Posttraumatic stress and symptom improvement in Norwegian tourists exposed to the 2004 tsunami – a longitudinal study

Norwegian tourists who had resided in a country affected by the 2004 tsunami were repatriated in the days following the tsunami and registered by the police upon their arrival in Norway. Permission from the Norwegian Data Inspectorate (Datatilsynet) and the Regional Committee for Medical Research Ethics (Regional Etisk Komité for medisinsk forskning) was obtained to make this information available for our study.

A postal questionnaire was sent to all the registered individuals aged ≥18 years (n = 2468) at two points of time: 6 months post-disaster (T1) and 24 months post-disaster (T2). A unique tracking number was attached to the questionnaires to prevent them from getting mixed up. Our population consisted of those who responded at both assessments (n = 674, see Figure 1). We sent a reminder for returning the first questionnaire few weeks after the initial distribution. At T2 the questionnaire was shortened and we sent two reminders following the initial distribution.

An information letter was sent along with the postal questionnaire. The letter contained information regarding the study and about methods for handling and storing data confidentially and stated that participation was voluntary. The persons were asked to fill in and return the questionnaire if they agreed with the terms and conditions mentioned in the information letter.

The mean age in the study sample was 43.0 years (SD = 13.0), 46.7% were men, 58.9% had ≥13 years of education, 73.4% were employed (before tsunami) and 71.8% were married or cohabitating (before tsunami) (Table 1).

The questionnaires included a variety of commonly used variables used in post disaster mental health research covering pre-, peri-, and post-disaster aspects [1, 17]: previous mental health, personality traits, disaster exposure, loss of loved ones, peritraumatic fear, social support, seeking professional help and exposure to other adverse life events. A more detailed discussion regarding the selection of the variables is given in a previous publication [21].

Responders and non-responders were compared by using independent t-test for continuous variables and Chi-square test for categorical variables. Paired samples t-test was used to examine changes in symptom scores from T1 to T2.

We used hierarchical multiple linear regression analysis to determine the adjusted effects of potential predictors of IES-R score at T1. Assumptions for linear regression were tested and found satisfactory [31]. This analysis was performed in four steps where each step contained a group of variables arranged in a theoretical and practical manner: Step 1 included socio-demographic variables such as age, gender, education, employment, marital status, and pre-tsunami mental health problems. Step 2 constituted the additional inclusion of exposure and peritraumatic fear. Step 3 included Big Five personality traits and in step 4 social support was included. To identify variables associated with symptom improvement the difference between IES-R scores at T2 and T1 were entered as dependent variable in a similar hierarchical multiple linear regression model. We wished to control for IES-R score at T1 (step 2). Thus, variables regarding exposure and peritraumatic fear were left out because they did not make any practical contribution to the model. In addition, these variables gave little clinical meaning as high levels of exposure and peritraumatic fear were associated with improvement. In the last step additional information regarding consulting a mental health specialist and experiencing additional adverse life events was included.

All analyses were conducted using the software Statistical Package for the Social Sciences (SPSS, version 16.0). All the categorical variables were dichotomized before running the analysis (Table 1). P-values <0.05 were considered significant.

Table 2 shows predictors of post traumatic stress at 6 months post-disaster (T1). In the final four step model female gender, increasing age, unemployment, disaster exposure i.e. witnessing death or suffering, being chased or caught by the waves and loss of loved ones, peritraumatic fear, low conscientiousness, neuroticism and low levels of social support were associated with posttraumatic stress. Peritraumatic fear, witnessing death or suffering and neuroticism were the strongest predictors of posttraumatic stress at T1 in the final model. The total explained variance was 43% (R = 0.65).

The total IES-R score declined from 24.6 (SD = 18.5) at T1 to 22.7 (SD = 18.3) at T2, p <0.001. At cluster level, we found a significant reduction in both intrusion and hyperarousal symptoms with the greatest reduction in the intrusion score (Figure 2). There was no significant reduction in avoidance score. According to caseness based on IES-R total score ≥33, the majority were identified as non-case at both assessments (57.7%). About 11 percent of the respondents went from being case at T1 to non-case at T2. On the other hand 7.4% went from being non-case at T1 to case at T2. A considerable number of respondents were identified as case at both assessments (20.8%) with relatively high IES-R total scores, 49.3 (SD = 12.2) at T1 and 48.4 (SD = 12.1) at T2, p = 0.352.

