Anxiety, Stress, & Coping, 2015 Vol. 28, No. 3, 272–286, http://dx.doi.org/10.1080/10615806.2014.956097
Latent classes of posttraumatic stress and growth Marianne Skogbrott Birkelanda*, Gertrud Sofie Hafstada, Ines Blixa and Trond Heira,b a
Norwegian Centre for Violence and Traumatic Stress Studies, 181 Nydalen, 0409 Oslo, Norway; b Institute of Clinical Medicine, Faculty of Medicine, University of Oslo, Oslo, Norway (Received 4 June 2014; accepted 15 August 2014) Background and Objectives: Potentially traumatic events may lead to different patterns of posttraumatic stress symptoms and posttraumatic growth. The objective of the present study was to identify subgroups with different patterns of posttraumatic reactions, and to determine whether these subgroups differed in terms of personal and social resources and indicators of adjustment. Design: This study used survey data collected 10 months after the 2011 Oslo bombing attack to investigate patterns of reactions among ministerial employees (N = 1970). Methods: We applied latent class analyses with covariates to extract subgroups of individuals. Results: Three classes of individual reactions were extracted, and these were similar among those who were and those who were not physically proximate to the bombing attack: “High stress/high growth” (27% and 11%, respectively), “Low stress/high growth” (74% and 42%, respectively), and “Low stress/low growth” (only among the not physically proximate: 47%). The classes differed in terms of gender, neuroticism, and social support as well as life satisfaction and daily functioning. Conclusions: Heterogeneous patterns of posttraumatic reactions were found. Physical proximity is not necessary to experience posttraumatic stress or growth after political violence. Among individuals with low stress, posttraumatic growth may not encompass higher life satisfaction or functioning. Keywords: posttraumatic growth; posttraumatic stress; adjustment; person-centered approach
Most of the literature on reactions after a potentially traumatic event has focused on describing negative effects of being exposed to trauma. However, it has been noted that many survivors also report positive changes and growth, defined as “the experience of positive change that occurs as a result of the struggle with highly challenging life crises” (Tedeschi & Calhoun, 2004, p. 1). Experiences of growth often coexist with symptoms of distress, although the literature has been mixed with regards to how these phenomena are related (Zoellner & Maercker, 2006). One possible explanation of the mixed nature of the previous literature is that different individuals may experience fundamentally distinct patterns of stress and growth. Based on differences in personality and social resources, individuals may have different ways to cope with a traumatic event. In this sense, not everyone should experience stress and growth in similar ways after a traumatic event. According to stress and coping theory, adjustment to a demanding event depends on one’s perceived or actual resources for coping (Lazarus & Folkman, 1984). Individuals with adequate personal and social *Corresponding author. Email: [email protected] © 2014 Taylor & Francis
Anxiety, Stress, & Coping
273
resources may adjust well to a traumatic event, and may therefore experience only a short period of high levels of stress before they are able to cope with the new situation. This coping may involve experiencing posttraumatic growth, but not necessarily. Healthy adjustment after a trauma does not demand experiencing the traumatic event as beneficial. On the other hand, individuals with low levels of resources may find the traumatic situation difficult, and experience higher levels of stress. As Tedeschi and Calhoun (2004) acknowledged, growth and stress may be experienced simultaneously. However, whether this experienced growth reflects actual positive changes has been debated, and some theories have suggested that posttraumatic growth may reflect a coping mechanism rather than real changes (Zoellner & Maercker, 2006). In line with this, empirical studies indicate that posttraumatic growth is not solely associated with positive outcomes (Blix, Hansen, Birkeland, Nissen, & Heir, 2013; Frazier, Conlon, & Glaser, 2001). This suggests that posttraumatic growth does not always convey healthy adjustment in an objective sense. Taken together, prior research and theory suggest that individuals adapt to trauma in heterogeneous ways. A line of studies has explored possible subgroups of posttraumatic stress and have typically found three classes of individuals which differed in terms of symptom patterns and symptom severity – often termed as no, intermediate, and pervasive symptom disturbance (Ayer et al., 2011; Breslau, Reboussin, Anthony, & Storr, 2005; Elhai, Naifeh, Forbes, Ractliffe, & Tamburrino, 2011; Steenkamp et al., 2012). Studies on how these subgroups relate to posttraumatic growth are lacking. Moreover, to our knowledge, no previous studies have explored subgroups of posttraumatic growth. Despite being multifaceted factors thought to involve different processes and domains of change, posttraumatic stress and posttraumatic growth have often been measured as broad and general factors. This may disguise possible relationships between the phenomena that lie behind these concepts. For example, some of the posttraumatic stress clusters may be related to only some of the domains of posttraumatic growth. A metaanalytic study found that intrusive and avoidant thoughts but not anxiety was positively related to posttraumatic growth (Helgeson, Reynolds, & Tomich, 2006). Another study found that intrusion, but not avoidance and hyperarousal was associated with posttraumatic growth (Shigemoto & Poyrazli, 2013). Furthermore, symptoms of distress may relate differently to growth in different domains. A study found that posttraumatic stress disorder (PTSD) symptoms were negatively related to perceived changes of self/positive life attitude, but unrelated to philosophy of life and relating to others (Powell, Rosner, Butollo, Tedeschi, & Calhoun, 2003). This suggests that some of the confusion in the literature can be attributed to the fact that both posttraumatic stress and growth seem to be multifaceted constructs and that the different facets reflect different phenomena that have different relationships with each other. If there are different patterns of posttraumatic reactions, it should be possible to distinguish between them on related phenomena such as proximity to the traumatic event (Grieger, Fullerton, & Ursano, 2004; Grieger, Waldrep, Lovasz, & Ursano, 2005; Hansen, Nissen, & Heir, 2013) and gender (Galea et al., 2002; Linley & Joseph, 2004; Tolin & Foa, 2006). In addition, based on stress and coping theory, personal and social resources may be crucial in determining coping and adjustment after trauma. This is in line with previous research that have found that high levels of neuroticism (Cox, MacPherson, Enns, & McWilliams, 2004), low levels of optimism (Ai, 2006), and low levels of social support (Brewin, Andrews, & Valentine, 2000) are associated with high levels of
274
M.S. Birkeland et al.
