2016).
There is a growing evidence-base supporting that elevated SOC is associated with increased positive outcomes, including general wellbeing (Nilsson et al., 2010) and satisfaction with life (Moksnes et al., 2013) and decreased negative outcomes, including depression and anxiety (Flannery & Flannery, 1990). In addition, cross-sectional research showed that increased SOC is associated with decreased symptom levels of posttraumatic stress disorder (PTSD) following different types of events (Schäfer et al., 2019). In longitudinal studies, people with higher SOC have been found to report lower posttraumatic stress following accidental injuries (Schnyder et al., 2008) and traumatic pregnancy loss (Engelhard et al., 2003).
It is conceivable that greater SOC may foster recovery from the death of a loved one. Manageability may help mourners to draw on available resources after their loss. Comprehensibility may fuel confidence that the consequences of the loss are understandable and controllable. Meaningfulness may strengthen the ability to commit to what is needed to adapt to the loss. To date, only few studies examined associations of SOC with psychological reactions to bereavement. Cross-sectional research showed that higher SOC was associated with decreased grief among women who suffered perinatal bereavement (Uren & Wastell, 2002). In a heterogeneous sample of bereaved people, Bachem and Maercker (2016) examined associations of SOC with symptoms of prolonged grief disorder (PGD)-a disorder encompassing symptoms of separation distress and other intense grief reaction (e.g., difficulties accepting the loss, anger) present to a disabling and distressing degree (APA, 2022). They found increased SOC to be associated with lower symptom-levels of PGD. Xiu et al. (2016) found similar associations between SOC and PGD in Swiss bereaved parents, but not Chinese bereaved parents. In people who lost loved ones to suicide, higher SOC scores were positively associated with healthy grief (assessed with the Texas Revised Inventory of Grief; Faschingbauer et al., 1977) and negatively with depression (Scharer & Hibberd, 2020) There is still a need to further our understanding of SOC in recovery from loss. For instance, although different studies applied either the SOC-13 or SOC-29 to bereaved samples, no studies have systematically investigated the factor structure of these measures among mourners. It is, therefore, uncertain if, among people exposed to deaths of loved ones, SOC encompasses three distinguishable factors or can better be understood as a unitary concept. This is especially relevant considering that the factor structures of the SOC-13 and SOC-29 have been debated (Grevenstein et al., 2016;Grevenstein & Bluemke, 2021). Furthermore, to our knowledge, apart from Wen et al.'s (2021) study, no studies have examined the degree to which SOC is longitudinally associated with responses to bereavement.
Moreover, it is important to establish whether their findings in a nonwestern group generalize to western mourners. Relevant too is to examine associations of SOC with different responses to loss, including healthy (e.g., satisfaction with life) and unhealthy responses (e.g., PGD and depression). Given that SOC bears relevance to understanding individual differences in responses to bereavement and has hardly been studied, it is imperative to further knowledge about SOC in grief. Such knowledge may guide future interventions with bereaved people and enhance our general understanding of grief.
The current study used data from a Danish research project to expand knowledge on the role of SOC in recovery from loss. In that project, elderly spousally bereaved people completed self-report measures at 2, 6, 13, 18, and 48 months post-loss (O'Connor, 2010a(O'Connor, , 2010bO'Connor et al., 2015). The present study was based on data obtained at 6 months post-loss (referred to as Wave 1, W1), 13 months post-loss (Wave 2, W2), and 18 months post-loss (Wave 3, W3). The study aim was twofold. First, we aimed to establish the optimal measurement of SOC from our data. Taking into account previous research investigating the factorial validity of SOC

Key Practitioner Message
• Among elderly spousally bereaved people, a sense of coherence was found to encompass three distinguishable elements of manageability, comprehensibility, and meaningfulness.
• The three elements of a sense of coherence were associated with concurrently assessed reactions to loss.
• Baseline increased meaningfulness (but not manageability and comprehensibility) predicted decreased unhealthy and increased healthy bereavement-responses later in time.
