Personality and Individual Differences 34 (2003) 315–322 www.elsevier.com/locate/paid Principal components analysis of the impact of event scale with children in war Patrick Smitha,*, Sean Perrina, Atle Dyregrovb, William Yulea a Department of Psychology, Institute of Psychiatry, De Crespigny Park, Denmark Hill, Camberwell, London SE5 8AF, UK b Center for Crisis Psychology, Fabrikkgaten 5, N-5037 Solheimsvik, Norway Received 22 June 2001; received in revised form 4 January 2002; accepted 5 February 2002 Abstract A new 13-item version of the Impact of Event Scale (Horowitz, Wilner, & Alvarez, 1979) was adminstered to 2976 9–14 year old children from Bosnia as part of a large epidemiological screening survey. The IES-13 consists of four intrusion items, four avoidance items, and five new arousal items. Factor analyses showed the scale to have an identical underlying factor structure as when it was used with British children who experienced a single-incident trauma. In addition, there was also a third factor of hyperarousal, closely related to intrusion. These results are discussed in the context of the cross-cultural validity of post traumatic stress reactions in children. # 2002 Elsevier Science Ltd. All rights reserved. Keywords: Children; War; Post traumatic stress disorder 1. Introduction Numerous studies have documented symptoms of post traumatic stress disorder (PTSD) in child survivors of war from various countries, including Cambodia (Kinzie, Sack, Angell, Manson, & Rath, 1986), the Lebanon (Saigh, 1991), Iraq (Dyregrov & Raundalen, 1992), Croatia (Kuterovac, Dyregrov, & Stuvland, 1994), Rwanda (Gupta, Dyregrov, Gjestad, & Mukanoheli, 1996), and Palestine (Thabet & Vostanis, 1999). However, controversy persists concerning the validity and utility of the post traumatic stress framework in these populations (see Bracken & Petty, 1998). Some writers (e.g. Summerfield, 1997) argue that using a western PTSD framework to investigate children’s reactions to living through war is ethnocentric, reductive, and runs the * Corresponding author. Tel.: +44-207-848-0506; fax: +44-207-848-5006. E-mail address: p.smith@iop.kcl.ac.uk (P. Smith). 0191-8869/02/$ - see front matter # 2002 Elsevier Science Ltd. All rights reserved. PII: S0191-8869(02)00047-8 316 P. Smith et al. / Personality and Individual Differences 34 (2003) 315–322 risk of pathologising and victimising whole populations. The arguments are broad and complex, but in some respects they concern the cross-cultural validity of PTSD in children. In an attempt to inform this debate, a number of workers have examined the factor structure of symptom checklists, self report questionnaires, and structured interview schedules to make inferences about the underlying construct of post traumatic stress reactions in diverse child populations. It is reasoned that while differences in absolute scores between culturally diverse groups are to be expected, measurement instruments should possess similar underlying factor structures when employed with different populations if they are indeed measuring the same construct. To date, empirical evidence is scarce. Sack, Seeley, and Clarke (1997) examined the factor structure of the Diagnostic Interview Schedule for Children and Adolescents (DICA, Welner, Reich, Herjanic, Jung, & Amando, 1987) which was used to assess a group of traumatised Cambodian refugee youth. The four factor solution (intrusion, avoidance, emotional numbing, and arousal) resembled that found in American adult populations (e.g. Watson, Kucala, Juba, Manifold, Anderson, & Anderson, 1991; Foa, Riggs, & Gershuny, 1995). This research group also examined the factor structure of the Impact of Event Scale (IES, Horowitz, Wilner, & Alvarez, 1979) as used with Cambodian youth (Sack, Seeley, Him, & Clarke, 1998). Here, a three factor solution (intrusion, avoidance, and emotional numbing) was found, identical to that reported by Yule, ten Bruggencate, and Joseph (1994) using the same instrument in their study of British child survivors of a shipping disaster. Last, Dyregrov, Kuterovac, and Barath (1996) also investigated the factor structure of the IES as used with a large group of children from Croatia during the early years of war in that country. In contrast to Yule et al. (1994), the Croatian study found a two factor solution of intrusion and avoidance, with the two items that had previously been found to form a ‘‘numbing’’ factor having poor loadings on both factors and contributing little to explaining the overall variance. The conclusion that has been drawn from these studies to date is that, with some discrepancies, post traumatic stress reactions in children are more similar across cultures than they are different. A similar data-analytic strategy to that described above is used in the present study. Here, the factor structure of a revised version of the Impact of Event Scale in children from Bosnia who survived 3 years of war is examined. Regardless of absolute scores, measurement instruments should possess similar underlying factor structures in