posttraumatic stress. Similarly, high levels of optimism and social support have been found to be related to higher levels of posttraumatic growth (Prati & Pietrantoni, 2009). Healthy adjustment after a trauma may be operationalized as low levels of posttraumatic stress, but also life satisfaction and functioning in daily life are indicators that suggest whether an individual has succeeded in coping with the traumatic event. High levels of posttraumatic stress are associated with impaired functioning in daily life and low life satisfaction (Schnurr, Lunney, Bovin, & Marx, 2009), but associations between posttraumatic growth and life satisfaction and functioning in daily life are unclear (Blix et al., 2013). Previous studies on the relationships between posttraumatic stress and posttraumatic growth are variable-centered, which means that the relationship between posttraumatic stress and posttraumatic growth is assumed to be generally the same among all individuals. Within a person-centered approach, it is assumed that there are individual differences in relationships between variables, and these differences are of particular interest in the study. The present study investigates the patterns of reactions in ministerial employees after the 2011 Oslo bombing attack. This was a politically motivated terror attack directed toward the Norwegian Government. A car bomb explosion in the executive governmental quarter in the city center shattered the governmental buildings, killed eight people and injured 209 more people. All employees belonged to a group of people that were the target of the assailant’s operation, and were collectively confronted with an event that involved threatened death or injury, a threat to the physical integrity of self and others, as well as a destroyed workplace and work environment. Political violence threatens both individuals and their environments (Sousa, 2013), and being collectively exposed to political violence has been found to have independent effects on health when individual exposure is taken into account (Giacaman, Shannon, Saab, Arya, & Boyce, 2007). The present study aimed to identify subgroups of individuals who may show different patterns of posttraumatic reactions after a potentially traumatic event. In addition, associations with personal and social resources such as low neuroticism, high optimism and high social support as well as indicators of adjustment such as life satisfaction and daily functioning were explored.
Methods Sample and design In April 2012, approximately 10 months after the 2011 Oslo bombing attack, all employees in 14 of the 17 Norwegian ministries were invited to participate in the survey study “Mental health and work environment factors in the aftermath of the Oslo terrorist attack July 22nd, 2011” (N = 3520). In total, 1970 employees, 56% of the total sample, responded. Among these, there were 1133 women and 837 men. The respondents were divided into two subsamples. Among the respondents of the study, 200 were present in the buildings during the bomb attack, and were considered physically proximate to the bomb attack, whereas the other 1690 respondents were considered not physically proximate to the bomb attack. The response rates in the subsamples were 59% and 54% respectively. The analyses were done twice, once for each of the subsamples.
Anxiety, Stress, & Coping
275
Strict procedures were followed to ensure informed consent, as well as confidentiality. The study was approved by the Regional committees for medical and health research ethics in Norway.
Measures Posttraumatic stress symptoms were measured by the PTSD Checklist specific (PCL-S; Hem, Hussain, Wentzel-Larsen, & Heir, 2012; Weathers & Ford, 1996). This checklist consists of 17 items measuring DSM-IV PTSD symptoms linked to the bomb explosion. The participants were asked to indicate on a 5-point scale to which extent they had been bothered by the 17 symptoms the last month. To test factor structure of this measure, confirmatory factor analyses were undertaken. In line with previous studies (Elhai & Palmieri, 2011), the DSM-IV three factor model (intrusion, avoidance, hyperarousal) did not fit the data well with comparative fit index (CFI) = .90 and root mean square error of approximation (RMSEA) = .06. Following the analytic strategy outlined by Simms, Watson, and Doebbeling (2002), models with one, two, three, and four-factor solutions were tested. A four-factor solution with re-experiencing, avoidance, dysphoria, and hyperarousal as correlated subfactors fitted the data best with CFI = .942 and RMSEA = .049 (Elhai & Palmieri, 2011; Yufik & Simms, 2010). In the further analyses, the items in these subfactors were averaged and used as observed variables and input in the latent class analyses (LCAs). The Cronbach’s alphas of the subfactors re-experiencing, avoidance, dysphoria, and hyperarousal were .96, .89, .97, and .91, respectively. Posttraumatic growth was measured by a short form of the posttraumatic growth inventory (PTGI-SF; Cann et al., 2010), which is a 10-item questionnaire that assesses the degree of perceived positive change experienced after a traumatic event. PTGI-SF is thought to reflect five underlying factors, and confirmatory factor analyses revealed a moderate fit: CFI = .922 and RMSEA = .113. Because this is a theoretically grounded factor structure, this structure was kept in further analyses. These factors were averaged and used as five observed variables, and the Cronbach’s alphas of the subfactors relating to others, new possibilities, personal strength, spiritual change, and appreciation of life were .66, .61, .77, .79, and .82, respectively. Optimism was measured by the revised version of Life orientation test (LOT-R; Scheier, Carver, & Bridges, 1994). After reversing the scores for negatively formulated items, items scores were averaged to yield an overall optimism score with high scores representing high optimism. Cronbach’s alpha was .81. Neuroticism was measured by the subscale measuring this facet from the 44-item Big Five Inventory (John, Donahue, & Kentle, 1991; Rammstedt & John, 2007). After reversing the scores for positively formulated items, items scores were averaged to yield an overall neuroticism score with high scores representing high neuroticism. Cronbach’s alpha was .83. Social support was measured by four items from the Crisis Support Scale (Elklit, Pedersen, & Jind, 2001), which is a short scale measuring received social support after trauma. This scale measures both positive and negative social support, but only positive social support was considered in the present paper. The items used concerned whether the respondents experienced having someone willing to listen, being able to talk about thoughts and feelings, getting sympathy and support from others, and getting practical help. The last item regarding positive social support (regarding contact with others in a
276
M.S. Birkeland et al.
similar situation) was omitted to avoid a roofing effect, since all respondents were ministerial employees who continued to work together after the bomb attack. Cronbach’s alpha was .89. Impaired functioning was measured by the Work and Social Adjustment Scale (WSAS). The WSAS consists of five questions, scored from no impairment (0) to very severe impairment (8), which were averaged. Life satisfaction was measured by the Cantril’s self-anchoring scale (Cantril, 1965). This scale is constructed as a ladder that ranges from 1 to 10, where 1 reflects “the worst imaginable life” and 10 reflects “the best imaginable life”. Compared with other measures of well-being, this measure seems to be more sensitive to change within a respondent’s life (Larsen, Diener, & Emmons, 1985).
