سوگ و داغداری طولانی مدت پس از مرگ اخیر یک دختر در میان مادران که ضرر و زیان دیستال در دوران خمرهای سرخ را تجربه کرده اند: اعتبار سوگ و داغداری طولانی مدت در کامبوج

Abstract This study addressed the validity of the prolonged grief (PG) construct in a Cambodian context. Eighty mothers who lost a young adult daughter stemming from a crowd stampede incident during the annual water festival were interviewed at the six-month post-loss point along with a control group of similarly aged women who were not recently bereaved. Both groups were assessed for PG, PTSD, anxiety, and depression symptoms and well as for the number of distal losses experienced during the Khmer Rouge (KR) regime – knowing that all the women were old enough to have lived through the KR regime. Support for the discriminant validity of PG was shown in a factor analysis in which its core symptoms were distinguished from anxiety, depression, and PTSD symptoms. Also, support was found for its incremental validity as shown in the unique sensitivity of PG in distinguishing the two groups when controlling for the other symptoms. Lastly, a positive relationship was found between the number of distal deaths experienced during the KR regime and PG symptom severity among the group of recently bereaved mothers, providing support for the predictive validity of PG. Implications as well as study limitations are discussed.

Introduction On November 22, 2010 during the annual Water Festival in Phnom Penh, Cambodia, a tragic incident occurred wherein approximately 350 Cambodians lost their lives as a result of crowd congestion on the Koh Pich bridge that led to panic and stampeding resulting in crushing and suffocation of people (Mydans, 2010). The gravity of this event for Cambodians was reflected in an official day of mourning that followed and a public statement by the prime minister Hun Sen that this was the worst tragedy encountered by Cambodians since the horrific Khmer Rouge regime between 1975 and 1979 where almost 25% of Cambodians died – often at the hands of violent means (Kiernan, 2002). Such a comparison highlights the shadow of the past regime that continues to haunt Cambodians who lived through it, as shown in the significant incidence of lifetime PTSD among KR survivors (Sonis et al., 2009). In the bereavement literature, there is evidence that those who experienced losses earlier in life are at greater risk for psychopathology including prolonged grief disorder (PGD) following the death of a significant other in adulthood (Silverman et al., 2001, Luecken, 2008 and Prigerson et al., 2009). It is also known that the death of a child is the type of loss associated with the most severe grief (Murphy, 2008 and Song et al., 2010). Finally, there is evidence that sudden loss – especially of a more shocking or gruesome nature, as what occurred in the Koh Pich bridge stampede incident – is more difficult to integrate than expected loss (Parkes, 2008). Knowing that a significant number of those who died in the Koh Pich incident were young women in their late teens through late 20s, all three of these risk factors may apply to their parents – having lived through the KR Regime and thus having experienced varying degrees of losses at a younger age, the unexpected nature of the recent death, and the fact that it involved the death of a child. Therefore, a high incidence of PGD is likely to be found among these parents. Such a loss may be especially debilitating for mothers, given her importance as a role model in the socialization of her daughter among Cambodians (Ebihara et al., 1994). To our knowledge, this is the first study to address the impact of the Koh Pich bridge stampede incident on adjustment to bereavement and how exposure to distal loss stemming from the KR period may affect response to this more recent loss. Specifically, it serves the dual aims of addressing the psychological consequences of such loss while simultaneously examining the validity of the PGD construct in a Cambodian context. PGD has been proposed as a distinct mental disorder stemming from the loss of a loved one (Prigerson et al., 2008 and Prigerson et al., 2009). This diagnosis requires the presence of severe grief symptoms of at least six months duration after the death. These symptoms include separation