• Helping bereaved people to experience life's demands as worthy of investment and engagement and empowering them to identify resources needed to adjust are important targets for bereavement care.
(e.g., Eriksson & Mittelmark, 2016), we compared different factormodels of the SOC-29 and the SOC-13. Specifically, for the SOC-29, we compared the fit of a one-factor model and a three-factor model, with items representing manageability, comprehensibility, and meaningfulness forming one or three factors, respectively. For the SOC-13, we evaluated the fit of a one-factor model, a three-factor model (with distinct factors for manageability, comprehensibility, and meaningfulness), plus a two-factor model, with items for manageability and comprehensibility forming one factor. Grevenstein et al. (2016) found the two-factor model of the SOC-13 to explain more variance in psychological distress than one-factor and three-factor models; this accords with the notion that meaningfulness is a distinct, more basic element of SOC, driving the motivation to mobilize resources to comprehend and manage life stressors. Considering that all three models have yielded at least some support in prior research (Grevenstein & Bluemke, 2021), we decided to compare their fit in the present work.
Our second aim was to examine associations of emerging SOC factors with indices of PGD, depression, and satisfaction with life, assessed concurrently (at 6 months post-loss, W1) and the association of SOC at W1 with these outcomes at W2 and W3 (13 and 18 months post-loss). We focused on negative and positive outcomes, considering that adjustment to loss involves both mitigating negative emotions and maintaining positive psychological functioning. Assuming that the analyses would show that, in our sample, SOC consists of multiple, distinguishable components, we intended to investigate both the association of the distinct components with the outcomes as well as the association of the SOC factors with outcomes when controlling for the shared variance between SOC factors. Considering the lack of knowledge on the role of SOC in recovery from loss, examining both the distinct and combined impact of SOC factors on bereavement outcomes was deemed both clinically and theoretically relevant, by elucidating potential underlying mechanisms of problems in recovery from loss and targets for interventions.
We had no hypotheses about the factor structure of the SOC items, considering that prior work has supported different factorial structures. With respect to our second aim, we anticipated that SOC would be negatively related with PGD and depression symptoms and positively with satisfaction with life. Again, because of the lack of research in bereaved samples, we had no specific hypotheses about the strengths of these relationships nor about the relative importance of SOC factors emerging from our analyses. were male (gender was unknown for n = 3). They had a mean of 8.19 years of public schooling (SD = 1.62; range 5-14, n = 12 missing). All had lost a spouse or partner. As noted, for all participants, the time elapsed since the loss was 6 months (at W1). Participants had cohabited with their partner 45.34 (SD = 10.73) years before the partner's death, on average. They had 2.61 (SD = 1.51, range 0-9, n = 6 missing) children on average; n = 190 (86%, n = 9 missing) had faced a period of the spouse's illness before the partner died. The self-reported cause of death was cancer in 91 (41.2%) cases, cardio vascular disease in 34 (15.4%) cases, dementia in five (2.3%) cases, an accident in three (1.4%) cases, diabetes in three (1.4%) cases, other causes (sometimes including combined causes, e.g., diabetes and cardio vascular) in 74 (33.5%) cases, and unknown in 11 (5%) cases.

| Measures
Data were gathered using self-report questionnaires. The first part contained questions on socio-demographics and loss-related variables.  Table 1.
An adapted version of the Inventory of Complicated Grief (Prigerson et al., 1995) was used to measure PGD symptoms. It includes 15 symptoms of PGD and other putative markers of disordered grief. Participants rated the frequency of symptoms in the prior month, on 5-point Likert scales with anchors 1 = none/almost never to 5 = a lot/all the time. The summed score was used as an index of PGD symptom severity. The ICG has sound psychometric properties (Prigerson et al., 1995). 1 The Beck Depression Inventory (BDI; Beck et al., 1988) was used to measure depression. It includes 21 sets of four statements representing depressive symptoms at increasing levels of severity. Statements are rated on a 0-3 scale (with higher scores representing more severe depression) and summed to get an index of depression. The measure has strong psychometric properties (Beck et al., 1988).
