Introduction
There is evidence that, in a significant minority of bereaved children, adolescents and adults, acute grief reactions turn into chronic debilitating distress, blocking reestablishment of normal routines (e.g., Dowdney
2008; Melhem et al.
2007,
2011). This evidence has led to the inclusion of Persistent Complex Bereavement Disorder in the Appendix of the Diagnostic and Statistical Manual of mental disorders [DSM-5; American Psychiatric Association (APA)
2013; for discussions see Boelen and Prigerson
2012; Kaplow et al.
2012; Wakefield
2012]. Consistent with many recent studies (see Kaplow et al.
2012), we refer to this syndrome as prolonged grief disorder (PGD). PGD includes symptoms of separation distress, numbness, avoidance and detachment present to the point of functional impairment at least 6 months following the death of a loved one (Prigerson et al.
2009). To date most research on PGD has been done in adults, but there is a growing body of research done in children and adolescents. There is strong evidence that symptoms of PGD are distinct from symptoms of depression, posttraumatic stress disorder (PTSD) and other anxiety disorders (Boelen and Prigerson
2012; Lichtenthal et al.
2004; Prigerson et al.
2009; Shear et al.
2011; Spuij et al.
2012a,
b).
In order to understand why some children cope with loss more easily than others, it is important to identify variables that mediate the development and maintenance of PGD. Given the lack of effective interventions for bereaved youth (Currier et al.
2007), this knowledge can be very helpful for the development of interventions for bereaved children who fail to recover from loss. Recent cognitive behavioral conceptualizations of PGD propose that negative thinking and avoidance behaviors play a significant role in the development and maintenance of the disorder (Boelen et al.
2006; Maccallum and Bryant
2013). Empirical findings, mostly from adults, have confirmed that some bereaved individuals develop negative thinking that contributes to the development of PGD (Boelen and Lensvelt-Mulders
2005; Boelen et al.
2003). However, little is known about negative thoughts associated with emotional distress in bereaved children. Findings of a preliminary study with 30 adolescent girls, aged 13–18 years (Boelen and Spuij
2008), showed that different types of cognitions were significantly associated with symptom levels of PGD and depression; global negative thinking about life and self and catastrophic misinterpretations were most strongly linked with these symptoms. To our knowledge, no research is available about specific grief related negative thoughts in bereaved children under the age of 13. This is remarkable, as many children are confronted with loss (Harrinson and Harrington
2001). Understanding the cognitive profile that is associated with specific grief related psychopathology in children is important for the conceptualization and treatment of this psychopathology.
Despite the widespread idea that alleviation of mood and anxiety symptom requires changes in underlying maladaptive thinking patterns (cf. Beck
1967; Ellis and Grieger
1977; Meiser-Stedman et al.
2009; Treadwell and Kendall
1996), there is a scarcity of measures to assess negative thinking in children who suffer from depression, PTSD or anxiety. Moreover, there is no measure of negative thoughts involved in the development and maintenance of PGD. Examples of measures that do assess negative thinking in youth are the Automatic Thoughts Questionnaire (ATQ, Hollon and Kendall
1980; Kazdin
1990), and the Negative Affectivity Self Statement Questionnaire (NASSQ; Ronan et al.
1994). Campbell et al. (
2000) stated that the majority of childhood cognitive measures, including the ATQ, are limited by the fact that they are downward extensions of questionnaires developed for adults. Other measures, like the NASSQ, have different versions for children and adolescents, which is a limitation for e.g. research purposes. In light of these limitations, Schniering and Rapee (
2002) developed the Children’s Automatic Thoughts Scale (CATS), a measure specifically designed to assess negative thinking in children and adolescents. The authors first started with a pool of items that was generated from in-depth interviews with children and adolescents with various forms of emotional problems. Items relating to affect or behavior were removed, so only items that reflect cognition remained. The final measure consists of 40 items covering a broad range of negative automatic thoughts (e.g., “There is something very wrong with me”, “I look like an idiot”, “I’m worthless” and “Most people are against me”).