Statistical analyses All data analyses and modeling were performed with Mplus Version 7.11 (Muthén & Muthén, 1998–2013). To correct for the somewhat skewed distributions, maximum likelihood estimation with robust errors (MLR) was applied. To determine model fit in the confirmatory and exploratory factor analyses, chi– squared test, degrees of freedom, MLR correction factor, RMSEA, and CFI were assessed. Values of RMSEA less than .05 and values of CFI above .95 were considered to denote a well-fitting model (Browne & Cudeck, 1992; Hu & Bentler, 1999). Differences between nested models were evaluated by assessing both differences in RMSEA, CFI, and Satorra–Bentler scaled chi-square difference tests (adjusted for MLR). To examine different patterns of posttraumatic symptoms and growth, LCA were undertaken. To account for effects of proximity on psychological distress, two LCAs were undertaken; one for the subsample of individuals which were present in the epicenter of the bomb attack (in the buildings during the bomb attack), and one for the subsample of individuals which were in different proximity to the bomb attack (but not in the buildings). Latent class analysis (LCA) sorts individuals into classes based on similar patterns of responses. All nine variables were entered as indicators of one single latent variable, and models for 1–5 latent classes were computed. The models were compared using Akaike’s information criterion (AIC), Bayesian information criterion (BIC), sample-size adjusted BIC, entropy, the Lo–Mendell–Rubin likelihood ratio test (LMR-LRT), and the bootstrapped likelihood ratio test (BLRT). AIC, BIC, and SaBIC are used to compare two or more non-nested models, where the smallest value overall represents the best fit of the hypothesized model. Entropy measures the latent classification accuracy, with higher values indicating better classification. Values higher than .80 are regarded acceptable (Ram & Grimm, 2009). The LMR-LRT and BLRT are similar to chi-square difference tests, and are used to compare the relative fit of two models with different parameter restrictions. The difference between the fit of a model with k classes and a model with k–1 classes (one fewer class) is tested (Nylund, Asparouhov, & Muthén, 2007; Tofighi & Enders, 2008). Based on these indicators, the model that best represented the data was chosen. To determine effects of the predictors (gender, age, neuroticism, optimism, and social support), the automatic three-step approach provided in Mplus 7.11 was used. By using these variables as auxiliary variables and specify them as (R3STEP), the variables was
Anxiety, Stress, & Coping
277
used as latent class predictors, without changing the model (Asparouhov & Muthen, 2013). Betas and standard errors were transformed to odds ratios with 95% confidence intervals. Then, to examine differences in distal outcomes across classes, the variables (life satisfaction and impaired functioning in daily life) were used as auxiliary variables specified as (DCON). This instructs Mplus to treat these variables as distal continuous outcomes without changing the model (Asparouhov & Muthen, 2013). Means with standard errors were transformed to means with 95% confidence intervals, and differences across classes were tested with Wald equality tests.
Missing data To minimize the chance that missing data might bias the results, it is crucial to try to determine the mechanisms that lie behind missingness. Most of the missing data was caused by individual survey nonresponse. Among those who responded, the percent of missing data across all variables ranged from 0% (gender) to 5% (the last item on the PCL scale). Missingness on the last item on the PCL scale was not predicted by the mean scores on neither re-experiencing, avoidance, dysphoria, and hyperarousal clusters, which increase the probability for this missingness to be explained by a “page fatigue effect” rather than being due to differences in the variables of interests. Full information maximum likelihood (FIML) estimation with robust standard errors was used to handle missing data. This approach uses all observed information to produce the maximum likelihood estimation of parameters. This is one of the best approaches currently available to handle missing data (Graham, 2009).
Results In Table 1, demographic characteristics of the subsamples are presented. Table 2 shows the fit statistics for models with increasing number of latent classes based on posttraumatic stress and posttraumatic growth among the subsample of individuals who were and were not physically proximate to the bomb attack. Among those physically proximate, the values of AIC, BIC, and SaBIC became lower the more classes that were included. The entropy values were generally high, and the two-class model yielded the highest value. The values of LMR-LRT indicate that the two-class model was superior to the one-class model, whereas the values of BLRT did not contribute to distinguishing Table 1. Characteristics of physically proximate and not physically proximate respondents. Socio-demographic variables Gender (female) Age Education (years) More than 16 13–16 Less than 13 Leadership ns, not significant.
Physically proximate (n = 207) % (n)/mean (SD)
Not physically proximate (n = 1763) % (n)/mean (SD)
61% (126) 45.4 (10.8)
57% (1007) 44.7 (10.9)
58% 31% 12% 17%
62% 27% 12% 17%
(119) (63) (24) (33)
(1082) (464) (202) (301)
χ 2/t value 1.067 (ns) 0.793 (ns) 1.627 (ns)
0.192 (ns)
278
M.S. Birkeland et al.
Table 2. Fit of models with 1–5 latent classes, based on posttraumatic stress and posttraumatic growth among physically proximate (n = 200), and not physically proximate respondents (n = 1690). Model
AIC
BIC
SaBIC
Physically proximate respondents (n = 200) 1 class 5489.790 5549.159 5492.134 2 class 5013.202 5105.555 5016.848 3 class 4801.158 4926.494 4806.107 4 class 4693.182 4851.501 4699.432 5 class 4611.406 4802.709 4618.959 Not physically proximate respondents (n = 1690) 1 class 38523.390 38621.175 38563.991 2 class 34378.399 34530.508 34441.556 3 class 31964.565 32170.999 32050.278 4 class 31231.835 31492.594 31340.105 5 class 30602.035 30917.119 30732.860
Entropy
LMR-LRT p value
BLRT
– 0.936 0.886 0.905 0.905
– 0.0001** 0.1024 0.4419 0.0648
– 0.0000** 0.0000** 0.0000** 0.0000**
– 0.879 0.905 0.858 0.880
– 0.0000** 0.0000** 0.0406* 0.5727
– 0.0000** 0.0000** 0.0000** 0.0000**
AIC, Akaike’s information criterion; BIC, Bayesian information criterion; SaBIC, Sample-size adjusted Bayesian information criterion; LMR-LRT Lo–Mendell–Rubin likelihood ratio test; BLRT, Bootstrapped likelihood ratio test. Bold text indicate best fitting models. *p < .05, **p < .01.