The Satisfaction with Life Scale is a five-item measure that was constructed by Diener et al. (1985) to measure the global judgement about one's life satisfaction. Participants are asked to rate how much they agree with items, on a 7-point scale, with anchors 1 = strongly disagree to 7 = strongly agree. The measure has good psychometric properties.
The SOC-29 was used to measure three domains of comprehensibility, manageability, and meaningfulness. The SOC-29 includes all 13 items from the SOC-13. Both versions were originally developed by Antonovsky (1987). Items are scaled along a dimensional scale, ranging from 1 to 7, with anchoring phrases at both extremes. Items of the SOC-13 are shown in Table S1. An example of an item on meaningfulness is "Do you have the feeling that you don't really care about what goes on around you?" rated on a scale from 1 = very seldom or never to 7 = very often (with scores being reversed before T A B L E 1 Zero order correlations, means, and standard deviations  being summed with other item scores). As noted above, evidence supported psychometric properties of the SOC-29 and SOC-13 but the factor structure of both scales has been debated (Eriksson & Mittelmark, 2016). One-factor, two-factor, and three-factor models evaluated in this study have all received support in different studies (Grevenstein & Bleumke, 2017).

| Statistical analyses
Our statistical analyses included four consecutive steps. First, to address our first aim, we examined the factor structure of SOC items using confirmatory factor analysis (CFA) implemented in MPlus (Muthén & Muthén, 1998-2012. Using data from the W1 sample and standardized root mean square residual (SRMR) with values <.08 indicating acceptable fit (Hu & Bentler, 1999). To compare the fit of the models, we considered the Akaike information criterion (AIC), Bayesian information criterion (BIC), and sample size adjusted Bayesian information criterion (SS-BIC). We used χ 2 difference tests to test the differences between nested models, using the Satorra-Bentler scaling correction (Satorra & Bentler, 2001).
As a second step, we calculated zero-order correlations between indices of SOC (at W1) and PGD, depression, and functional impairment (at W1-W3). In addition, we examined whether scores on these indices differed as a function of the socio-demographic and lossrelated variables we assessed, using analysis of variance and Pearson correlations. As a third step, we examined whether participants who discontinued participation at W2 differed from those continuing par- For items included in the SOC-13, the one-factor model did not fit well. The two-factor model fit the data better than the one-factor model (Δχ 2 = 22.15, Δdf = 1, p < .001), but also did not provide good fit estimates (e.g., CFI and TLI were too low, RMSEA was too high).
Modification indices pointed out that errors of the same two items that had correlated errors in the SOC-29 were correlated in the SOC-13. In addition, errors terms of one of the manageability items (Have you felt like a loser?) and one of the comprehensibility (Have you overestimated/underestimated the importance of things happening?) were also correlated; this was also potentially due to content overlap, considering that these items both represented critical self-evaluation.
The SOC-13 three-factor model fit the data better than the onefactor model (Δχ 2 = 22.79, Δdf = 3, p < .001), but did not have good fit estimates. Again, modification indices pointed at correlations between the errors of the two aforementioned item pairs. The adjusted three-factor model, with correlations between these error terms, fit better than the one-factor model (Δχ 2 = 93.43, Δdf = 5, p < .001) and the three-factor model (Δχ 2 = 84.03, Δdf = 2, p < .001) and had good fit estimates ( Table 2).
The two-factor model with correlated errors and the three-factor model with correlated errors had similar fit estimates andaccordingly-the χ 2 difference test was not significant (Δχ 2 = 3.39, Δdf = 2, p = .184). Because research on SOC in grief is sparse, it was deemed informative to retain the three-factor model, distinguishing the three factors put forth by Antonovsky (1979), in subsequent analyses. Items and factor loading of the SOC-13 are shown in Table S1.
Correlations of the comprehensibility factor with the manageability and meaningfulness factors were .91 and .88, respectively; the correlation between the manageability and meaningfulness factors was .72 (ps < .001).