Given the lack of measures for negative thinking in children in general and more specific for bereaved children, our aim was to develop a measure for the latter group. Specifically, the current study sought to enhance knowledge on the role of negative thinking in PGD and other forms of emotional distress among children confronted with the death of a loved one. To this end, as described in more detail below, we constructed the so called Grief Cognitions Questionnaire for Children (GCQ-C) for bereaved children aged 8–18 years, drawing from earlier work on negative cognitions in adult PGD (Boelen and Lensvelt-Mulders
2005). Because of the limitations mentioned by Campbell et al. (
2000) about the usefulness of adult measures in children, we asked child psychologists, children and adolescents to help us with the development of this specific child version. We describe this process below. Furthermore, our aim was to evaluate several psychometric properties of this GCQ-C. Specifically, we studied the dimensionality, internal consistency, and temporal stability of the GCQ-C. In addition, we studied several predictions concerning the validity of the GCQ-C. With respect to the concurrent validity, we expected that stronger endorsement of negative cognitions tapped by the GCQ-C would be associated with increased symptom levels of PGD, depression, and PTSD as well as Internalizing and Externalizing Problems. In addition, we expected elevated negative thinking—tapped by the GCQ-C—to be associated with reduced
coping efficacy, defined as the participant’s own perceived capability to adjust to the loss (cf. Benight and Bandura
2004). With respect to convergent and divergent validity, we expected scores on the GCQ-C to be more strongly associated with cognitions associated with depressive and anxious states (tapped by the subscales social threat, physical threat and personal failure of the CATS; cf. Schniering and Rapee
2002), than cognitions associated with anger (tapped by the subscale hostile intent of the CATS; cf. Schniering and Lyneham
2007). In addition, it was predicted that, compared to the summed scores on the CATS (tapping cognitions related to depressive, anxious, and angry states not specifically linked with the loss participants had confronted), summed scores on the GCQ-C (tapping loss related negative thoughts) would be more strongly associated with the degree to which children and adolescents experienced impairments in functioning as a result of the loss. Finally, we explored the extent to which scores on both questionnaires varied as a function of demographic and loss related variables (e.g., time since loss and cause of death).
Results
Characteristics of the Study Samples
Table
1 shows background and loss related characteristics of both samples. As can be seen, most participants had lost a parent. Most losses were due to illness and most losses were experienced as being unexpected. Scores on the IPG-C (minimum 30, maximum 90) and the CDI are also shown in Table
1. Scores on the IPG-C did not differ between Sample 1 (
M = 52.24, SD = 12.14) and Sample 2 (
M = 52.69, SD = 11.99);
t(149) = −0.228,
p = 0.74). In comparison with reference groups from Timbremont et al. (
2008) scores on the CDI in both samples fell in the subclinical range.
Dimensionality
The exploratory factor analysis resulted in the emergence of four factors with eigenvalues greater than 1.00 (i.e., 11.76, 1.45, 1.19, 1.08). However, there were reasons to conclude that the GCQ-C items are best characterized as one factor. Firstly, the first factor explained almost 45 % of the variance in the GCQ-C with the second through fourth factor each adding very little the variance explained by the first factor. Secondly, inherent to that, the scree plot revealed a clear break after the first component. Thirdly, as shown in Table
2, in the one factor solution, all 20 items had factor-loadings ≥0.50. Finally, in the models with more than one factor, several items loaded highly on more than one factor and factors could not be interpreted in a meaningful way. Overall, the findings suggested that, within the present dataset, the GCQ-C items clustered together into one underlying dimension of “overall loss related negative thinking”.
Table 2
Abbreviated items of the GCQ-C and factor-loadings in the one-factor solutions
1 | Since s/he died, I think of myself as a weak person | 0.88 |
2 | Since s/he is dead, I am of no use to anyone anymore | 0.87 |
3 | Since s/he is dead, I think the world is bad | 0.61 |
4 | Since s/he is dead, I think the world is worthless | 0.86 |
5 | I think that others should pay attention to how I am doing | 0.50 |
6 | I don’t show others what I think and feel, because I am afraid that that only makes things worse for them | 0.68 |
7 | I should have seen to it that s/he would not have died | 0.75 |
8 | I blame myself for not having cared for him/her better than I did | 0.82 |
9 | I don’t think that I will ever feel better; I don’t feel confidence about the future without him/her | 0.88 |
10 | I think that the future will be no fun without him/her | 0.82 |
11 | As long as I am sad, I don’t have to let him/her go | 0.66 |
12 | I want to hold on to my sorrow for as long as possible | 0.60 |
13 | It is not nice toward him/her, when I will begin to feel less sad | 0.72 |
14 | I think that my feelings about this loss are not normal | 0.75 |
15 | My life has little to offer now s/he died | 0.85 |
16 | My life is worthless since s/he died | 0.82 |
17 | Sometimes I think that something is wrong with me, because I feel so sad about his/her death | 0.83 |
18 | When I think about his death, I feel frightened about all the things I feel | 0.87 |
19 | I continue to think that the thing that happened to him/her can also happen to me | 0.79 |
20 | Ever since s/he died, I continue to think that I can also die | 0.78 |
Internal Consistency
Cronbach’s alpha for the 20 items of the GCQ-C was 0.95 in Sample 1 (n = 83), 0.89 in Sample 2 (n = 63), and 0.93 in the combined sample (N = 151). In line with the results of the exploratory factor analysis, in the combined sample the item-total correlations were all positive and ranged from 0.42 (item 5: “I think that others should pay attention to how I am doing.”) to 0.87 (item 18: “When I think about his death, I feel frightened about all the things I feel.”). In the combined sample, the internal consistency did not increase with a deletion of a single item.
