Background
Bereavement is a universal human experience. While most people react to a loss with intense pain and may even develop an increased risk of mental or physical health problems [
1], a majority adjust successfully to this stressful life-event without professional help. A minority of bereaved people, however, suffer from persistent grief symptoms of clinical relevance that are accompanied by functional impairment. Both new diagnostic classification systems of mental health conditions acknowledge the syndrome. The Diagnostic and Statistical Manual for Mental Disorders in its 5th edition (DSM 5) classifies it as ‘Persistent Complex Bereavement Disorder’ (PCBD) and considers it a condition for further study [
2]. The International Classification of Diseases in its 11th edition (ICD-11) will probably include ‘Prolonged Grief Disorder’ (PGD) as a stress-related disorder [
3,
4]. Concerning symptom duration, ICD 11 may allow a diagnosis as early as six months after the loss occurred, whereas DSM 5 requires symptoms to last twelve months. A recent meta-analysis estimates the pooled prevalence of PGD after bereavement to be 9.8% [
5]. The present article will use the well-established term ‘complicated grief’ to refer to clinically relevant grief symptomatology, since the present assessment of grief symptoms follows the Inventory of Complicated Grief [
6]. This also allows for drawing on existing research of grief rumination and bereavement outcome.
Given the severe distress that may be experienced after bereavement, a thorough understanding of the malleable factors that may contribute to the development and persistence of mental health problems is important. Thought processes are both malleable (for a review: [
7]) and are assumed to play an important role in the potential transition from ‘normal’ to complicated grief and its maintenance [
8]. Special attention has been paid to trans-diagnostic thought processes such as repetitive thinking, i.e. the ‘process of thinking attentively, repetitively or frequently about one’s self and one’s world’ ([
9], p. 909). Repetitive thinking about the deceased, the loss and its circumstances and consequences seems inherent to the acute grieving process [
10]. Some forms of repetitive thinking, however, such as rumination, have been associated with poor bereavement outcome, both concurrently and prospectively (cf. [
11‐
13]).
Two types of rumination have been studied in some detail in adjustment to loss. The first investigations in this area were conducted in the mid-nineties. They focused on clarifying the role of depressive rumination after bereavement. Depressive rumination was defined as repetitively and passively focusing on depressive symptoms and on their possible causes and consequences [
11]. A frequently used theory to understand the effects of depressive rumination on psychopathological symptoms is the Response Styles Theory (RST). The RST proposes that depressive rumination fuels depression by increasing the accessibility of negative thought content, impairing instrumental behaviour and problem solving, and driving away social support [
11,
14]. In an attempt to differentiate adaptive from maladaptive forms of depressive rumination (cf. [
15]) and to minimise the content overlap of the assessment of rumination with that of symptoms of depression, two sub-facets of depressive rumination were introduced, namely brooding and reflection [
16]. ‘Brooding’ implies a passive comparison of the aversive current situation with some unachieved standard, and ‘reflection’ indicates actively focusing inward to engage in cognitive problem solving in order to overcome depressive symptoms. Longitudinally, brooding has been associated with more depressive symptoms, whereas reflection seems to be associated with less depressive symptoms [
16]. Concerning adaptation to bereavement, RST conceptualises rumination as a confrontation strategy as it entails thinking repeatedly about post-loss emotions. Previous research indicates that all three constructs (i.e. depressive rumination, brooding, and reflection) are associated with psychopathological symptoms after bereavement [
17‐
23]. However, it also suggests that another type of rumination, namely grief rumination, is potentially more predictive of mental health problems in adjustment to bereavement, consistently explaining more variance in post-loss symptoms of depression, posttraumatic stress and complicated grief concurrently and longitudinally ([
20,
24,
25]; for a review: [
13]).
In contrast to depressive rumination, grief rumination is not limited to analysing feelings of depression, as negative post-loss emotions are not restricted to sadness or helplessness but may also entail many other emotions including yearning, anger or irritability [
1]. Thus, rumination after loss likely focuses on a wider array of loss-related feelings [
19]. Additionally, typical topics of rumination will differ in grief and depression. Similarly to rumination after traumatic events [
26,
27], grief rumination may focus strongly on reconciling the event with previously held beliefs about the meaningfulness or fairness of the world (i.e., thinking about why the event happened and the injustice of the loss), and counterfactual thinking (i.e., thinking about possible courses of action that might have prevented the event’s occurrence).