| Differences between people continuing versus discontinuing participation
Of all 221 participants at Wave 1, the 176 who still participated at Error terms of one manageability item (Has it happened that people whom you counted on disappointed you?) and one comprehensibility item (Has it happened in the past that you were surprised by the behaviour of people whom you thought you knew well?) were correlated. b Error terms of the same manageability item and comprehensibility item were correlated; error terms of one manageability item (Many people-even those with a strong character-sometimes feel like sad sacks [losers] in certain situations. How often have you felt this way in the past?) and one comprehensibility item (When something happened, have you generally found that: [you overestimated or underestimated its importanceyou saw things in the right proportion]) were also correlated.
who still participated at Wave 3 did not differ from the 52 discontinuing their participation on all variables assessed at Wave 1. Table 3 summarizes outcomes of the regression analyses regarding the concurrent associations between SOC and outcomes. All three SOC factors were inversely associated with PGD severity in distinct analyses (Models 1-3). Comprehensibility and meaningfulness were significantly correlated with W1 PGD when all SOC factors were considered simultaneously (Model 4). Similar outcomes were found in the regression analyses with depression severity as dependent variable.

| Associations of SOC factors with PGD, depression, and satisfaction with life at W1-W3
As for satisfaction with life, all three SOC factors were correlated with satisfaction with life in distinct analyses (Models 1-3); meaningfulness was the only SOC factor associated with this outcome, when all factors were considered simultaneously (Model 4).  Abbreviations: DV, dependent variable; IV, independent variable. *p < .05. **p < .01. ***p < .001.

| Additional regression analyses
We examined if the outcomes of the regressions summarized in Tables 2-4 changed when we controlled for the three loss-related variables that were associated with some of the outcomes, namely: number of children, dichotomized cause (cancer vs. other cause), and having had a period of illness before the death (yes/no). There were two minor changes: The association of comprehensibility with PGD at W1 was no longer significant in the model that included all three SOC factors (Model 4; β = À.15, p = .14). The association of meaningfulness with W3 PGD, controlling W1 PGD at the other SOC factors, passed the threshold for significance (β = À.14, p < .05). All other significant findings remained significant and all other nonsignificant findings remained nonsignificant. Thus, associations of SOC factors with outcomes reported in Tables 2-4 were not moderated by the three loss-related variables associated with some of the outcomes (number of children, dichotomized cause of loss, and a period of illness preceding the death).
As noted, the two-factor model (with one factor including the comprehensibility and manageability items and a second factor spanning meaningfulness items) fit the data equally well as the threefactor model ( Table 2). For exploratory reasons, we reran all the regression analyses predicting PGD, depression, and satisfaction with life, but instead of including the three SOC factors, we now included two SOC factors (factor 1: comprehensibility/manageability, factor 2: meaningfulness). Outcomes are summarized in Tables S2-S4. Outcomes were generally consistent with outcomes of the regressions with the three SOC factors. In all analyses where the two distinguished comprehensibility and manageability factors were significantly associated with the dependent variable (e.g., satisfaction with life at W2 and W3), the combined factor (spanning comprehensibility and manageability items) predicted the outcome as well. In all analyses where the distinct factors did not predict outcomes, the combined also did not. So, it appears that combining the manageability and comprehensibility items did not enhance explanatory power of the analyses.
T A B L E 4 Summary of regression analyses with comprehensibility, manageability, and meaningfulness predicting prolonged grief, depression, and satisfaction with life at W2

| DISCUSSION
A gradually growing evidence base shows that a sense of coherence (SOC), encompassing a sense that one's internal and external world are comprehensible, manageable, and meaningful, offers protection against the negative impact of stressful life events (Engelhard et al., 2003;Schäfer et al., 2019;Schnyder et al., 2008). So far, very few studies have examined the role of SOC in responses to bereavement. The current study used data from a large Danish multi-wave study, among elderly spousally bereaved individuals (e.g., O'Connor, 2010a(e.g., O'Connor, , 2010bO'Connor et al., 2015), to further knowledge about this issue.