distress characterized by intense yearning for the deceased and intense emotional pain, and additional cognitive, affective, and behavioral symptoms linked to the death. In order to qualify as a PGD diagnosis, these symptoms are at a level of severity such as to significantly interfere with the bereaved’s occupational and social functioning. A large body of literature exists in support of the validity of PGD as a distinct diagnosis from other comorbid disorders stemming from bereavement including depression and anxiety disorders as well as PTSD (Prigerson et al., 2008 and Prigerson et al., 2009). Although the majority of studies addressing PGD have been conducted on Western populations, there has been an increasing number of such studies with non-Western cultures including Taiwan (Chiu et al., 2010), Rwanda (Schaal et al., 2009), and Pakistan (Prigerson et al., 2002). Of particular relevance to the present study, Stammel et al. (2013) recently examined PGD among Cambodian survivors of the KR regime. Their study provided support for the predictive validity of PGD stemming from distal loss in showing that KR survivors who experienced a greater number of deaths or who lost a close relative during the KR period were at greater risk for developing PGD. The present study extends the work of Stammel et al. (2013) in contributing to the further validation of the PG construct in a Cambodian setting addressing other aspects of validity in addition to its predictive validity, including its discriminant validity and its incremental validity over other bereavement-related measures of distress. An important evidence base for PGD is its distinct clinical phenomenology when compared with other co-morbid symptom constellations shown in factor analytic studies involving bereaved individuals in which PGD symptoms are shown to load on a separate factor from these other symptoms. In a number of such studies, PGD items have been distinguished from bereavement-related anxiety and depression as well as PTSD symptoms (Boelen et al., 2010, Golden and Dalgleish, 2010 and Ogrodniczuk et al., 2003) – thereby providing support for the discriminant validity of PG symptoms. The present study extends this work to a Cambodian context in showing that PG symptoms can be distinguished from anxiety, depression, and PTSD symptoms in a factor analysis. In addition to providing support for the discriminant validity of PGD, this study addressed its incremental validity in terms of its sensitivity in detecting degree of distress as a function of time since the death beyond other bereavement-related symptom measures. This was addressed by including a similarly aged control group of women from the same villages where the bereaved mothers who lost their daughters in the Koh Pich bridge stampede resided. Because the mothers in this control group were also known to have experienced past albeit more distal losses – in having survived the KR regime and its aftermath – important differences between the two groups in time since the loss provided a basis for determining whether PG symptoms were a more sensitive index in distinguishing the two groups than anxiety, depression, and PTSD symptom measures. Research on the trajectory of bereavement-related distress has shown that symptoms typically decrease over time (Bonanno et al., 2008). However, different types of bereavement-related symptoms are also known to have different trajectories of change over time. For example, Thompson et al. (1991) found that among older widowed adults the reduction in depression and general psychopathology symptoms leveled out at six months post-loss whereas grief-specific symptoms continued to decrease over a 30-month post-loss period. Such results suggest that PG symptoms may be a more sensitive marker of bereavement-related distress than other co-morbid symptoms of bereavement. We therefore expected that PG symptoms would provide greater sensitivity than these other symptom measures toward distinguishing the group of mothers who recently lost daughters in the Koh Pich stampede from a control group of women who experienced loss at a more distal point in time. This study also addressed another aspect of the predictive validity of PG in terms of the effects of earlier losses on response to later losses. In a