between models. Based on these indices, the two-class model was considered representing the best fitting model for the respondents who were physically proximate to the bomb attack. Also among the not physically proximate respondents, AIC, BIC, and SaBIC became lower the more classes that were included. The three-class model yielded the highest entropy. The values of LMR-LRT indicate that the adding classes increased the model fit, until the fifth class was added. This suggests that a four-class model fit the data best. The values of BLRT did not contribute to distinguishing between models. Inspection of the three- and four-class models showed that the fourth class were similar to one of the other extracted classes and did not contribute significantly to the interpretation of the meaning of subgroups. Based on these indices, the three-class model was considered representing the best fitting model for the respondents who were not physically proximate to the bomb attack. The classes were named according to how they differed from the others. Clusters scores generally below 2 were considered relatively low, and cluster scores generally above 3 were considered relatively high. The means and 95% confidence intervals for the two classes among the physically proximate respondents can be seen in Figure 1. In this subsample, 27% reported high levels of all subclusters of posttraumatic stress and domains of posttraumatic growth, whereas 74% reported low levels of all subclusters of posttraumatic stress and relatively high levels of the domains of posttraumatic growth. The mean probabilities of class membership were 0.974 and 0.987, which indicate excellent discrimination among the classes. Similarly, Figure 2 presents means and 95% confidence intervals for the three classes among the not physically proximate respondents. 11% reported relatively high levels of all subclusters of posttraumatic stress and domains of posttraumatic growth, and 42% reported low levels of all subclusters of posttraumatic stress and relatively high levels of the domains of posttraumatic growth. In addition, 47% of the respondents reported relatively low levels of both stress and growth. The mean probabilities of class
Anxiety, Stress, & Coping
279
5
4.5
4
3.5
3 Low stress/high growth 74% (n = 147) 2.5
High stress/high growth 27% (n = 53)
2
1.5
1
Figure 1. Means with 95% confidence intervals for posttraumatic stress and growth for the two latent classes, “low stress/high growth” and “high stress/high growth,” among those physically proximate to the bomb attack.
5
4.5
4
3.5
3
Low stress/low growth 47% (n = 794)
2.5
Low stress/high growth 42% (n = 712)
2
High stress/high growth 11% (n = 184)
1.5
1
Figure 2. Means with 95% confidence intervals for posttraumatic stress and growth for the three latent classes, “low stress/low growth,” “low stress/high growth,” and “high stress/high growth,” among those not physically proximate to the bomb attack (n = 1690).
280
M.S. Birkeland et al.
Table 3. Odds ratios (95% CI) of memberships in latent classes “High stress/high growth,” and “Low stress/high growth” among those physically proximate to the bomb attack, according to female gender, age, optimism, neuroticism, and social support (n = 200). “High stress/high growth” vs. “Low stress/high growth” Female gender Age Optimisma Neuroticisma Social supporta
4.88** (1.66–14.34) 0.98 (0.95–1.02) 0.39 (0.08–1.84) 4.06** (1.57–10.45) 0.47* (0.23–0.96)
a
Standardized. *p < .05, **p < .01.
membership were 0.961, 0.949, and 0.961, indicating excellent discrimination among the classes. Tables 3 and 4 show odds ratios of being classified in the latent classes in the two subsamples, dependent of levels on predictors. In both subsamples, female gender, higher neuroticism and lower social support increased the probability for being classified in the “High stress/high growth” class. Age and optimism were poor predictors of class memberships. Tables 5 and 6 show means with 95% confidence intervals in distal outcomes across the classes. The “High stress/high growth” classes reported lower life satisfaction and more impairment in daily functioning than the “Low stress/high growth” and “Low stress/ low growth.”
Discussion In the present study, we examined the relationship between posttraumatic stress and posttraumatic growth in ministerial employees after a terror attack against the Norwegian Table 4. Odds ratios (95% CI) of memberships in latent classes “High stress/high growth,” “Low stress/low growth,” and “Low stress/high growth,” among those not physically proximate to the bomb attack according to direct exposure, female gender, age, optimism, neuroticism, and social support (n = 1690). “High stress/high growth” vs. “Low stress/low growth” Female gender Age Optimisma Neuroticisma Social supporta a Standardized. *p < .05, **p < .01.
“High stress/high growth” vs. “Low stress/high growth”
“Low stress/high growth” vs. “Low stress/low growth”
3.03** (1.99–4.62)
1.49 (0.97–2.28)
2.04** (1.61–2.59)
1.01 (0.99–1.03) 0.96 (0.55–1.66) 9.47** (5.00–17.91) 0.50** (0.39–0.65)
0.99 (0.97–1.00) 0.83 (0.48–1.50) 4.24** (2.36–7.59) 0.52** (0.41–0.67)
1.02* (1.01–1.03) 1.15 (0.81–1.64) 2.23** (1.50–3.32) 0.96 (0.81–1.14)
Anxiety, Stress, & Coping
281
Table 5. Means (95% CI) and Wald equality tests of life satisfaction and impaired functioning in daily life as distal outcomes of the latent classes “high stress/high growth” and “low stress/high growth” among those physically proximate to the bomb attack (n = 195). 1. High stress/high growth 2. Low stress/high growth Life satisfaction Impaired functioning
5.32 (4.83–5.80) 4.18 (3.70–4.66)
7.15 (6.91–7.34) 0.86 (0.67–1.04)
Wald equality tests (p < .05) 12
ministries. Based on the four clusters of posttraumatic stress and the five domains of posttraumatic growth, three classes of individual reactions were extracted. A class of “High stress/high growth” was found in both the physically proximate (27%) and not physically proximate (11%) subsample. Similarly, both the physically proximate and the not physically proximate subsamples included a group of individuals who reported “Low stress/high growth (74% and 42%, respectively). In addition, the not physically proximate subsample included a group of people who reported “Low stress/low growth” (47%). The combinations of “High stress/high growth” and “Low stress/low growth” show a positive linear relationship between posttraumatic stress and posttraumatic growth. This is in accordance with a hypothesis that both distress and growth may develop simultaneously after exposure to a potentially traumatic event (Tedeschi & Calhoun, 2004). This finding is also congruent with the theory that posttraumatic growth may be distorted illusions that might help people counterbalance emotional distress and promote wellbeing (Taylor & Brown, 1994; Taylor, Kemeny, Reed, Bower, & Gruenewald, 2000). The relatively large group of individuals in the “Low stress/high growth” classes suggests the possibility of another relationship between posttraumatic stress and growth. This bomb attack was directed toward individuals embedded in a highly resourceful community, namely the Norwegian governmental system. This is a high-resource community in physical environment, personal characteristics, and social capital. After this event, these resources were mobilized, and problems were addressed in an efficient way. Being a part of this, and seeing that the community is in fact a resilient community may be source of posttraumatic growth perhaps also in absence of posttraumatic stress. This is accordant with Norris, Stevens, Pfefferbaum, Wyche, & Pfefferbaum’s (2008) theory of community resilience. Table 6. Means (95% CI) and Wald equality tests of life satisfaction and impaired functioning in daily life as distal outcomes of the latent classes “high stress/high growth,” “low stress/low growth,” and “low stress/high growth” among those not physically proximate to the bomb attack (n = 1644).
