تجزیه و تحلیل عاملی تأییدی از مدل ترکیبی از پرسشنامه تروما دانشگاه هاروارد و پرسشنامه سوگ و داغدیدگی بازنگری پیچیده: آیا ما سوگ پیچیده و یا استرس پس از سانحه را اندازه گیریمی کنیم؟

Abstract The aim of this study was to assess the factorial structure of complicated grief (CG) and investigate the relationship between CG and posttraumatic stress disorder (PTSD) through the assessment of models combining both constructs. The questionnaire was completed by elderly, married respondents with a history of at least one significant, interpersonal loss (145 males and 147 females, 60–81 years). Confirmatory factor analysis (CFA) supported a two-factor model of separation and traumatic distress in CG. To investigate the relationship between CG and PTSD three combined models were specified and estimated using CFA. A model where all five factors, the two factors of CG and the three factors of PTSD, as defined by the DSM-IV, were allowed to correlate provided the best fit. The results indicated a considerable overlap between the dimensions of CG and PTSD, and complicated grief is construct that appears to be largely accounted for by especially the intrusive component of PTSD.

.

3. Results 3.1. Prevalence and internal consistency The prevalence of CG in this study indicated by scores above the cut-off point of 36 was 9% (M = 21.78; SD = 9.74; range 16–40). The internal consistency of the scores were acceptable for ICG-R both on the total scale and subscales, and the MIIC were relatively high (total score: α = .94; MIIC = .52. Separation anxiety: α = .89; MIIC = .66. Traumatic distress: α = .92; MIIC = .50). The prevalence of participants meeting clinical cut-offs of the three core criteria of PTSD as measures by the HTQ was 6% (HTQ-total: M = 25.04; SD = 7.44; range 16–40). The internal consistency and the inter-item correlations of the HTQ scores on the total scale and subscales were acceptable (total score: α = .88; MIIC = .33. Intrusion: α = .76; MIIC = .46. Avoidance: α = .78; MIIC = .34. Arousal: α = .80; MIIC = .45). Kappa analysis was used to determine the degree of co-morbidity between CG and PTSD. The analysis indicated a moderate degree of co-morbidity beyond chance (κ = .48, p < .0005). Semi-partial correlations between HTQ and ICG-R total scores controlling for types of loss individually revealed little change in the strength of correlations between the two constructs. Only when controlling for loss of a parent and loss of a spouse a small fall in the strength of the correlation between HTQ and ICG-R was observed. This indicates that it is unlikely that the nature of the loss moderated the relationship between the two constructs. 3.2. Factor structure of complicated grief In accordance with previous EFA of measures of CG, and the common use of the ICG-R, we tested a one-factor model of ICG-R that included all items. We also tested the theoretical derived two-factor model (correlated and uncorrelated) consisting of symptoms of separation anxiety and traumatic distress. The fit indices of the alternative models of ICG-R are reported in Table 1. On the basis of the criteria associated with RMSEA and SRMR, and IFI and CFI, the one-factor model and the correlated two-factor model were judged to exhibit reasonable model fit. The chi-square statistics were statistically significant, but this should not lead to rejection of the models, because the large sample size increases the power of the test (Tanaka, 1987). Comparing the one-factor model and the correlated two-factor model, the correlated two-factor model had a significant better fit (ΔS-Bχ2 = 10.35; Δdf = 1, p < .005). Also, the RMSEA, ECVI and SRMR values were in favour of the correlated two-factor model, as were the chi-square/df ratio (one-factor model: 2.93; two-factor-c model: 2.77), and the IFI and CFI values. Thus, it was judged that the two-factor model represented an adequate description of the data and is the best of the alternative models. The two factors correlated very highly correlated (r = .84) and factor loadings ranged from .60 to .66 on the separation anxiety factor and from .38 to .68 on the traumatic distress factor. The loadings can be seen in Table 2. All loadings were statistically significant (p < .05). Table 1. Fit Indices for the alternative models of the Inventory of Complicated Grief. Index One-factor model Two-factor model correlated Two-factor model uncorrelated S-Bχ2, df (p) 263.44, 90 (.00) 247.18, 89 (.00) 302.42, 90 (.00) RMSEA 90% CI .081 (.070–.093) .078 (.067–.090) .090 (.079–.100) ECVI 90% CI 1.11 (.96–1.29) 1.06 (.91–1.24) 1.25 (1.08–1.44) IFI .95 .96 .94 CFI .95 .96 .94 SRMR .073 .071 .20 Note: S-Bχ2, Satorra–Bentler scaled chi-square; RMSEA, Root Mean Square Error of Approximation; ECVI, Expected Cross-Validation Index; IFI, Incremental Fit Index, CFI, Comparative Fit Index; SRMR, Standard Root Mean-square Residual. Table options Table 2. Standardized factor loadings for the correlated two-factor model of the ICG-R. Question Separation anxiety Traumatic distress Q1 .60 Q2 .65 Q3 .66 Q4 .66 Q5 .42 Q6 .64 Q7 .68 Q8 .53 Q9 .63 Q10 .63 Q11 .65 Q12 .38 Q13 .65 Q14 .61 Q15 .56 Note: All factor loadings significant (p < .05). Table options 3.3. Factor structure of PTSD Three models of PTSD were tested using the HTQ. The original DSM-IV three-factor model consisting of intrusion, avoidance, and arousal, and a one-factor PTSD model were estimated. Following recent studies of the structure of PTSD symptoms, we also tested the alternative four-factor model proposed by Simms et al. (2002). The fit indices of the alternative models of HTQ are reported in Table 3. On the basis of meeting all criteria associated with RMSEA and SRMR, and IFI