A model that is currently often used to understand the negative consequences of grief rumination is the Rumination as Avoidance Hypothesis (RAH [
10]). The RAH conceptualises rumination as an avoidance strategy because when ruminating about, for example, alternative outcomes of the situation (counterfactual thinking), one may avoid confronting the reality and permanence of the loss. Rumination would thus impede acceptance of the loss and hinder its contextualisation within existing autobiographical knowledge [
8]. Previous research suggests that experiential avoidance and thought suppression longitudinally mediate the relationship between grief rumination and symptoms of complicated grief [
25]. Experimental approaches have also corroborated the link between grief rumination and avoidance [
28,
29]. Additionally, grief rumination has been investigated in another longitudinal study of recently bereaved participants [
20]. In this sample, while simultaneously controlling for baseline symptom levels and other loss-related variables, grief rumination was a stronger predictor of later symptom levels of grief than was depressive rumination. This analysis also provided the first evidence of a distinction between adaptive and possibly maladaptive facets of grief rumination. Rumination about emotional reactions to the loss was regarded as potentially adaptive, since it was longitudinally associated with lower symptom levels. Rumination about the injustice of the loss was considered potentially maladaptive, since it was longitudinally associated with higher symptom levels.
Given grief rumination’s potential theoretical and clinical relevance, the Utrecht Grief Rumination Scale (UGRS) was recently developed to specifically assess grief rumination [
19]. The UGRS is based on theories of depressive rumination [
16,
30], trauma-related rumination [
26], and grief-relevant rumination [
8]. It captures five typical themes of post-loss rumination: (1) personal emotional reactions to the loss, (2) injustice of the death, (3) counterfactual thoughts about the circumstances of the death, (4) meaning and consequences of the loss, and (5) the reactions of others to the loss. It was originally published in Dutch [
24]; an English version has been developed and its cross-cultural equivalence confirmed [
19]. In confirmatory factor analyses of the data of the Dutch and British samples, a single-level factor structure with five correlated factors provided the best model fit, even though a hierarchical model with a second-order factor performed almost equally as well [
19]. In English and Dutch samples, the UGRS has demonstrated very good psychometric properties. It showed excellent internal consistency (α = .90) and, as a first indication of its validity, the UGRS contributed to the prediction of depression, posttraumatic stress and complicated grief over and above demographic and loss-related variables and other measures of rumination [
19,
24].
Clearly, more international research is needed to distinguish potentially adaptive and maladaptive facets of rumination at different time points in the grieving process. We also need to elucidate the pathways via which rumination contributes to the development and maintenance of mental health problems and, specifically, complicated grief. As prerequisite to this long-term goal, the present study aimed to develop a German version of the UGRS, to investigate its psychometric properties (e.g. reliability, item-correlations, factor structure), and to test its concurrent and discriminant validity by examining associations between the UGRS, reflection and brooding, and symptoms of anxiety, depression and complicated grief.
Discussion
This is the first study to present and validate a German version of the UGRS. In a sample of recently bereaved participants, the German UGRS was shown to have an identical factor structure to the original UGRS and the UGRS and its subscales demonstrated very good item properties, internal consistency and convergent, divergent, concurrent and discriminant validity.
The reliability of the UGRS in this study is high and comparable to those found for the English and Dutch versions. The internal consistencies of the subscales are – considering their extreme brevity – also very good. All item difficulties are in the medium range, which is recommended for maximum discriminatory power. Overall, the item-whole correlations with the total scale were high, with the exception of item 10, which was in the medium range.