Life satisfaction Impaired functioning
1. High stress/ high growth
2. Low stress/low growth
3. Low stress/high growth
Wald equality tests (p < .05)
6.10 (5.84–6.36)
7.50 (7.41–7.59)
7.46 (7.34–7.56)
1 < 2, 3
2.65 (2.36–2.93)
0.26 (0.21–0.30)
0.48 (0.41–0.54)
2
Latent classes of posttraumatic stress and growth Marianne Skogbrott Birkelanda*, Gertrud Sofie Hafstada, Ines Blixa and Trond Heira,b a
Norwegian Centre for Violence and Traumatic Stress Studies, 181 Nydalen, 0409 Oslo, Norway; b Institute of Clinical Medicine, Faculty of Medicine, University of Oslo, Oslo, Norway (Received 4 June 2014; accepted 15 August 2014) Background and Objectives: Potentially traumatic events may lead to different patterns of posttraumatic stress symptoms and posttraumatic growth. The objective of the present study was to identify subgroups with different patterns of posttraumatic reactions, and to determine whether these subgroups differed in terms of personal and social resources and indicators of adjustment. Design: This study used survey data collected 10 months after the 2011 Oslo bombing attack to investigate patterns of reactions among ministerial employees (N = 1970). Methods: We applied latent class analyses with covariates to extract subgroups of individuals. Results: Three classes of individual reactions were extracted, and these were similar among those who were and those who were not physically proximate to the bombing attack: “High stress/high growth” (27% and 11%, respectively), “Low stress/high growth” (74% and 42%, respectively), and “Low stress/low growth” (only among the not physically proximate: 47%). The classes differed in terms of gender, neuroticism, and social support as well as life satisfaction and daily functioning. Conclusions: Heterogeneous patterns of posttraumatic reactions were found. Physical proximity is not necessary to experience posttraumatic stress or growth after political violence. Among individuals with low stress, posttraumatic growth may not encompass higher life satisfaction or functioning. Keywords: posttraumatic growth; posttraumatic stress; adjustment; person-centered approach
Most of the literature on reactions after a potentially traumatic event has focused on describing negative effects of being exposed to trauma. However, it has been noted that many survivors also report positive changes and growth, defined as “the experience of positive change that occurs as a result of the struggle with highly challenging life crises” (Tedeschi & Calhoun, 2004, p. 1). Experiences of growth often coexist with symptoms of distress, although the literature has been mixed with regards to how these phenomena are related (Zoellner & Maercker, 2006). One possible explanation of the mixed nature of the previous literature is that different individuals may experience fundamentally distinct patterns of stress and growth. Based on differences in personality and social resources, individuals may have different ways to cope with a traumatic event. In this sense, not everyone should experience stress and growth in similar ways after a traumatic event. According to stress and coping theory, adjustment to a demanding event depends on one’s perceived or actual resources for coping (Lazarus & Folkman, 1984). Individuals with adequate personal and social *Corresponding author. Email: [email protected] © 2014 Taylor & Francis
Anxiety, Stress, & Coping
273
resources may adjust well to a traumatic event, and may therefore experience only a short period of high levels of stress before they are able to cope with the new situation. This coping may involve experiencing posttraumatic growth, but not necessarily. Healthy adjustment after a trauma does not demand experiencing the traumatic event as beneficial. On the other hand, individuals with low levels of resources may find the traumatic situation difficult, and experience higher levels of stress. As Tedeschi and Calhoun (2004) acknowledged, growth and stress may be experienced simultaneously. However, whether this experienced growth reflects actual positive changes has been debated, and some theories have suggested that posttraumatic growth may reflect a coping mechanism rather than real changes (Zoellner & Maercker, 2006). In line with this, empirical studies indicate that posttraumatic growth is not solely associated with positive outcomes (Blix, Hansen, Birkeland, Nissen, & Heir, 2013; Frazier, Conlon, & Glaser, 2001). This suggests that posttraumatic growth does not always convey healthy adjustment in an objective sense. Taken together, prior research and theory suggest that individuals adapt to trauma in heterogeneous ways. A line of studies has explored possible subgroups of posttraumatic stress and have typically found three classes of individuals which differed in terms of symptom patterns and symptom severity – often termed as no, intermediate, and pervasive symptom disturbance (Ayer et al., 2011; Breslau, Reboussin, Anthony, & Storr, 2005; Elhai, Naifeh, Forbes, Ractliffe, & Tamburrino, 2011; Steenkamp et al., 2012). Studies on how these subgroups relate to posttraumatic growth are lacking. Moreover, to our knowledge, no previous studies have explored subgroups of posttraumatic growth. Despite being multifaceted factors thought to involve different processes and domains of change, posttraumatic stress and posttraumatic growth have often been measured as broad and general factors. This may disguise possible relationships between the phenomena that lie behind these concepts. For example, some of the posttraumatic stress clusters may be related to only some of the domains of posttraumatic growth. A metaanalytic study found that intrusive and avoidant thoughts but not anxiety was positively related to posttraumatic growth (Helgeson, Reynolds, & Tomich, 2006). Another study found that intrusion, but not avoidance and hyperarousal was associated with posttraumatic growth (Shigemoto & Poyrazli, 2013). Furthermore, symptoms of distress may relate differently to growth in different domains. A study found that posttraumatic stress disorder (PTSD) symptoms were negatively related to perceived changes of self/positive life attitude, but unrelated to philosophy of life and relating to others (Powell, Rosner, Butollo, Tedeschi, & Calhoun, 2003). This suggests that some of the confusion in the literature can be attributed to the fact that both posttraumatic stress and growth seem to be multifaceted constructs and that the different facets reflect different phenomena that have different relationships with each other. If there are different patterns of posttraumatic reactions, it should be possible to distinguish between them on related phenomena such as proximity to the traumatic event (Grieger, Fullerton, & Ursano, 2004; Grieger, Waldrep, Lovasz, & Ursano, 2005; Hansen, Nissen, & Heir, 2013) and gender (Galea et al., 2002; Linley & Joseph, 2004; Tolin & Foa, 2006). In addition, based on stress and coping theory, personal and social resources may be crucial in determining coping and adjustment after trauma. This is in line with previous research that have found that high levels of neuroticism (Cox, MacPherson, Enns, & McWilliams, 2004), low levels of optimism (Ai, 2006), and low levels of social support (Brewin, Andrews, & Valentine, 2000) are associated with high levels of
274
M.S. Birkeland et al.