study on conjugal bereavement, Silverman et al. (2001)found that the incidence of PG disorder was higher among recently widowed older adults who had experienced death of a parent in childhood relative to those who did not experience such childhood adversity. Other studies have similarly found that death of a parent in childhood may be a risk factor for later psychopathology (Luecken, 2008). Finally, multiple losses have been associated with more severe grief (Mercer and Evans, 2006 and Stammel et al., 2013). Thus, among mothers who lost daughters in the Koh Pich bridge stampede those who experienced loss of a parent or a greater number of losses during the KR period would have more severe PG symptoms – knowing that many of these mothers were children during that time. We expected that this would especially hold true among mothers who were children during the KR period in light of evidence that significant losses in childhood, especially the death of a parent, have long-term consequences for psychological adjustment (Silverman et al., 2001 and Luecken, 2008). 1.1. Present study This study builds upon previous work of Stammel et al. (2013) toward further validation of the PG construct in a Cambodian context. Their study examined PG stemming from distal losses during the KR regime but did not address the impact of more recent bereavement or the effect of more distal loss during the KR regime on response to more recent loss. The present between-group design study involving mothers who have lost daughters stemming from the Koh Pich crowd stampede incident with similar aged mothers from the same villages who had not experienced a recent loss provided such an opportunity. Knowing that the mothers in both groups were born prior to or during the KR regime, and knowing that virtually all Cambodians who lived through that period were faced with varying numbers of losses of loved ones, the study design provides a means for assessing the PG construct in terms of its sensitivity in distinguishing the effect of recent loss from more distal loss over-and-above anxiety, depressive, and PTSD symptoms. It also provides the opportunity for examining the cumulative effects of more distal and recent losses on PG. We investigated the degree of psychological distress among mothers at the six-month post-loss point of their daughter’s deaths stemming from the Koh Pich bridge stampede incident. This amount of time following the loss was sufficient for determining the incidence PGD, given the criterion that PG symptoms must remain elevated for at least six months in order to receive a PGD diagnosis (Prigerson et al., 2008 and Prigerson et al., 2009). In light of the unexpected and shocking nature of this loss and relatively close proximity to the time of the death at the time of assessment, we expected that these women as a group would have a high incidence of PGD as well as a high incidence of clinical levels of bereavement-related anxiety, depression, and PTSD symptomatology. Toward further validation of the PG construct in a Cambodian setting, this study addressed the following hypotheses: (1) following previous work largely conducted on Western samples, PG symptoms were expected to load on a distinct factor from anxiety, depression, and PTSD symptoms – providing support for the discriminant validity of PG in terms of its distinct phenomenology; (2) in support of its incremental validity in terms of its sensitivity in distinguishing the bereaved and control groups (who should differ in degree of bereavement-related distress as a function of time since the loss) PG symptoms were expected to successfully distinguish the two groups (when controlling for anxiety, depression, and PTSD symptoms); and (3) in support of the predictive validity, among the recently bereaved mothers, those who encountered a greater number of losses or experienced the death of a parent during the KR period were expected to have more severe PG symptoms. Moreover, the effects of these distal losses were expected to be more pronounced among the younger recently bereaved mothers – knowing that these women were children during the KR regime and therefore especially likely to be affected by such early losses.

