Non-Western sample

Terrorism is now a global concern. Every day there are headlines reporting terrorist activities in industrialised Western nations and elsewhere, including Africa and Asia. Despite the magnitude of this threat to worldwide mental health, there are few published studies that systematically determine the prevalence of – and risk factors for – PTSD following a terrorist attack. One relevant paper (Reference Pfefferbaum, North and FlynnPfefferbaum et al, 2001) showed a close relationship between injury and post-traumatic stress in a directly exposed group of 21 individuals 8 months following the simultaneous US embassy bombing in Dar es Salaam, Tanzania, in which 11 people were killed and 80 wounded. As in our study, the people surveyed in this convenience sample were highly exposed to the traumatic event: a mean score of 6.7 (s.d.=1.9) out of a possible 8 on their measure of hearing and feeling the explosion, and a 31% (6 of 21) injury rate, with one individual rating his/her injury 3 (‘some injury’) out of 4 and five others rating theirs 2 (‘a little injury’) out of 4. Initial reactions, such as ‘nervous or afraid’, ‘felt helpless’ and ‘thought I would die’, were rated on a scale of 1 to 5. There was also a battery of questions on post-traumatic stress symptoms, e.g. ‘I had dreams about it’, ‘I tried not to talk about it’ and ‘I felt watchful and on guard’. In this traumatised sample, injury significantly predicted post-traumatic stress symptoms (R ²=0.21) in general as well as intrusion and arousal symptom clusters in particular; however, injury did not predict avoidance/numbing symptoms at 8 months. The measures of hearing and feeling the explosion and initial reaction did not predict post-traumatic stress symptoms or symptom clusters. In our larger sample, however, presence of injury, witnessing the explosion and peritraumatic reactions of fear, helplessness and feeling threatened all significantly predicted PTSS within 3 months of the incident.

The DSM–IV criteria for PTSD have not been fully validated in developing countries. Despite the growing literature on post-traumatic morbidity in Africa, which describes results similar to those in Western studies, it is possible that a different set of symptoms would better represent psychiatric impairment after trauma among, for instance, middle-class Kenyans. Jenkins (Reference Jenkins, Marsella, Friedman and Gerrity1996), after finding certain PTSD criteria inapplicable among Salvadoran women, suggested some criterion modifications for different cultures. In the case of African populations, somatic symptoms such as intense heat (‘central heat’; Reference IfabumuyiIfabumuyi, 1981) or the social repercussions of numbing might be more salient indicators of PTSD. The roles of dreams, ancestors, witches and fate may need to be assessed with appropriate terminology. Alternatively, the concept of PTSD could be a cultural category fallacy (cf. Reference KleinmanKleinman, 1977), in the sense that no such diagnostic entity, as configured in DSM–IV, for example, exists outside of Western industrialised nations. Although by no means conclusive, the similarity between findings in Western and non-Western studies of PTSD argues against this, and suggests that the DSM–IV criteria at least approximate a universal phenomenon.

Sample size

In addition to its short-term, non-Western focus, the current study is unique for its large sample size (n=2627). Even in this age of telephone interviews and web-based surveys – neither of which were feasible in this investigation – there are only three other studies of terrorist-related PTSD with populations exceeding 2000: Pfefferbaum et al (Reference Pfefferbaum, Doughty and Reddy2002) on children's response to the Oklahoma City tragedy; Silver et al (Reference Silver, Holman and McIntosh2002) on the reactions to the disaster of 11 September 2001 of Americans living outside of New York; and Schlenger et al (Reference Schlenger, Caddell and Ebert2002) on the reactions to the latter tragedy of a US national sample including New Yorkers.