posttraumatic stress. Similarly, high levels of optimism and social support have been found to be related to higher levels of posttraumatic growth (Prati & Pietrantoni, 2009). Healthy adjustment after a trauma may be operationalized as low levels of posttraumatic stress, but also life satisfaction and functioning in daily life are indicators that suggest whether an individual has succeeded in coping with the traumatic event. High levels of posttraumatic stress are associated with impaired functioning in daily life and low life satisfaction (Schnurr, Lunney, Bovin, & Marx, 2009), but associations between posttraumatic growth and life satisfaction and functioning in daily life are unclear (Blix et al., 2013). Previous studies on the relationships between posttraumatic stress and posttraumatic growth are variable-centered, which means that the relationship between posttraumatic stress and posttraumatic growth is assumed to be generally the same among all individuals. Within a person-centered approach, it is assumed that there are individual differences in relationships between variables, and these differences are of particular interest in the study. The present study investigates the patterns of reactions in ministerial employees after the 2011 Oslo bombing attack. This was a politically motivated terror attack directed toward the Norwegian Government. A car bomb explosion in the executive governmental quarter in the city center shattered the governmental buildings, killed eight people and injured 209 more people. All employees belonged to a group of people that were the target of the assailant’s operation, and were collectively confronted with an event that involved threatened death or injury, a threat to the physical integrity of self and others, as well as a destroyed workplace and work environment. Political violence threatens both individuals and their environments (Sousa, 2013), and being collectively exposed to political violence has been found to have independent effects on health when individual exposure is taken into account (Giacaman, Shannon, Saab, Arya, & Boyce, 2007). The present study aimed to identify subgroups of individuals who may show different patterns of posttraumatic reactions after a potentially traumatic event. In addition, associations with personal and social resources such as low neuroticism, high optimism and high social support as well as indicators of adjustment such as life satisfaction and daily functioning were explored.
Methods Sample and design In April 2012, approximately 10 months after the 2011 Oslo bombing attack, all employees in 14 of the 17 Norwegian ministries were invited to participate in the survey study “Mental health and work environment factors in the aftermath of the Oslo terrorist attack July 22nd, 2011” (N = 3520). In total, 1970 employees, 56% of the total sample, responded. Among these, there were 1133 women and 837 men. The respondents were divided into two subsamples. Among the respondents of the study, 200 were present in the buildings during the bomb attack, and were considered physically proximate to the bomb attack, whereas the other 1690 respondents were considered not physically proximate to the bomb attack. The response rates in the subsamples were 59% and 54% respectively. The analyses were done twice, once for each of the subsamples.
Anxiety, Stress, & Coping
275
Strict procedures were followed to ensure informed consent, as well as confidentiality. The study was approved by the Regional committees for medical and health research ethics in Norway.
Measures Posttraumatic stress symptoms were measured by the PTSD Checklist specific (PCL-S; Hem, Hussain, Wentzel-Larsen, & Heir, 2012; Weathers & Ford, 1996). This checklist consists of 17 items measuring DSM-IV PTSD symptoms linked to the bomb explosion. The participants were asked to indicate on a 5-point scale to which extent they had been bothered by the 17 symptoms the last month. To test factor structure of this measure, confirmatory factor analyses were undertaken. In line with previous studies (Elhai & Palmieri, 2011), the DSM-IV three factor model (intrusion, avoidance, hyperarousal) did not fit the data well with comparative fit index (CFI) = .90 and root mean square error of approximation (RMSEA) = .06. Following the analytic strategy outlined by Simms, Watson, and Doebbeling (2002), models with one, two, three, and four-factor solutions were tested. A four-factor solution with re-experiencing, avoidance, dysphoria, and hyperarousal as correlated subfactors fitted the data best with CFI = .942 and RMSEA = .049 (Elhai & Palmieri, 2011; Yufik & Simms, 2010). In the further analyses, the items in these subfactors were averaged and used as observed variables and input in the latent class analyses (LCAs). The Cronbach’s alphas of the subfactors re-experiencing, avoidance, dysphoria, and hyperarousal were .96, .89, .97, and .91, respectively. Posttraumatic growth was measured by a short form of the posttraumatic growth inventory (PTGI-SF; Cann et al., 2010), which is a 10-item questionnaire that assesses the degree of perceived positive change experienced after a traumatic event. PTGI-SF is thought to reflect five underlying factors, and confirmatory factor analyses revealed a moderate fit: CFI = .922 and RMSEA = .113. Because this is a theoretically grounded factor structure, this structure was kept in further analyses. These factors were averaged and used as five observed variables, and the Cronbach’s alphas of the subfactors relating to others, new possibilities, personal strength, spiritual change, and appreciation of life were .66, .61, .77, .79, and .82, respectively. Optimism was measured by the revised version of Life orientation test (LOT-R; Scheier, Carver, & Bridges, 1994). After reversing the scores for negatively formulated items, items scores were averaged to yield an overall optimism score with high scores representing high optimism. Cronbach’s alpha was .81. Neuroticism was measured by the subscale measuring this facet from the 44-item Big Five Inventory (John, Donahue, & Kentle, 1991; Rammstedt & John, 2007). After reversing the scores for positively formulated items, items scores were averaged to yield an overall neuroticism score with high scores representing high neuroticism. Cronbach’s alpha was .83. Social support was measured by four items from the Crisis Support Scale (Elklit, Pedersen, & Jind, 2001), which is a short scale measuring received social support after trauma. This scale measures both positive and negative social support, but only positive social support was considered in the present paper. The items used concerned whether the respondents experienced having someone willing to listen, being able to talk about thoughts and feelings, getting sympathy and support from others, and getting practical help. The last item regarding positive social support (regarding contact with others in a
276
M.S. Birkeland et al.
similar situation) was omitted to avoid a roofing effect, since all respondents were ministerial employees who continued to work together after the bomb attack. Cronbach’s alpha was .89. Impaired functioning was measured by the Work and Social Adjustment Scale (WSAS). The WSAS consists of five questions, scored from no impairment (0) to very severe impairment (8), which were averaged. Life satisfaction was measured by the Cantril’s self-anchoring scale (Cantril, 1965). This scale is constructed as a ladder that ranges from 1 to 10, where 1 reflects “the worst imaginable life” and 10 reflects “the best imaginable life”. Compared with other measures of well-being, this measure seems to be more sensitive to change within a respondent’s life (Larsen, Diener, & Emmons, 1985).