. Results 3.1. Time since the death among control group Prior to conducting analyses addressing the main set of hypotheses, it was important to confirm that the control group differed from the bereaved group in terms of time since the death since this is known to affect symptom severity (Bonanno et al., 2008). The average time elapsed since the death was 15.43 years (S.D.=12.28) – significantly longer than the 6-month post-loss point among bereaved group participants. Among the deaths reported by control participants, 31.6% involved death of a parent, 7.6% death of a child, 13.9% death of a spouse, 11.4% death of a sibling, 20.3% death of a relative, and 15.2% other. 3.2. Bereaved and control group symptom means Independent t-tests were conducted to compare the bereaved and control groups on their mean symptom levels for prolonged grief, depression, anxiety, and PTSD symptoms ( Table 2). The bereaved group was significantly higher than the control group on each of the symptom measures. It is noteworthy that mean symptom levels in the bereaved group for anxiety (M=2.13) and depression (M=2.11) were higher than the clinical cutoff of 1.75 as was the mean for the PTSD total score (M=46.23) which was greater than the clinical cutoff score of 44. This highlights the high levels of distress experienced by the bereaved mothers at six months following the death of her daughter. Table 2. Bereaved and control group symptoms means of disordered level of symptoms with T-test. N M S.D. t PG Bereaved 80 3.32 0.97 11.568⁎⁎⁎ Control 79 1.79 0.66 PTSD Bereaved 80 46.23 15.34 7.393⁎⁎⁎ Control 72 29.68 11.82 Anxiety Bereaved 80 2.13 0.84 4.330⁎⁎⁎ Control 79 1.60 0.69 Depression Bereaved 80 2.11 0.80 4.647⁎⁎⁎ Control 79 1.59 0.60 ⁎⁎⁎ p<0.001. Table options A Chi-square analysis was performed to compare the incidence of disordered symptom levels in the bereaved and control groups. This required first dichotomizing participants into disordered versus non-disordered levels based on the clinical cutoff levels for anxiety, depression, and PTSD total score respectively. Participants were similarly dichotomized based on the criteria for determining prolonged grief disorder previously described in Section 2 (Table 3). The incidence of clinical levels of each of the symptoms was significantly higher in the bereaved group relative to the control group. A striking difference was shown for prolonged grief in which 47.5% of participants in the bereaved group had a prolonged grief diagnosis whereas only 1.3% of the participants in the control group had prolonged grief. A similarly high incidence of PTSD was shown in which the prevalence of probable PTSD among the bereaved mothers was 57.5%, as was the prevalence of 63.8% for anxiety and 61.3% for depression. Table 3. Bereaved and control group incidence of disordered levels of symptoms. Group Bereavement Control χ2 PG Without 42 78 45.90⁎⁎⁎ With 38 1 PTSD Without 45 66 24.13⁎⁎⁎ With 35 6 Anxiety Without 29 54 16.42⁎⁎⁎ With 51 25 Depression Without 31 55 15.25⁎⁎⁎ With 49 24 ⁎⁎⁎ p<0.001. Table options 3.3. Convergent validity for prolonged grief A Pearson correlation analysis was conducted to examine the relationship among the symptom measures (Table 4). Support for the convergent validity of the PG-13 measure of prolonged grief is shown in its strong to moderate positive correlation with total PTSD symptoms (r=0.76, p<0.001), anxiety (r=0.54, p<0.001), and depression (r=0.62, p<0.001). Table 4. Correlations among symptom measures. Prolonged grief Anxiety Depression Prolonged grief — Anxiety 0.54 — Depression 0.62 0.81 — PTSD 0.76 0.68 0.75 Table options 3.4. Factor analysis comparing pg with depression, anxiety, and PTSD symptoms Prior to comparing PG symptoms with the other symptom measures, a principal axis exploratory factor analysis was conducted on the full study sample in order to determine the factor structure of the PG-13 measure within a Cambodian context. In many of the previous studies with Western samples, PG symptoms have been shown to load on a single factor (Boelen et al., 2003 and Prigerson et al., 1995). It would therefore be of interest to know whether a single factor structure for PG would also hold true in a non-Western Cambodian sample. Eigenvalue greater than 1 and scree plot criteria, in conjunction with interpretability of the factors, served as bases for determining its factor