Risk factor significance

Another advantage of our investigation is its clear findings on risk factors for attack-related PTSD symptoms, concerning which other research findings are not always consistent. Most importantly, the Nairobi data show a strong link between injury and PTSS (P<0.0001). Other studies showing the importance of injury severity include those of Pfefferbaum et al (Reference Pfefferbaum, North and Flynn2001), described above; Abenheim et al (Reference Abenheim, Dab and Salmi1992), on 254 survivors of terrorist attacks in France; and Wilson et al (Reference Wilson, Poole and Trew1997), on police officers in Ireland who witnessed terrorist acts. On the other hand, Curran et al (Reference Curran, Bell and Murray1990) demonstrated an inverse relationship between injury severity and PTSD for 26 people involved in the 1987 Enniskillen bombing in Northern Ireland, and Tucker et al (Reference Tucker, Pfefferbaum and Nixon2000), studying 85 individuals after the Oklahoma City bombing, showed a significant association between injury and PTSD in a univariate analysis that was not significant in the multivariate analysis. Several factors consistently reported to be associated with PTSD were further confirmed in our analysis. For example, measures of proximity to the event (in our study, being outside v. inside a nearby building) have predicted PTSD in several studies of terrorist attacks: Galea et al (Reference Galea, Ahern and Resnick2002a ,Reference Galea, Resnick and Ahern b ), in a telephone survey of 988 Manhattanites, demonstrated the significance of residence below Canal Street on 11 September; and Bernard et al (Reference Bernard, Baron and Mueller2002) showed higher PTSD rates among staff at two schools near the World Trade Center site compared with staff at two New York City schools over 8 km away. Other variables found to be significant in our study that typically correlate with PTSD after terrorist attacks include female gender, not being married and various negative long-term sequelae such as increased financial difficulty or losing possessions in the attack (Reference Easton and TurnerEaston & Turner, 1991; Galea et al, Reference Galea, Ahern and Resnick2002a ,Reference Galea, Resnick and Ahern b ). Some risk factor variables were interrelated: for example, there was an association between financial loss and both anticipating financial difficulty and receiving assistance.

Prevalence

The prevalence of PTSS in this recently traumatised civilian cohort – 35% (or 24%, when peritraumatic feelings are included in the analysis) – is comparable with data from other terrorist incidents. In his evidence-based review prior to the 11 September disaster in the USA, Gidron (Reference Gidron2002) showed a mean post-attack PTSD prevalence of 28%; this value was derived from six US studies and was unduly affected by one study of police officers (PTSD prevalence 5%), small samples, and variability in sampling, timing and assessment. Estimates of PTSD prevalence following the 11 September disaster in large samples of US populations have been reported as 7.5% in Manhattan and 20% below Canal Street in New York at 1–2 months (Reference Galea, Ahern and ResnickGalea et al, 2002a ); 7.5% in Manhattan, as well as 24% with increased smoking and 36% with increased marijuana use at 1–2 months (Reference Vlahov, Galea and ResnickVlahov et al, 2002); 9% in Manhattan at 1–2 months (Reference Galea, Resnick and AhernGalea et al, 2002b ); 11% in New York City and 4% in the rest of the USA at 1–2 months (Reference Schlenger, Caddell and EbertSchlenger et al, 2002); 17% at 2 months and 6% at 6 months in the US population outside New York City (Reference Silver, Holman and McIntoshSilver et al, 2002); and 15–23% near the site and 6–8% over 8 km away at 4–6 months (Reference Bernard, Baron and MuellerBernard et al, 2002). Hence, it is not surprising that our group composed primarily of directly exposed civilians assessed at 1–3 months had a PTSS prevalence of roughly 35%.

Convenience sample

One of the shortcomings of our study is its lack of randomisation. Operation Recovery's primary objective in the months following the incident was service delivery and there was no opportunity to form randomised groups. Participants were recruited at nearby businesses, at a clinic devoted to people affected by the blast and at public gatherings such as the annual agricultural exposition. The profile of the resulting cohort was skewed towards highly exposed, educated professionals. Respondents might have been seeking help for high levels of distress, and a bias towards increased trauma in a sample would naturally inflate the prevalence of PTSD. However, this shortcoming is shared by many studies: with the exception of the investigation of the responses of 2000 schoolchildren to the Oklahoma bombing (Reference Pfefferbaum, Doughty and ReddyPfefferbaum et al, 2002), no large, randomised study of PTSD after a terrorist attack had been conducted until the telephone and internet surveys conducted after the New York 11 September tragedy. Notably, the results of these surveys agree in many ways with the data from our Kenyan sample. Also, we made a concerted attempt to include everyone who had been near to the blast, by contacting every office in the surrounding city blocks and inviting workers who were present during the blast to participate in the study. Although this predisposed to a large proportion of educated professionals in the sample, it provided an accurate reflection of the population present on a weekday morning in the centre of the business district. These middle-class participants were local Black Kenyans.