Regression analyses showed that high IES-R score at T1 and emotional stability were both positively related to symptom improvement from 6 to 24 months (Table 3). Having been referred to a mental health specialist was negatively related to symptom improvement. We further explored the effect of having been referred to a mental health specialist by analysing symptom change in a subpopulation with high level of posttraumatic stress at T1 (IES-R cut-off ≥33). Significant symptom improvements were found whether participants were referred to a mental health specialist or not (Table 4). In this subpopulation there was no significant difference in symptom improvement between those who had been referred to a mental health specialist and those who had not (p = 0.269).

In agreement with many studies of natural disasters, we found that disaster exposure and peritraumatic reactions were among the strongest predictors of posttraumatic stress [3, 32‐35]. The results of our study also are comparable to studies of Thai and Sri Lankan tsunami survivors [36‐38]. We found that witnessing abandoned children, seriously injured persons or dead bodies were more strongly related to posttraumatic stress than exposure to water or loss of loved ones. The pathologizing effect of horror is well documented in studies of children witnessing domestic violence [39] and soldiers witnessing atrocities [40, 41], but is often given less importance when assessing exposure in disaster survivors [42]. Witnessing horror was also a key variable in predicting general psychological distress in a previous study of Norwegian tourists [21]. Our findings indicate that witnessing horror can be just as, or even more, traumatizing than the exposure to danger. Exposure to grotesque sights should therefore be taken into account when conducting surveys of disaster survivors [43]. It is still unclear whether the impact of witness exposure is more severe as a part of survival than persons who are only exposed to such grotesque sights in the aftermath. Some studies do suggest that witness exposure without danger to life constitutes a lesser risk for enduring posttraumatic stress symptoms [21, 44].

We found a slight but significant decrease in the IES-R sum score from 6 months to 24 months post-disaster. The decrease in posttraumatic stress was more modest than generally expected after natural disasters [4, 33]. This expectation was particularly well-founded as we studied a population who escaped secondary stressors and returned to stable home communities after a short period of time [19, 20, 45]. It is likely that among many tourists the duration of significant posttraumatic stress reactions were less than 6 months, thus resulting in little additional decrease in symptom level from 6 months to 24 months post-disaster. In addition, many of the included persons were not directly exposed to the dangers of the tsunami and may have experienced little or no symptoms in the aftermath. However, the finding also indicates that many who were markedly exposed and experienced personal life threat or suffered loss developed enduring posttraumatic stress symptoms.

At first look our results may presume that Norwegian tourists experienced little variation in level of posttraumatic stress from T1 to T2. However, based on status of being case (IES-R total score ≥33), we could identify various symptom trajectories. This is in agreement with other follow-up studies of posttraumatic stress symptoms [37, 46] and with the understanding that persons with enduring symptoms often experience a fluctuating course rather than a stable and consistent level of posttraumatic stress [47‐49].

We found that consulting mental health specialists did not facilitate symptom improvement, which was somewhat surprising as we expected that treatment would alter the course in favor of symptom reduction. However, similar findings were reported in a follow-up study after the NYC terror attacks [50]. The National comorbidity data showed that in more than one third of the affected individuals the symptoms persisted for many years even when treatment was given [13]. It is argued that symptoms may be maintained and triggered by day-today adverse life experience. In addition, it is likely that secondary trauma stressors, such as loss of home or livelihood, play an important role in development of enduring posttraumatic stress following natural disasters [51]. However, the majority of our population was protected from secondary adversities as they were able to return to unaffected home communities.

One may argue that the regression analysis (Table 3) solely represent the fact that persons referred to a mental health specialist are those with highest levels of posttraumatic stress. Further investigation of this finding did show a slightly higher symptom level at T1 among persons referred to a specialist (Table 4, analyses restricted to persons with IES-R cut-off ≥33). However, in this subpopulation (IES-R cut-off ≥33), there was no significant difference in symptom improvement between persons referred and not referred to a mental health specialist. These findings do not support a negative relation between referral and symptom improvement, but they do indicate lack of efficacy of treatment given in specialized mental health care. Nevertheless, these findings must be interpreted with caution as our study was not designed as an effect-study.

It is important to bear in mind that we asked the participants whether they had been referred to a mental health specialist instead of asking directly about receiving treatment in the specialist healthcare. The question was chosen to include people who were referred to a specialist, but had not yet received treatment at the time of the first survey (T1). One may question whether the patients in this study actually received proper treatment or were just referred to treatment. However, a considerable part of those referred to treatment was satisfied by the support received from psychiatric services, which indicates that they actually received some kind of treatment [52]. Furthermore, if some persons declined the treatment offered as a result of symptom improvement, this would not change our findings regarding lack of treatment efficacy.