structure. The results of this analysis indicated that a single factor structure best characterized the PG-13. The factor loadings for this single factor solution (eigenvalue=6.32) ranged from 0.64 to 0.86 and explained 57.41% of the variance. An internal consistency analysis confirmed that a single factor structure was highly satisfactory (alpha=0.92). Each of the item-total scale correlation scores was high (ranging from 0.38 to 0.75) such that the Cronbach alpha for the total scale score was reduced when any of the items were deleted. To address the hypothesis regarding the unique phenomenology of PG symptoms as distinct from other bereavement-related symptoms, a principal axis factor analysis with a promax oblique rotation was conducted on the full set of PG-13, PCL PTSD, and Hopkins-25 anxiety and depression symptom items. An oblique rotation was selected because the symptom measures were significantly correlated with one another. The bereaved and control groups were combined for this analysis in order to ensure a sufficient sample size. Conducting the factor analysis on the combined groups was also advantageous in ensuring that a broader range of symptoms scores were likely to be represented. Although the rule-of-thumb constraint that there should be no more than one item for each five participants (Gorsuch, 1983) was not satisfied, there is evidence that sample size is more important in factor recovery in Monte Carlo analyses (Preacher and MacCallum, 2002), and that a 3:1 ratio of participants to variables may be acceptable in sample sizes greater than 100 (Arrindell and van der Ende, 1985) — a condition met in the combined groups sample that consisted of 159 participants and for which a total of 53 items were included in the factor analysis. Ten factors were extracted based on the eigenvalue criterion of greater than 1 that explained 69.95% of the variance. The first four factors extracted were easily interpretable. Most importantly, the prolonged grief items loaded on a separate factor (factor 2) in this 10-factor solution; all of the items except for PG-13 item 5 had loadings of greater than.40. The anxiety (factor 1) and depression (factor 3) items also loaded on separate factors. Although the PTSD re-experiencing items were distinguished on a separate factor (factor 4) that also included some avoidance items, the remaining PTSD items were spread among the additional factors. In light of this, a series of forced factor solutions were performed that involved progressively decreasing the number of factors specified. A 4-factor solution was determined to be the most interpretable factor structure in terms of largely providing support for simple structure characterized by high loadings on one factor (>0.40) and corresponding low loadings on the other factors (<0.20) for each item. This 4-factor solution was also justified based on the scree plot criterion and on conceptual grounds in distinguishing the four different types of symptoms that were included in the factor analysis. The pattern matrix of this 4-factor solution that explained 56.07% of the variance is displayed in Table 5. Table 5. Factor analysis: PG, PTSD, depression, & anxiety. Symptoms Loading on factor 1 (Anxiety) Loading on factor 2 (Prolonged grief) Loading on factor 3 (Depression) Loading on factor 4 (PTSD) Prolonged Grief (PG-13) Longing/yearning 0.058 0.888 −0.050 −0.090 Emotional pain 0.051 0.914 −0.027 −0.075 Avoiding reminders −0.117 0.760 −0.088 0.108 Shocked by loss 0.055 0.600 −0.053 0.266 Role confusion 0.023 0.232 0.378 0.182 Difficulty accepting loss 0.066 0.495 0.005 0.159 Mistrust others 0.091 0.475 −0.051 0.055 Bitter feeling 0.050 0.875 −0.102 −0.112 Difficulty moving on −0.104 0.526 0.092 0.136 Emotional numbness 0.097 0.648 0.165 −0.041 Feeling empty/meaningless −0.163 0.477 0.440 0.008 PTSD Checklist (PCL-C) Intrusive thoughts/images 0.062 0.397 −0.078 0.397 Disturbing dreams −0.040 0.129 0.158 0.432 Flashbacks −0.159 0.085 0.178 0.615 Feelings of upset −0.175 0.262 0.055 0.641 Physical reactions 0.272 0.150 −0.283 0.707 Avoiding thoughts −0.085 0.162 −0.037 0.698 Avoiding activities/situations −0.010 0.273 −0.194 0.632 Trouble remembering −0.015 −0.073 0.004 0.780 Loss of interest 0.125 0.200 0.123 0.429 Feeling distant from others −0.218 0.098 