Questionnaire

A second weakness of our study is its use of an unvalidated psychometric instrument. The questionnaire was constructed in the first month following the embassy bombing with the help of international trauma experts and it went through many revisions before it was used to collect data. Incidentally, virtually no one required assistance with comprehending the English-language questionnaire. Nevertheless, an opportunity was lost to use a more standard trauma scale or some other tool based on DSM–IV. It should be noted, however, that the reported studies on this topic have made use of a wide variety of non-standard measures of post-traumatic stress, and that commonly used standard instruments for assessing PTSD have not yet been validated in African populations. The unvalidated self-report instrument in this study, then, allowed only an approximation of PTSD caseness and not a diagnosis. Questions in one section of the survey did roughly correspond to the DSM–IV criteria and so a PTSS algorithm (one or more re-experiencing symptoms, three or more avoidance/numbing symptoms and two or more arousal symptoms) was plausible. This algorithm did not include retrospectively reported criterion A (exposure and peritraumatic reaction) symptoms, nor did it address decline in function, overall subjective distress or (unequivocally) duration of symptoms. However, our procedure was somewhat validated by the fact that there was a strongly significant association (P<0.0001) between PTSS and each of the criterion A reaction questions (feeling afraid, helpless or threatened during the event). Furthermore, 91% of this sample experienced the explosion directly, which addresses the exposure aspect of DSM–IV criterion A. Regarding decline in function or increased distress, many respondents were seeking mental health treatment either at a clinic or at their workplace. Sequelae such as bereavement, financial difficulties and receiving assistance were closely aligned with PTSS. Once again, the small amount of published research on this subject includes several examples of the use of similar methods to measure post-traumatic stress. Last, although other anxiety and depressive disorders are important sequelae of trauma, appropriate testing for these disorders would have made the (already long) questionnaire impracticably lengthy, and this might have led to an underestimation of post-traumatic morbidity.

Missing responses

A third problem with our study is the unfavourable response rate. Lower levels of trauma exposure and of serious injury characterised the 256 people who were omitted, thereby possibly raising the actual prevalence of PTSS above that observed. A few important variables were removed from the analysis owing to poor response (e.g. changes in sexual relationship). Some interesting results presented above had upwards of one-third non-response (e.g. having felt afraid, helpless or threatened; numbers of children and dependants) and so are difficult to interpret. Related to the issue of non-response in surveys is the problem of accuracy. This was most pronounced in our sample in the responses on substance use. Vlahov et al (Reference Vlahov, Galea and Resnick2002) in a telephone survey of 988 Manhattan residents confirmed the intuitive notion that cigarette, alcohol and marijuana use increased after the World Trade Center attacks and their use was linked to cases of PTSD and depression. Our investigation – with only about 15% missing data on this subject – showed not only low rates of increased use (alcohol 5%, smoking 3%, illicit drugs 1%) but also unusually high rates of reported abstinence (alcohol 61%, smoking 73%, illicit drugs 78%). If the tendency of this population was to underreport symptoms in other potentially stigmatising areas such as post-traumatic stress criteria, then the already highly traumatised people in this sample may be more troubled than the descriptive statistics indicate.

The 35% prevalence of significant PTSD symptoms in this highly exposed sample a few months after the Nairobi bombing is comparable with the prevalence found in studies of Western populations affected by terrorism. Likewise, frequently reported predictors of PTSD such as female gender, injury, peritraumatic response and financial sequelae were confirmed in this large non-Western sample. It will be important to learn more from this cohort, and from the unfortunately growing number of similar groups, so that we can further refine our diagnoses, identify those at greatest risk and effectively treat the victims of terrorism.