and CFI, the three-factor model and the four-factor model were judged to exhibit reasonable model fit. Again, the chi-square statistics were statistically significant. Comparing the three-factor model and the four-factor model, there where no significant difference in fit (ΔS-Bχ2 = 2.43; 3 Δdf = 3, p = .49). However, the three-factor model had a lower chi-square/df ratio (2.00) compared with the four-factor model (2.21), lower RMSEA, ECVI and SRMR values, and slightly higher values for IFI and CFI. On that basis, and because of the preference of the most parsimonious model, it was judged that the three-factor DSV-IV model represents an adequate description of the data and a better model than the four-factor model. The three symptom clusters correlated highly (r = .65–.81) and factor loadings ranged from .64 to .72 on the intrusion factor, from .39 to .71 on the avoidance factor and from .61 to .73 on the arousal factor. All loadings were statistically significant (p < .05). Table 3. Fit Indices for the alternative models of the Harvard Trauma Questionnaire. Index One-factor model Three-factor model (DSM-IV) Four-factor model (Simms et al., 2002) S-Bχ2, df (p) 322.08, 104 (.00) 201.84, 101 (.00) 216.47, 98 (.00) RMSEA 90% CI .085 (.074–.095) .059 (.047–.070) .064 (.053–.076) ECVI 90% CI 1.33 (1.15–1.53) .93 (.81–1.09) 1.01 (.87–1.16) IFI .95 .98 .97 CFI .95 .98 .97 SRMR .075 .064 .065 Note: S-Bχ2, Satorra–Bentler scaled chi-square; RMSEA, Root Mean Square Error of Approximation; ECVI, Expected Cross-Validation Index; IFI, Incremental Fit Index, CFI, Comparative Fit Index; SRMR, Standard Root Mean-square Residual. Table options 3.4. Combined models ICG-R and HTQ consisted of two completely separate sets of items reflecting a set of two and a set of three factors, respectively. The univariate correlation between the HTQ and CG total scores was high and statistically significant (r = .68, p < .0005). To investigate the overall relationship between CG and PTSD three combined models were specified as described in the introduction of this study and estimated using the ICG-R and the HTQ. The fit indices for these three models of ICG-R and HTQ are reported in Table 4. On the basis of meeting the criteria associated with RMSEA and SRMR only, Model 3 was judged to exhibit reasonable model fit. Comparing Model 3 with Model 1 (ΔS-Bχ2 = 35.63; Δdf = 10, p < .0005) and Model 2 (ΔS-Bχ2 = 20.49; 6 Δdf = 6, p < .005), Model 3 had a significantly better fit. Also, Model 3 had a lower chi-square/df ratio and ECVI values than Model 1 and Model 2. Moreover, Model 3 met the .95 criteria for the IFI and CFI, considered to reflect acceptable model fit. On that basis it was judged that Model 3 represents a reasonable description of the data, suggesting that CG and PTSD are two correlated, but different constructs. Looking at the specific formulation of items from the two scales, it appears that item 1 on the HTQ and item 15 on the ICG-R relate to recurrent thoughts about the death. Item 4 on the HTQ and item 21 on the ICG-R relate to feeling detached from other people, whereas item 14 on the HTQ and item 20 on the ICG-R relate to feeling there is no real future after the loss. Possible strong correlations between the above items must be considered when investigating the relationship between CG and PTSD. Therefore, in an additional analysis of Model 3, the three sets of almost identical items in the HTQ/ICG-R were allowed to correlate. This alteration did not result in a better fit (results not shown). Factor correlations of Model 3 ranged from r = .39 to r = .88 on the two CG factors and from r = .37 to r = .73 on the three PTSD factors. All correlations were statistically significant (p < .05). Table 4. Fit Indices for the combined models. Index Model 1: one-factor model Model 2: five-factor model (ICG-R and HTQ uncorrelated) Model 3: five-factor model (ICG-R and HTQ correlated) S-Bχ2, df (p) 1489.27, 434 (.00) 840.28, 430 (.00) 756.14, 424 (.00) RMSEA 90% CI .091 (.086–.097) .057 (.052–.063) .052 (.046–.058) ECVI 90% CI 5.54 (5.15–5.96) 3.34 (3.07–3.64) 3.09 (2.84–3.37) IFI .95 .98 .98 CFI .95 .98 .98 SRMR .089 .24 .067 Note: S-Bχ2, Satorra–Bentler scaled chi-square; RMSEA, Root Mean Square Error of Approximation; ECVI, Expected Cross-Validation Index; IFI, Incremental Fit Index, CFI, Comparative Fit Index; SRMR, Standard Root Mean-square Residual. Table options The correlations between the five factors in Model 3 are shown in Fig. 1. Traumatic distress was very highly correlated with intrusion (r = .87) and highly correlated with the two other clusters of PTSD symptoms (r = .49–.56). Separation anxiety was very highly correlated with intrusion and avoidance (r = .73–.86) and highly correlated with arousal (r = .56). The dependent correlations among the CG and PTSD factors (Model 3) were compared following the approach suggested by Steiger (1980). The association between separation anxiety and arousal was significantly weaker than the associations found between separation anxiety and avoidance (t = −7.3, p < .001) and intrusion (t = −12.7, p < .001), respectively. Moreover, a stronger association was found between separation anxiety and intrusion as compared to separation anxiety and avoidance (t = −6.8, p < .001). Furthermore, traumatic distress was significantly more associated with intrusion than arousal (t = −15.4, p < .001) and avoidance (t = −18.8, p < .001). Finally, traumatic distress was more associated with avoidance than arousal (t = −2.1, p < .05). Model 3 with standardized factor correlations. All correlations given are ... Fig. 1. Model 3 with standardized factor correlations. All correlations given are significant at p < .05