0.349 0.314 Emotional numbness 0.032 0.056 0.241 0.434 Sense of foreshortened future −0.166 0.023 0.399 0.452 Sleep disturbance 0.390 −0.041 0.130 0.337 Irritability 0.293 −0.040 0.135 0.327 Difficulty concentrating 0.304 −0.007 0.142 0.388 Hyper-alertness 0.145 −0.178 0.149 0.642 Easily startled 0.126 −0.069 0.196 0.492 Hopkins Symptoms Checklist-25 (HSCL-25) Depressive symptoms Feeling low 0.595 0.067 0.192 −0.015 Blaming self 0.174 −0.061 0.437 0.217 Crying 0.225 0.168 0.293 0.123 Loss of sexual interest 0.210 0.135 0.175 −0.171 Poor appetite 0.628 0.159 −0.213 0.115 Sleep disturbance 0.538 −0.043 0.310 0.032 Hopelessness about future 0.171 0.077 0.653 −0.019 Feeling blue 0.144 0.080 0.684 0.015 Loneliness 0.034 −0.051 0.870 −0.079 Suicidal thoughts −0.012 −0.109 0.493 −0.026 Feeling trapped 0.110 −0.117 0.550 0.160 Worries 0.229 0.020 0.668 −0.089 Loss of interest −0.068 0.035 0.566 0.201 Feeling everything is an effort 0.465 −0.093 0.502 −0.089 Feeling of worthlessness 0.034 −0.061 0.707 0.069 Anxiety symptoms Feeling fearful 0.529 −0.114 −0.273 0.467 Faintness/dizzy/weakness 0.405 0.043 0.147 0.206 Nervousness 0.789 0.004 −0.099 −0.008 Heart racing 0.741 −0.038 −0.059 0.152 Trembling 0.829 0.054 0.014 0.028 Feeling tense 0.845 0.109 0.039 −0.172 Headaches 0.397 0.160 0.389 −0.138 Panic 0.699 0.006 0.157 −0.118 Restlessness 0.736 −0.113 0.143 −0.087 Extraction method: principal axis factoring. Rotation method: promax with Kaiser normalization. Table options Most noteworthy, the Table 5 pattern matrix 4-factor solution results identify a distinct factor for the PG-13 items on factor 2. Apart from item 5, each of the PG-13 items have loadings of greater than 0.40 on factor 2, while the factor 2 loadings are less than 0.20 on the other symptom measure items except for PTSD item 1. Moreover, apart from item 5 and 10, the PG-13 items have loadings of less than 0.20 on the other factors. Although the other symptom measures were not as clearly distinguished from each other, factor 1 identified the anxiety items, while the majority of depression items loaded on factor 3, and the PTSD items largely loaded on factor 4. 3.5. Incremental validity of PG in distinguishing the bereaved and control groups A logistic regression analysis was conducted to provide support for the incremental validity of PG symptoms in terms of its sensitivity in distinguishing the bereaved and control group’s over-and-above the other symptom measures. Here, the symptom measures were included as predictors of bereaved versus control group status. The results are displayed in Table 6. Indeed, PG symptoms were the only significant predictor of group status when entered simultaneously with the other symptom measures in the regression model. Table 6. Symptom severity predicting bereaved versus control group status. B se Wald Constant 4.283 0.771 30.865⁎⁎⁎ Prolonged grief −1.871 0.378 24.487⁎⁎⁎ Anxiety 0.087 0.476 0.034 Depression 0.877 0.588 2.224 PTSD −0.039 0.027 2.182 ⁎⁎⁎ p<0.001. Table options A similar logistic regression analysis was employed wherein the symptom measures were entered as categorical variables based on the clinical cutoff criteria for each of these measures in predicting group status (Table 7). Replicating the previous finding, PG diagnosis was shown to be uniquely predictive of group status whereas none of the other symptom diagnoses distinguished the two groups. Both of the logistic regression analyses therefore provided clear-cut support for the incremental validity of PG in terms of its sensitivity in distinguishing recent versus more distal bereavement when compared depression, anxiety, and PTSD within a Cambodian context. Table 7. Symptom diagnosis predicting bereaved versus control group status. B se Wald Constant 1.745 0.646 7.308⁎⁎ Prolonged grief −3.551 1.067 11.077⁎⁎⁎ Anxiety −0.356 0.522 0.465 Depression 0.225 0.525 0.185 PTSD −0.962 0.589 2.669 ⁎⁎ p<0.01. ⁎⁎⁎ p<0.001. Table options 3.6. Effect of KR loss on PG among Koh Pich bereaved mothers A final set of analyses addressed the predictive validity of PG in terms of its sensitivity in registering the impact of the number of distal deaths during the KR regime and whether such losses included death of a parent on PG. The control group was not included in the analysis because the focal loss