diverse populations if they are indeed measuring the same construct. It is therefore hypothesised that the underlying factor structure of this scale when used with children from Bosnia will be similar to that when it is used with children from Britain who were exposed to a civilian peacetime disaster. Second, it is hypothesised that five additional new items relating to arousal will form a meaningful third factor. These data were collected as part of a large epidemiological screening survey for a United Nations Children’s Fund (UNICEF) psychosocial programme for children in Mostar, Bosnia, details of which have been reported elsewhere (Smith, Perrin, Yule, Hacam, & Stuvland, 2002). Briefly, children in Mostar survived over 3 years of war; and the population from the Eastern part of town were besieged for 9 months, cut off from any international military or humanitarian aid. During the long siege, including a bitter winter, many families in East Mostar lived in basements to shelter from artillery and sniper fire. It is estimated that more than 100,000 shells landed in the town during the war, that over half of the buildings on the Eastern side of town were damaged or destroyed, and that more than 1500 civilians were killed (UNICEF, 1994). Current data were collected at the beginning of 1996 (two years after the signing of the Washington Agreement P. Smith et al. / Personality and Individual Differences 34 (2003) 315–322 317 which brought a semblance of peace to the region), as an integral part of a service oriented project, in order to identify vulnerable children and to guide service planning. 2. Methods 2.1. Participants Participants were 2 976 children from Mostar, Bosnia-Herzegovina, aged between 9 and 14 years old (49% boys, 51% girls). All 9–14 year old primary school children in the municipal district were invited to participate. In general, return rates were good to excellent, ranging from 54 to 94% across 10 primary schools, with an overall return rate of 84%. The mean age of children was 12.11 years (SD=1.69). The proportion of children in each age band was similar, indicating an unbiased estimate of the population. Most children (73.4%) were born in Mostar and most (97.6%) identified as Muslim. The majority of children (84.4%) lived at home with both parents, and most of them (89.0%) also had one or more sibling living at home. 2.2. Procedure Data collection was via schools. Teachers were trained to administer the IES-13 by class. Teachers were given letters asking for informed consent from parents, and these were given to every child in the relevant grade. One week later, questionnaires were delivered to schools and administered by class, except to those children whose parents had refused. Questionnaires were collected from schools on the same day. 2.3. Materials A revised version of the Impact of Event Scale was administered as part of the screening battery (Smith et al., in press). The Bosnian language version of this measure was obtained through translation and blind back translation (Bracken & Barona, 1991). The original 15-item version of the IES (Horowitz et al., 1979) has been widely used with adults to assess the degree of subjective distress, characterised by intrusion and avoidance, following exposure to trauma. Although the IES-15 is commonly used in assessing children, there is evidence that a number of items are misinterpreted by children (Yule et al., 1994; Dyregrov et al., 1996), and so an abbreviated 8-item version has been developed (Yule, 1998; Dyregrov & Yule, 1995). Using a cutoff of 17, this shorter version of the scale has been found to work efficiently at discriminating PTSD cases, misclassifying only 10% (Dyregrov & Yule, 1995; Stallard, Velleman, & Baldwin, 1999). In order to assess the third symptom cluster of PTSD symptoms—increased physiological arousal—five new items were added. These were designed to reflect the 5 DSM-IV Cluster D symptoms, and their wording was finalized after clinic piloting. In piloting with children attending clinic, all the newly translated items were easily understood, and those children scoring highly were more likely to have PTSD. The new IES-13 thus measures symptoms of intrusion (4 items), avoidance (4 items), and arousal (5 new items). In the current translations, each item was presented in question form because clinic piloting had suggested that this was easier for children to respond to than a 318 P. Smith et al. / Personality and Individual Differences 34 (2003) 315–322 statement format. Each question was answered on the original 4-point scale (Not at all, Rarely, Sometimes, Often), scored 0, 1, 3, 5 with no reversed items. 2.4. Analysis Reliability scaling and principal components analysis was carried out using the Statistical Package for Social Sciences (SPSS, 1993). 3. Results 3.1. Reliability of translated measure For the 13-item Impact of Event Scale, Cronbachs’s alpha was 0.80; and for the eight items contributing to the 8-item short total, Cronbachs’s alpha was minimally lower at 0.75. Regarding the three subscales, Cronbach’s alpha for the intrusion items was 0.70; and for the avoidance items it was 0.73. However, lower internal consistency was found for the five new arousal items, with Cronbach’s alpha at 0.60. 