We did not measure details about the treatment received in specialist mental health care, but other studies of Norwegian tsunami survivors imply that common treatment received in specialised setting was supportive psychotherapy with or without the use of psychotropic drugs [53]. Our results may question the effectiveness of psychiatric treatment of enduring posttraumatic stress symptoms. It also raises concern about a possible negative effect of psychiatric treatment on symptom improvement [54, 55]. It has been suggested that in some settings treatment may enhance perceived helplessness and result in decreased coping self-efficacy [56, 57].

Our finding that social support was inversely related to posttraumatic stress at T1 is consistent with previous findings [2, 5]. However, social support was not related to symptom improvement from T1 to T2. Our finding indicates that the main positive effect of social support regard to posttraumatic stress is pronounced in the temporal proximity to the disaster. This is in line with previous findings where initial levels of social support did not predict the course of chronic PTSD symptoms [58]. The diminishing effect of social support in our study may be explained by persisting symptoms of avoidance and emotional numbing (Figure 2), which may diminish the supportive efforts of others, result in less social support or even deteriorate social relations [59].

In this study emotional stability proved consistently related to both low levels of posttraumatic stress at T1 and symptom improvement from T1 to T2. This could be expected as individuals with low levels of emotional stability (neuroticism) tend to respond emotionally negative to threat or loss [60]. The neuroticism subscale in Big Five Inventory (BFI) overlap to some extent with measures of posttraumatic stress (i.e getting easily stressed, irritated and having difficulties concentrating). Due to the study design it may be difficult to be sure of whether the measured personality traits are actual signs of neuroticism or merely represent symptoms of posttraumatic stress. However, instructions given for filling in IES-R scale largely differs from the instructions given for BFI. The first scale focuses on trauma-related symptoms experienced past seven days, while the latter scale queries whether a specific trait fits the person in general. In addition, a relation between neuroticism and posttraumatic stress has been reported in several other studies [11, 61].

It was found that experience of other stressful life events was negatively related to symptom improvement (Table 2, p = 0.054). This finding may partly explain the strong inverse relation between neuroticism and symptom improvement in the present study as individuals with high neuroticism are likely to experience negative life events and tend to show an increased level of emotional reactivity to those events [62]. Consistent with previous studies our results imply that neuroticism plays an important role in the maintenance of long term posttraumatic stress symptoms [33, 63]. Neuroticism is related to a variety of mental and physical disorders, partially mediated by inappropriate coping strategies [62, 64].

This study had some methodological advantages. Almost all Norwegians who were tourists in the disaster area were invited to participate, reducing sample selection bias. Both women and men in all age groups were represented in the study population. The participants were similar to the age- and gender-adjusted Norwegian population with regard to employment and marital status. Educational level, however was higher than average [22, 65]. The relatively fast evacuation of the Norwegian tourists reduced the impact of secondary disaster stressors, limiting the impact of trauma to the primary tsunami experience. Finally, regardless of impairment level, our sample received affordable and easily accessible medical and psychiatric care as well as ample community support [52, 66].

Limitations of our study include relatively low response rates, which imply interpreting the results with cautionous. Although 75% of T1 participants also participated at T2, our study population constituted only 27% of the total sample. Selective participation may have biased the results towards increased morbidity interfering with variables of interest like personality traits, coping behaviour and seeking treatment. Due to the directionality of the dropout bias, the included participants seem to represent most of the heavily exposed Norwegian tourists in the tsunami-stricken areas [22]. Thus the prevalence rates of cases based on high symptom score (IES-R total score ≥33) may be somewhat overestimated as compared to the whole adult population of Norwegian tourist who experienced the tsunami. However, the main focus of the current study was to investigate how variables were correlated to posttraumatic stress which is less likely to be affected by systematic response rate biases.

The information was gathered by the use of postal questionnaires. Thus, participants from the same household may have interacted during the filling-in process. This may have influenced the results [67].

Data regarding disaster exposure and peritraumatic reactions were ascertained retrospectively and may be affected by recall bias. Generally, the risk of recall bias increases as the time passes after the index traumatic event. A previous tsunami study showed that the memories for stressful events may amplify as time passes [22]. The study also showed that individuals with recall amplification did not improve in PTSD symptom severity from 6 to 24 months, whereas individuals who did not have recall amplification showed a reduction in PTSD symptoms over time. For this reason we have avoided the use of variables solely based on personal appraisal, and rather included more objectively defined criteria when measuring exposure.

Our results may not be applicable to other populations that were exposed to chronic stressors.

The fact that the participants were abroad on Christmas holiday suggests that they may represent a more privileged and healthy sub-population of the Norwegian population. However, we do not know in what way this could have affected the development and course of posttraumatic stress symptoms in the aftermath of the tsunami disaster.