A first aim was to identify the optimal measurement of SOC with the data available for the current investigation. Therefore, we first examined the factor structure of different measurement models of the SOC-29 and the SOC-13. For the SOC-29, one-factor and threefactor models did not fit the data well. A three-factor model with correlated errors of two similarly worded items approached (but did not pass) the threshold for acceptable model fit. In our examination of the factor structure of the SOC-13, the two-factor and three-factor models fit the data better than the one-factor model but did not pass the threshold for good fit. Modification indices revealed that error terms of two items referring to critical views on other people and two other items referring to a critical self-evaluation were correlated.
When allowing residual correlations for these two item-pairs, both the two-factor and three-factor models fit the data well; the fit of these models did not differ significantly.
The findings indicate that, in our sample, the more parsimonious set of 13 items yielded a better measurement of SOC than the combination of 29 items. Our findings are generally consistent with research findings showing that the two-factor model represents a good model of the SOC-13 (Grevenstein & Bluemke, 2021) and other studies supporting the three-factor structure (see Bonacchi et al., 2012). That said, some studies have shown that the scale is adequately represented by a one-factor model (e.g., Antonovsky, 1993) or alternative three-factor models (e.g., Bachem & Maercker, 2016). Because studies T A B L E 5 Summary of regression analyses with sense of coherence predicting prolonged grief, depression, and satisfaction with life at Wave 3 Considering these findings, the following stands out. Meaningfulness consistently turned out to be the most important element of SOC in predicting consequences of bereavement. This is consistent with Antonovsky's (1987) original notion that meaningfulness has the strongest bearing on people's abilities to manage adversities. Apparently, the sense that consequences of one's loss-experience are challenges that are worthy to investigate time and energy in-implicated in elevated meaningfulness-has an impact on adjustment to loss. The sense that one's external and internal worlds are understandable and that one has the resources needed to manage the consequences of the loss-implicated in elevated comprehensibility and manageabilityare seemingly less important. The findings are reminiscent of prior theorizing and research pointing at the importance of meaning attribution and meaning making in grief (Neimeyer et al., 2014;Smid, 2020). The present findings add to this prior work by showing that, apart from the ability to assign meaning to a specific loss-experience, the broader life orientation that stressors are worth facing attenuates the emotional pain of bereavement. Notable too is that increased meaningfulness was associated with increased satisfaction with life and, albeit to a lower degree, decreased levels of PGD and depression. This is consistent with prior theorizing and research indicating that elevated SOC both acts as buffer between negative events and distress and directly inflates positive outcomes in the face of these events (Schäfer et al., 2019). Following spousal loss, higher meaningfulness likely encourages people to continue investing energy in activities that were fulfilling before the loss and explore options for new activities. This type of behaviour is likely to attenuate sadness, yearning, and other grief reactions and to maintain satisfaction with the life.
Strengths of this study include its longitudinal design, multiple follow-up assessments, and relatively homogeneous sample. The study also has some limitations. First, the use of self-report measures to assess all the variables may have inflated associations between variables because of shared method variance. Second, although the ICG is still commonly used, it was constructed way before recent concep- supports that SOC positively affects recovery from loss, some practical implications for supporting bereaved people may be considered. For instance, helping people to see the world as more comprehensible, manageable, and-particularly-more meaningful, and empowering them to identify resources needed to adjust may be important targets for interventions and health promotion activities (Super et al., 2016). Psychological interventions tailored specifically for recognizing and increasing a sense of meaning with life would, thus, be called for. Inspiration for these types of interventions may be gained from the perspective of positive psychology (Seligman & Csikszentmihalyi, 2000), narrative psychology (Nolen-Hoeksema, 2000), and third wave cognitive behavioural psychology such as compassion focused therapy (Gilbert, 2014) or mindfulness based cognitive therapy (Segal et al., 2013).

CONFLICT OF INTEREST
The authors declare to have no conflict of interest.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author (PAB) upon reasonable request.