Concerning the factorial structure, our analyses mirror the result of the original analyses in English and Dutch samples. They demonstrated that, even though the lower-order model with five correlated factors provided the best fit, a higher-order model with Grief Rumination as a second-order latent construct was only marginally outperformed [
19]. Importantly, the fit indices of both competing models differed even less in our sample compared to the original analyses. Both absolute and comparative fit indices award a very small advantage to the lower-order model, albeit only marginally. Parsimony is an important aspect in evaluating model fit. Of note, the parsimony-adjusted fit index PCFI favours the hierarchical model. This argument for conceptualising the UGRS as containing one higher-order factor dovetails with the theoretical viewpoint, because grief rumination was intended as a unidimensional construct in which the subscales each represent a recurrent theme of repetitive thinking. Naturally, the themes will vary among bereaved people and not all themes will occur equally often in bereaved persons. However, the latent process of grief-specific rumination is thought to represent the unifying construct underlying these themes. Given the only marginal superiority of a lower-order model in the confirmatory factor analysis results of the English and Dutch UGRS [
19] and our own data, it seems preferable to conceptualise the UGRS sum score as an indicator of grief-specific rumination.
The validity of the UGRS was supported further in our sample. Importantly, supporting the convergent and divergent validity of the UGRS, grief rumination was more closely associated with maladaptive types of ruminative thought (i.e. brooding) than more adaptive types (i.e. reflection). This replicates previous findings [
19] even to the point that the UGRS subscales Injustice and Counterfactuals demonstrated no significant association with reflection. Both the UGRS sum score and all subscales showed high correlations with complicated grief. This concurs with theoretical accounts that view rumination as a mechanism which perpetuates complicated grief [
8]. Associations between the UGRS and symptoms of anxiety and depression, which are often present concurrently in persons who suffer from complicated grief [
43], were significant, yet the level of the associations also differed significantly: grief rumination was more closely associated with disabling grief than with symptoms of depression or anxiety, providing a first indication of the UGRS’s discriminant validity. The regression analysis additionally demonstrated that the association between UGRS scores and symptoms of complicated grief holds even when simultaneously considering demographic and loss-related variables, depressive or anxiety symptoms, and other facets of rumination. The proportion of variance explained was high. Of note, UGRS scores explained incremental variance in ICG scores, over and above the aforementioned constructs. This speaks for the importance of grief-specific rumination as an independent construct.
Lastly, we investigated whether UGRS scores differed between participants with and without an increased likelihood of receiving a diagnosis of complicated grief. To this end, we classified participants according to the established symptom level cut-off in the ICG as candidates or non-candidates for complicated grief, applying different time criteria. The loss had to have happened at least six months ago (time criterion ICD-11) or twelve months ago (time criterion DSM 5). Irrespective of the time criterion, the UGRS scores were elevated strongly (Cohen’s d = 1.66 and d = 1.60) in both candidate groups when compared with the group with less likelihood of receiving a diagnosis of complicated grief. This speaks for the discriminant validity of the UGRS.
Several limitations must be borne in mind when interpreting the results. All data are based on online self-reporting and are cross-sectional in nature. The sample consisted predominantly of conjugally bereaved females, which mirrors the over-representation of this subgroup in most grief research [
44], but is not representative of the general bereaved population. While we have no reasons to assume that the associations under investigation are different for lower-educated people and men, future studies on rumination following loss should aim to recruit more representative samples. The findings refer to a convenience sample of bereaved persons, which allows comparison with previous UGRS psychometric research, but they require replication in a clinical sample of patients who suffer from complicated grief in order to ascertain that the associations remain stable across the complete range of disabling grief symptoms. Complicated grief is a diagnosis that is undergoing (re-) conceptualisations with the revisions of the classification systems. Even though the ICG is one of the best established and most-used instruments for assessing disabling grief symptoms, it is unclear to what extent the ICG and its established cut-offs concur with the diagnostic criteria of Persistent Complex Bereavement Disorder (DSM 5) or Prolonged Grief Disorder (ICD-11), which limits the generalisability of our results. Additionally, information on grief symptom levels in our sample was based on self-report: no clinical diagnosis can be made based solely on this information. Therefore, we have referred to our diagnostic categories based on the ICG cautiously as ‘candidates for complicated grief’ in order to account for this diagnostic uncertainty.