event for which a number of control group participants answered the PG-13 occurred after the end of the KR regime. The first analysis involved determining whether the number of losses during the KR regime had an effect on PG symptom severity among the Koh Pich stampede bereaved mothers. Because mothers who were younger during the KR were likely to be more psychologically affected by KR losses, age was expected to moderate the impact of KR number of losses on PG symptoms such that a stronger effect would be found among younger participants. This was addressed in a multiple regression analysis in which KR number of losses and the mother’s age were entered as predictors, along with their interaction, of PG symptoms. The analysis was conducted using the SPSS macro supplied by Hayes and Matthes (2009) for identifying moderational effects with continuous predictor variables. This macro provides syntax for identifying the effect of a focal predictor at specified values of a moderator variable. In the present study context, the effect of mothers’ KR number of losses was treated as the focal predictor while mother’s age was included as a moderator. Interaction effects were probed at −10+1 S.D. value points for mother’s age on her PG symptoms. Mean centered scores for the focal predictor and moderator variable were also specified for the analysis. The results indicated that no significant main effects were found for either KR number of losses (B=−0.0772, se=0.1138; t=−0.6784, ns) or mother’s age (B=−0.0025, se=0.0165, t=−0.1507, ns) in predicting PG symptom severity; however, a significant interaction effect was found (B=0.0576, se=0.0183, t=3.1500, p<0.01). Surprisingly, as shown in the conditional effect of KR number of losses at −10+1 S.D. value points for age in Table 8, the interaction effect was opposite to that predicted. Here, a significant negative relationship was found between KR number of losses and PG symptoms among younger mothers, (B=−0.4540, se=0.1939, t=−2.3409, p<0.05 at S.D.−1) whereas a significant positive relationship was shown among the older ones (B=0.2995, se=0.1301, t=2.3027, p<0.05 at S.D.+1). Table 8. Conditional effect of number of KR losses on PG symptoms at values of the moderator age. Mean age B se t 42.75 (−1 S.D.) −0.4540 0.1939 −2.3409⁎ 49.29 (0 S.D.) −0.0772 0.1138 −6784 55.83 (+1 S.D.) 0.2995 0.1301 2.3027 ⁎ p<0.05. Table options A comparable logistic regression analysis was conducted to determine whether age moderated the impact of KR trauma exposure on the likelihood of PG diagnosis. Again, the results indicated that KR number of losses was not a significant predictor of PG diagnosis (B=−0.1776, se=0.3122; z=0.5690, ns), nor was age (B=0.0296, se=0.0413; z=0.7159, ns), whereas a significant interaction effect was found (B=0.1387, se=0.0548; z=2.5328, p<0.05). Results for the conditional effect of KR trauma exposure on PG diagnosis at −10+1 S.D. points for age are shown in Table 9. A similar pattern to that shown previously was found wherein a significant negative relationship was found between KR number of losses and PG diagnosis among younger mothers (B=−1.0844, se=0.5511; z=−1.9675, p<0.05 at −1S.D.), whereas a positive relationship was found among the older mothers (B=0.7291, se=0.3841; z=1.8984, p=0.06 at +1 S.D.). Table 9. Conditional effect of number of KR losses on PG diagnosis at values of the moderator age. Mean age B se z 42.75 (−1 S.D.) −1.0844 0.5511 −1.9675⁎ 49.29 (0 S.D.) −0.1776 0.3122 −0.5690 55.83 (+1 S.D.) 0.7291 0.3841 1.8984 (p=0.06) ⁎ p<0.05. Table options A regression analysis was conducted to determine whether type of loss in terms of whether it involved death of a parent during the KR regime, in conjunction with age, was predictive of PG symptoms. There were no significant main effects shown for type of loss (B=−0.2063, se=0.3789; t=−0.5444, ns) and mother’s age (B=−0.0089, se=0.0172, t=−0.5122, ns), or a significant interaction effect (B=0.1064, se=0.0565, t=1.8827, ns), in predicting PG symptom severity. In a similar logistic regression analysis, treating PG diagnosis as the criterion variable, similarly no support was shown for type of loss (B=−1.5711, se=0.1.4198; z=−1.1066, ns) or age (B=0.0255, se=0.0428; z=0.5956, ns), or their interaction (B=0.3242, se=0.2058; z=1.5752, ns).