Statistical analyses All data analyses and modeling were performed with Mplus Version 7.11 (Muthén & Muthén, 1998–2013). To correct for the somewhat skewed distributions, maximum likelihood estimation with robust errors (MLR) was applied. To determine model fit in the confirmatory and exploratory factor analyses, chi– squared test, degrees of freedom, MLR correction factor, RMSEA, and CFI were assessed. Values of RMSEA less than .05 and values of CFI above .95 were considered to denote a well-fitting model (Browne & Cudeck, 1992; Hu & Bentler, 1999). Differences between nested models were evaluated by assessing both differences in RMSEA, CFI, and Satorra–Bentler scaled chi-square difference tests (adjusted for MLR). To examine different patterns of posttraumatic symptoms and growth, LCA were undertaken. To account for effects of proximity on psychological distress, two LCAs were undertaken; one for the subsample of individuals which were present in the epicenter of the bomb attack (in the buildings during the bomb attack), and one for the subsample of individuals which were in different proximity to the bomb attack (but not in the buildings). Latent class analysis (LCA) sorts individuals into classes based on similar patterns of responses. All nine variables were entered as indicators of one single latent variable, and models for 1–5 latent classes were computed. The models were compared using Akaike’s information criterion (AIC), Bayesian information criterion (BIC), sample-size adjusted BIC, entropy, the Lo–Mendell–Rubin likelihood ratio test (LMR-LRT), and the bootstrapped likelihood ratio test (BLRT). AIC, BIC, and SaBIC are used to compare two or more non-nested models, where the smallest value overall represents the best fit of the hypothesized model. Entropy measures the latent classification accuracy, with higher values indicating better classification. Values higher than .80 are regarded acceptable (Ram & Grimm, 2009). The LMR-LRT and BLRT are similar to chi-square difference tests, and are used to compare the relative fit of two models with different parameter restrictions. The difference between the fit of a model with k classes and a model with k–1 classes (one fewer class) is tested (Nylund, Asparouhov, & Muthén, 2007; Tofighi & Enders, 2008). Based on these indicators, the model that best represented the data was chosen. To determine effects of the predictors (gender, age, neuroticism, optimism, and social support), the automatic three-step approach provided in Mplus 7.11 was used. By using these variables as auxiliary variables and specify them as (R3STEP), the variables was
Anxiety, Stress, & Coping
277
used as latent class predictors, without changing the model (Asparouhov & Muthen, 2013). Betas and standard errors were transformed to odds ratios with 95% confidence intervals. Then, to examine differences in distal outcomes across classes, the variables (life satisfaction and impaired functioning in daily life) were used as auxiliary variables specified as (DCON). This instructs Mplus to treat these variables as distal continuous outcomes without changing the model (Asparouhov & Muthen, 2013). Means with standard errors were transformed to means with 95% confidence intervals, and differences across classes were tested with Wald equality tests.
Missing data To minimize the chance that missing data might bias the results, it is crucial to try to determine the mechanisms that lie behind missingness. Most of the missing data was caused by individual survey nonresponse. Among those who responded, the percent of missing data across all variables ranged from 0% (gender) to 5% (the last item on the PCL scale). Missingness on the last item on the PCL scale was not predicted by the mean scores on neither re-experiencing, avoidance, dysphoria, and hyperarousal clusters, which increase the probability for this missingness to be explained by a “page fatigue effect” rather than being due to differences in the variables of interests. Full information maximum likelihood (FIML) estimation with robust standard errors was used to handle missing data. This approach uses all observed information to produce the maximum likelihood estimation of parameters. This is one of the best approaches currently available to handle missing data (Graham, 2009).
Results In Table 1, demographic characteristics of the subsamples are presented. Table 2 shows the fit statistics for models with increasing number of latent classes based on posttraumatic stress and posttraumatic growth among the subsample of individuals who were and were not physically proximate to the bomb attack. Among those physically proximate, the values of AIC, BIC, and SaBIC became lower the more classes that were included. The entropy values were generally high, and the two-class model yielded the highest value. The values of LMR-LRT indicate that the two-class model was superior to the one-class model, whereas the values of BLRT did not contribute to distinguishing Table 1. Characteristics of physically proximate and not physically proximate respondents. Socio-demographic variables Gender (female) Age Education (years) More than 16 13–16 Less than 13 Leadership ns, not significant.
Physically proximate (n = 207) % (n)/mean (SD)
Not physically proximate (n = 1763) % (n)/mean (SD)
61% (126) 45.4 (10.8)
57% (1007) 44.7 (10.9)
58% 31% 12% 17%
62% 27% 12% 17%
(119) (63) (24) (33)
(1082) (464) (202) (301)
χ 2/t value 1.067 (ns) 0.793 (ns) 1.627 (ns)
0.192 (ns)
278
M.S. Birkeland et al.
Table 2. Fit of models with 1–5 latent classes, based on posttraumatic stress and posttraumatic growth among physically proximate (n = 200), and not physically proximate respondents (n = 1690). Model
AIC
BIC
SaBIC
Physically proximate respondents (n = 200) 1 class 5489.790 5549.159 5492.134 2 class 5013.202 5105.555 5016.848 3 class 4801.158 4926.494 4806.107 4 class 4693.182 4851.501 4699.432 5 class 4611.406 4802.709 4618.959 Not physically proximate respondents (n = 1690) 1 class 38523.390 38621.175 38563.991 2 class 34378.399 34530.508 34441.556 3 class 31964.565 32170.999 32050.278 4 class 31231.835 31492.594 31340.105 5 class 30602.035 30917.119 30732.860
Entropy
LMR-LRT p value
BLRT
– 0.936 0.886 0.905 0.905
– 0.0001** 0.1024 0.4419 0.0648
– 0.0000** 0.0000** 0.0000** 0.0000**
– 0.879 0.905 0.858 0.880
– 0.0000** 0.0000** 0.0406* 0.5727
– 0.0000** 0.0000** 0.0000** 0.0000**
AIC, Akaike’s information criterion; BIC, Bayesian information criterion; SaBIC, Sample-size adjusted Bayesian information criterion; LMR-LRT Lo–Mendell–Rubin likelihood ratio test; BLRT, Bootstrapped likelihood ratio test. Bold text indicate best fitting models. *p < .05, **p < .01.
between models. Based on these indices, the two-class model was considered representing the best fitting model for the respondents who were physically proximate to the bomb attack. Also among the not physically proximate respondents, AIC, BIC, and SaBIC became lower the more classes that were included. The three-class model yielded the highest entropy. The values of LMR-LRT indicate that the adding classes increased the model fit, until the fifth class was added. This suggests that a four-class model fit the data best. The values of BLRT did not contribute to distinguishing between models. Inspection of the three- and four-class models showed that the fourth class were similar to one of the other extracted classes and did not contribute significantly to the interpretation of the meaning of subgroups. Based on these indices, the three-class model was considered representing the best fitting model for the respondents who were not physically proximate to the bomb attack. The classes were named according to how they differed from the others. Clusters scores generally below 2 were considered relatively low, and cluster scores generally above 3 were considered relatively high. The means and 95% confidence intervals for the two classes among the physically proximate respondents can be seen in Figure 1. In this subsample, 27% reported high levels of all subclusters of posttraumatic stress and domains of posttraumatic growth, whereas 74% reported low levels of all subclusters of posttraumatic stress and relatively high levels of the domains of posttraumatic growth. The mean probabilities of class membership were 0.974 and 0.987, which indicate excellent discrimination among the classes. Similarly, Figure 2 presents means and 95% confidence intervals for the three classes among the not physically proximate respondents. 11% reported relatively high levels of all subclusters of posttraumatic stress and domains of posttraumatic growth, and 42% reported low levels of all subclusters of posttraumatic stress and relatively high levels of the domains of posttraumatic growth. In addition, 47% of the respondents reported relatively low levels of both stress and growth. The mean probabilities of class
Anxiety, Stress, & Coping
279
5
4.5
4
3.5
3 Low stress/high growth 74% (n = 147) 2.5
High stress/high growth 27% (n = 53)
2
1.5
1
Figure 1. Means with 95% confidence intervals for posttraumatic stress and growth for the two latent classes, “low stress/high growth” and “high stress/high growth,” among those physically proximate to the bomb attack.