3.2. Factor structure Principal components analysis was carried out first on the eight intrusion and avoidance items. After varimax rotation, two clear factors corresponding to intrusion and avoidance items emerged, accounting for 54.6% variance. Comparison data from a re-analysis of these eight items from the ‘‘Jupiter’’ shipping disaster (Yule et al., 1994) gave an equivalent two factor solution, with similar loadings for each item (see Table 1). Second, principal components analysis was carried out on the full 13-item version. The initial Varimax-rotated solution gave two factors with an Eigenvalue greater than 1 (see Table 2). This 2 Table 1 Principal components analysis of IES-8: Bosnian (Mostar) and British (Jupiter sinking) data Loading Mostar Jupiter Communality Factor 1: Factor 2: Communality Factor 1: Factor 2: intrusion avoidance intrusion avoidance 1. Did you think about it when you didn’t mean to? 4. Did you have waves of strong feeling about it? 8. Did pictures about it pop into your mind? 9. Did other things keep making you think about it? 2. Did you try to remove it from your memory? 6. Did you stay away from reminders of it? 7. Did you try not to talk about it? 10. Did you try not to think about it? 0.50 0.57 0.54 0.53 0.58 0.50 0.56 0.59 0.69 0.75 0.71 0.72 0.02 0.23 0.17 0.16 0.15 0.08 0.18 0.09 0.76 0.67 0.73 0.75 0.84 0.68 0.76 0.64 0.74 0.63 0.60 0.82 0.90 0.81 0.84 0.73 0.29 0.09 0.18 0.28 0.03 0.15 0.29 0.37 0.82 0.74 0.79 0.86 319 P. Smith et al. / Personality and Individual Differences 34 (2003) 315–322 Table 2 Principal components analysis of IES-13: Bosnian (Mostar) data, two factor solution Factor loading 1. Did you think about it when you didn’t mean to? 2. Did you try to remove it from your memory? 3. Did you have difficulties paying attention or concentrating? 4. Did you have waves of strong feeling about it? 5. Did you startle more easily or feel more nervous than you did before it happened? 6. Did you stay away from reminders of it? 7. Did you try not to talk about it? 8. Did pictures about it pop into your mind? 9. Did other things keep making you think about it? 10. Did you try not to think about it? 11. Did you get easily irritable? 12. Were you more alert and watchful even when there was no obvious need to be? 13. Did you have sleep problems? Communality Factor 1: intrusion+arousal Factor 2: avoidance 0.42 0.56 0.28 0.45 0.40 0.63 0.03 0.53 0.66 0.63 0.16 0.75 0.07 0.09 0.10 0.50 0.56 0.45 0.45 0.60 0.34 0.21 0.23 0.15 0.64 0.66 0.16 0.58 0.41 0.67 0.73 0.19 0.09 0.76 0.07 0.22 0.25 0.49 0.10 factor solution splits the scale into Avoidance and Intrusion+Arousal items. The two factors together accounted for 42.1% of the variance. Last, factor loadings were re-examined when the third factor, with an Eigenvalue of 0.93, and accounting for 7.1% variance, was included, after an examination of scree plots. As predicted, when a three factor solution was forced, the underlying structure corresponds to Intrusion (Factor 1), Avoidance (Factor 2), and Arousal (Factor 3). The only exception is item 12, an arousal item which remains loaded onto the Intrusion factor. The three factors together accounted for 49.3% of the variance (Table 3). The above analyses were carried out for boys and girls separately (cf. Dyregrov et al., 1996), but factor loadings did not differ substantially according to gender. 4. Discussion This study confirms findings from earlier work with children and youth from diverse cultures by showing that intrusion and avoidance on the IES are robust and separable factors. This result is consistent with that of Dyregrov et al. (1996), who found a two-dimensional structure to the IES15 when used with children from Croatia. It is also in line with Sack et al.’s (1998) study, which reported intrusion and avoidance as the two main factors to the IES when used with Cambodian youth. Previous studies were not entirely consistent as to the presence of a third factor on the IES-15, termed ‘‘emotional numbing’’. For the current work, those items which had been found to contribute to a putative numbing factor were omitted, and only the four items with strongest loadings onto the intrusion and avoidance factors were used. Re-analysis of these eight items 320 P. Smith et al. / Personality and Individual Differences 34 (2003) 315–322 Table 3 Principal components analysis of IES-13: Bosnian (Mostar) data, three-factor solution Factor loading Communality 1. Did you think about it when you didn’t mean to? 2. Did you try to remove it from your memory? 3. Did you have difficulties paying attention or concentrating? 4. Did you have waves of strong feeling about it? 5. Did you startle more easily or feel more nervous than you did before it happened? 6. Did you stay away from reminders of it? 7. Did you try not to talk about it? 8. Did pictures about it pop into your mind? 9. Did other things keep making you think about it? 10. Did you try not to think about it? 11. Did you get easily irritable? 