Temporal Stability
Thirty children completed the GCQ-C 3–9 weeks after the first time and 25–34 weeks after the first assessment. The test–retest correlation for the 30 children between scores at assessment 1 and assessment 2 was r = 0.84 (p < 0.001). The test–retest correlation for the scores at assessment 1 and assessment 3 was r = 0.73 (p < 0.001).
Concurrent Validity
Correlations of the GCQ-C with the IPG-C, CDI, CPSS, and CBCL scores are shown in Table
3. As predicted, the GCQ-C total score was significantly and positively correlated with indices of PGD (IPG-C), depression (CDI), posttraumatic stress (CPSS) and Internalizing Problems (CBCL-Internalizing); correlations were all large. Unexpectedly, GCQ-C scores were not significantly correlated with the CBCL-Externalizing scale and the CBCL Total Problems score. Also as we expected, higher scores on the GCQ-C were correlated with lower scores on the item tapping coping self-efficacy;
r = −0.24,
n = 147,
p < 0.01.
Table 3
Correlations between study measures in children
Prolonged grief disorder (IPG-C) | 0.76** |
Depression (CDI) | 0.67** |
Posttraumatic stress (CPSS) | 0.78** |
CATS-physical threat | 0.78** |
CATS-social threat | 0.66** |
CATS-personal failure | 0.78** |
CATS-hostile intent | 0.54** |
Internalising problems (CBCL) | 0.18* |
Externalising problems (CBCL) | 0.02 |
Total problems (CBCL) | 0.14 |
Internalising problems (YSR) | 0.74** |
Externalising problems (YSR) | 0.30** |
Total problems (YSR) | 0.63** |
Convergent and Divergent Validity
Correlations of the GCQ-C with the CATS subscales (
N = 151) are also shown in Table
3. As predicted, the GCQ-C was more strongly associated with Personal Failure (
r = 0.78), Physical Threat (
r = 0.78) and Social Threat (r = 0.66), than with the Hostility subscale (
r = 0.54, all
ps < 0.01). Differences between correlations were significant (
r = 0.78 vs.
r = 0.54,
z = −4.97
p < 0.001;
r = 0.78 vs.
r = 0.54,
z = −4.87
p < 0.001;
r = 0.66 vs.
r = 0.54,
z = −2.11,
p < 0.02.
Finally, both the GCQ-C total score and the summed score on the CATS were significantly associated with the IPG-C item that assesses impairment in functioning as a result of the loss. However, in contrast with what was expected, the correlation of the GCQ-C total score with functional impairment (r = 0.46, p < 0.001) was equal to the correlation of the CATS score with functional impairment (r = 0.45, p < 0.001). Consistent with that, both the correlation of the GCQ-C total score with the summed functional impairment items of the CPSS and the correlation of the CATS score with this index were r = 0.53, p < 0.001.
Scores on the GCQ-C did not vary as a function of time from loss, age, gender, relationship to the deceased, cause of death, and whether or not the death was expected (all ps > 0.08).
Discussion
The aims of the present study were to examine psychometric properties of the Grief Cognition Questionnaire for Children (GCQ-C) and, more generally, to enhance knowledge about maladaptive thinking in bereaved children (aged 8–18 years). The GCQ-C was modeled after the adult Grief Cognitions Questionnaire (Boelen et al.
2003) and designed to measure bereavement-related negative cognitions that are assumed to play a role in the development and persistence of childhood PGD and other symptoms of post loss psychopathology.
With regard to the dimensionality of the GCQ-C, outcomes of an exploratory factor analysis showed that the GCQ-C represented one underlying dimension of negative loss-related cognitions. This finding is inconsistent with research based on the adult GCQ, in which negative cognitions clustered into different factors representing different themes, including negative cognitions about the self, life, the future, and catastrophic misinterpretations of grief reactions (Boelen and Lensvelt-Mulders
2005). These findings suggest that there are differences between thinking patterns of bereaved children and adults such that bereaved children differentiate between different types of cognitions associated with loss less than adults do. Indeed, it is possible that adults distinguish between different themes (including the view of themselves, their lives, and futures) more clearly, such that some of these themes are appraised more negatively than others, whereas children’s negative appraisals generalize across different themes. In a related vein, it is possible that children do not necessarily experience less sadness or grief following loss compared to adults, but are less aware of the precise content of their cognitions that underly these feelings. It is also possible that self-report questionnaire are adequate means to assess different types of cognitions in adults, but may be less suitable to assess nuanced differences in cognitions about different themes in children. More research on this topic in larger populations is needed to gain more knowledge about the phenomenology of maladaptive thinking in bereaved children and differences with maladaptive thoughts patterns in adults. Reliability analyses showed that the GCQ-C had adequate internal consistency and temporal stability.