5
4.5
4
3.5
3
Low stress/low growth 47% (n = 794)
2.5
Low stress/high growth 42% (n = 712)
2
High stress/high growth 11% (n = 184)
1.5
1
Figure 2. Means with 95% confidence intervals for posttraumatic stress and growth for the three latent classes, “low stress/low growth,” “low stress/high growth,” and “high stress/high growth,” among those not physically proximate to the bomb attack (n = 1690).
280
M.S. Birkeland et al.
Table 3. Odds ratios (95% CI) of memberships in latent classes “High stress/high growth,” and “Low stress/high growth” among those physically proximate to the bomb attack, according to female gender, age, optimism, neuroticism, and social support (n = 200). “High stress/high growth” vs. “Low stress/high growth” Female gender Age Optimisma Neuroticisma Social supporta
4.88** (1.66–14.34) 0.98 (0.95–1.02) 0.39 (0.08–1.84) 4.06** (1.57–10.45) 0.47* (0.23–0.96)
a
Standardized. *p < .05, **p < .01.
membership were 0.961, 0.949, and 0.961, indicating excellent discrimination among the classes. Tables 3 and 4 show odds ratios of being classified in the latent classes in the two subsamples, dependent of levels on predictors. In both subsamples, female gender, higher neuroticism and lower social support increased the probability for being classified in the “High stress/high growth” class. Age and optimism were poor predictors of class memberships. Tables 5 and 6 show means with 95% confidence intervals in distal outcomes across the classes. The “High stress/high growth” classes reported lower life satisfaction and more impairment in daily functioning than the “Low stress/high growth” and “Low stress/ low growth.”
Discussion In the present study, we examined the relationship between posttraumatic stress and posttraumatic growth in ministerial employees after a terror attack against the Norwegian Table 4. Odds ratios (95% CI) of memberships in latent classes “High stress/high growth,” “Low stress/low growth,” and “Low stress/high growth,” among those not physically proximate to the bomb attack according to direct exposure, female gender, age, optimism, neuroticism, and social support (n = 1690). “High stress/high growth” vs. “Low stress/low growth” Female gender Age Optimisma Neuroticisma Social supporta a Standardized. *p < .05, **p < .01.
“High stress/high growth” vs. “Low stress/high growth”
“Low stress/high growth” vs. “Low stress/low growth”
3.03** (1.99–4.62)
1.49 (0.97–2.28)
2.04** (1.61–2.59)
1.01 (0.99–1.03) 0.96 (0.55–1.66) 9.47** (5.00–17.91) 0.50** (0.39–0.65)
0.99 (0.97–1.00) 0.83 (0.48–1.50) 4.24** (2.36–7.59) 0.52** (0.41–0.67)
1.02* (1.01–1.03) 1.15 (0.81–1.64) 2.23** (1.50–3.32) 0.96 (0.81–1.14)
Anxiety, Stress, & Coping
281
Table 5. Means (95% CI) and Wald equality tests of life satisfaction and impaired functioning in daily life as distal outcomes of the latent classes “high stress/high growth” and “low stress/high growth” among those physically proximate to the bomb attack (n = 195). 1. High stress/high growth 2. Low stress/high growth Life satisfaction Impaired functioning
5.32 (4.83–5.80) 4.18 (3.70–4.66)
7.15 (6.91–7.34) 0.86 (0.67–1.04)
Wald equality tests (p < .05) 12
ministries. Based on the four clusters of posttraumatic stress and the five domains of posttraumatic growth, three classes of individual reactions were extracted. A class of “High stress/high growth” was found in both the physically proximate (27%) and not physically proximate (11%) subsample. Similarly, both the physically proximate and the not physically proximate subsamples included a group of individuals who reported “Low stress/high growth (74% and 42%, respectively). In addition, the not physically proximate subsample included a group of people who reported “Low stress/low growth” (47%). The combinations of “High stress/high growth” and “Low stress/low growth” show a positive linear relationship between posttraumatic stress and posttraumatic growth. This is in accordance with a hypothesis that both distress and growth may develop simultaneously after exposure to a potentially traumatic event (Tedeschi & Calhoun, 2004). This finding is also congruent with the theory that posttraumatic growth may be distorted illusions that might help people counterbalance emotional distress and promote wellbeing (Taylor & Brown, 1994; Taylor, Kemeny, Reed, Bower, & Gruenewald, 2000). The relatively large group of individuals in the “Low stress/high growth” classes suggests the possibility of another relationship between posttraumatic stress and growth. This bomb attack was directed toward individuals embedded in a highly resourceful community, namely the Norwegian governmental system. This is a high-resource community in physical environment, personal characteristics, and social capital. After this event, these resources were mobilized, and problems were addressed in an efficient way. Being a part of this, and seeing that the community is in fact a resilient community may be source of posttraumatic growth perhaps also in absence of posttraumatic stress. This is accordant with Norris, Stevens, Pfefferbaum, Wyche, & Pfefferbaum’s (2008) theory of community resilience. Table 6. Means (95% CI) and Wald equality tests of life satisfaction and impaired functioning in daily life as distal outcomes of the latent classes “high stress/high growth,” “low stress/low growth,” and “low stress/high growth” among those not physically proximate to the bomb attack (n = 1644).
Life satisfaction Impaired functioning
1. High stress/ high growth
2. Low stress/low growth
3. Low stress/high growth
Wald equality tests (p < .05)
6.10 (5.84–6.36)
7.50 (7.41–7.59)
7.46 (7.34–7.56)
1 < 2, 3
2.65 (2.36–2.93)
0.26 (0.21–0.30)
0.48 (0.41–0.54)
2