With respect to the concurrent validity, the GCQ-C total score was significantly and positively correlated (moderate to strong correlations) with self-report indices of PGD, depression and PTSD. It was also significantly, but with a weak correlation, correlated with internalizing problems as reported by parents, but not with externalizing problems. This finding is in line with prior research (e.g. Melhem et al.
2007; Spuij et al.
2012a) indicating that clinicians should ask children as informants about their grief related cognitions. Furthermore, as expected, higher scores on the GCQ-C were correlated with lower scores on the perceived ability to cope with the loss indicating that more negative cognitions associated with the loss coincided with a more negative view of children about their own ability to deal with the challenges brought about by the loss.
Several findings supported the convergent and divergent validity of the GCQ-C. GCQ-C scores were significantly related to automatic thoughts associated with a broad spectrum of negative emotional states, tapped by the CATS (Schniering and Rapee
2002); as predicted, the GCQ-C was more strongly associated with CATS subscales measuring cognitions associated with personal failure, physical threat, and social threat, than with the CATS subscale measuring cognitions associated with Hostility. There were no differences between the GCQ-C and the CATS in terms of their associations with indices of functional impairment associated with the loss.
It is noteworthy that endorsement of negative cognitions tapped by the GCQ-Q did not differ as a function of the background and loss related variables that we assessed. These findings accord with prior findings of a weak linkage between background and loss-related variables and PGD symptoms (Spuij et al.
2012a) and suggest that background and loss-related variables generally are not associated with high levels of loss related psychological problems.
Some limitations should be taken into account when interpreting the present findings. First, the sample size (
N = 151) may be considered somewhat small for exploratory factor analyses. Notably though, Guadagnoli and Velicer (
1988) argued that a factor is reliable, regardless of sample size, when it has four or more loadings greater than 0.6 and that with all communalities above 0.6 samples less than 100 may be adequate and with communalities in the 0.5 range samples between 100 and 200 can be good enough. This is why we decided that although the
N was quite small, it was worth to explore the factors in this sample. Second, given the variety of methods of recruitment, most of the analyses that we conducted relied on a rather homogeneous group, as most children lost a parent who died from cancer. Maybe there are differences in origin and severity of grief cognitions between subgroups (e.g., children who suffered from traumatic loss, children who lost a sibling or other relative) that were unidentified in the current study because of sample size constraints. Third, CBCL’s for participants were completed by parents. Because of their own grief, these parents possibly were not the most reliable informants of their children’s problems. Thus, it would be relevant for future studies to correlate grief cognitions of children and PGD severity also with indices of children’s problems obtained from other, more distant informants such as teachers. Fourth, the present study tested a Dutch version of the GCQ-C. Although it is conceivable that the present findings are generalizable to other Western—including English-speaking—cultures, the psychometric properties of versions of the questionnaire in other languages remain to be tested. As a related point, the cultural background of participants had a Western origin. Thus, generalization of the present findings to non-Western subgroups or religions remains to be determined. A final limitation of this study was its cross-sectional design that precluded any conclusions about the direction of causality between negative cognitions and emotional problems. It would be interesting for future research to use prospective-longitudinal designs to disentangle the linkage between negative thinking and emotional problems further.
Notwithstanding these considerations, results of the present study indicate that the GCQ-C is a promising measure for the assessment of negative loss-related cognitions in children and adolescents confronted with bereavement. As such it may be used in future studies on the role of cognitive variables in emotional problems after bereavement. To our knowledge, this is the first study that provides evidence for a strong link between negative thinking associated with different themes on the one hand, and elevated PGD and other symptoms of depression and PTSD on the other hand. Moreover, the measure may be useful in clinical practice for the identification of cognitions that are important targets of the treatment of PGD-symptoms. Given the lack of effective interventions for childhood grief (Currier et al.
2007), it is important to know which factors can mediate emotional problems in children after a loss. We recently conducted two studies that provided preliminary evidence that cognitive behavioral interventions are fruitful in ameliorating PGD symptoms and other symptoms of distress in bereaved children and adolescents (Spuij et al.
2013,
2015). The findings of these prior studies and the current study suggest that negative thinking plays an important role in childhood PGD and should be targeted in the treatment of this condition. This is in line with research from Meiser-Stedman et al. (
2009) who showed that maladaptive cognitions in children with PTSD are causally implicated in the unfolding and maintenance of posttraumatic stress responses over time and that these should be the focus of treatment of PTSD in children.