12. Were you more alert and watchful even when there was no obvious need to be? 13. Did you have sleep problems? Factor 1: intrusion Factor 2: avoidance Factor 3: arousal 0.48 0.60 0.46 0.56 0.47 0.64 0.06 0.14 0.73 0.32 0.13 0.76 0.07 0.05 0.09 0.23 0.15 0.66 0.16 0.60 0.51 0.57 0.54 0.49 0.60 0.54 0.24 0.25 0.21 0.70 0.65 0.14 0.17 0.44 0.66 0.72 0.16 0.06 0.76 0.07 0.19 0.08 0.01 0.17 0.26 0.11 0.71 0.21 0.35 0.18 0.10 0.55 from the Jupiter disaster were compared with factor analysis of the same eight items for the data from Mostar, and both analyses revealed the same underlying structure of intrusion and avoidance, with similar factor loadings for each item in both groups. In line with this factor structure, internal consistency for these two subscales was satisfactory. That is, the present study confirms that intrusion and avoidance are robust factors of the IES in children from different cultures. The second part of the present study examined the existence of a third factor, hyperarousal. Five new items, relating to DSM-IV (Criterion D) symptoms and derived from clinic piloting were added to the eight items from the original adult scale to make up the IES-13. However, in factor analysis, the existence of a third ‘‘arousal’’ factor was not entirely clear-cut. When the standard criterion of including factors with eigenvalues of greater than one was used, a two factor solution was found: intrusion and avoidance were again confirmed as separable factors, but all of the arousal items loaded onto the same factor as the intrusion items. This suggests that intrusion and arousal are closely linked. Examination of scree plots showed that a three-factor solution would be valid, and when a three-factor solution was forced, a third ‘‘arousal’’ factor emerged. All items in the three-factor solution loaded onto their predicted factors, except item 12, relating to hypervigilance. Hypervigilance is likely to change its meaning when measured in a situation of ongoing threat. The suggestion that there is a meaningful third arousal factor in children’s post traumatic stress reactions is in line with studies which have examined the factor structure of standardised interview schedules (Sack et al., 1997) as completed by Cambodian youth, and of course is in accord with DSM-IV criteria. The implication from the present factor analyses however, is that intrusion and arousal are closely linked, and are not entirely independent constructs. This was also reflected in the relatively low internal consistency of the arousal subscale in the present study. Further studies are needed to ascertain whether this is also the case when the IES is P. Smith et al. / Personality and Individual Differences 34 (2003) 315–322 321 used with groups of children from other cultures. A study examining the relationship of the IES13 to diagnoses derived from structured clinical interviews in children attending a specialist trauma clinic in the UK is currently in progress. The present finding in this group of children from Bosnia of a factor structure to the IES which is similar to that found among children from other cultures suggests that post traumatic stress is not a ‘‘culture-bound’’ syndrome. This was also the conclusion of Goenjian, Yehuda, Pynoos, and Steinberg (1996), using an entirely different methodology, when working with children from Armenia after the earthquake. That group found that child survivors not only presented with cardinal symptoms of PTSD, but also showed disturbances to salivary cortisol levels and basal heart rates. The presence of these physiological markers of hyperarousal in children, which have previously been found to be typical of PTSD sufferers in North America, and which are presumably less affected by cultural influences than scores on translated self report inventories, lends support to the argument for post traumatic stress manifesting across cultures. As with the adult literature (e.g. de Silva, 1998), there is a paucity of research into the cross cultural aspects of post traumatic stress in children. The available evidence suggests that stress reactions among children, including post traumatic stress, are not culture bound. The present study shows how children from Bosnia react in ways that are more similar than different to children from other cultures. This framework for understanding children’s reactions in turn proved invaluable in assisting local professionals to set up services for war affected children. References Bracken, B. A., & Barona, A. (1991). State of the art procedures for translating, validating, and using psychoeducational tests in cross cultural assessment. School Psychology International, 12(1), 119–132. Bracken, P. J., & Petty, C. (1998). Rethinking the trauma of war. London: Save the Children/Free Association Books. de Silva, P. (1998). Cultural aspects of post traumatic stress disorder. In W. Yule (Ed.), Post traumatic stress disorders: concepts and therapy. Chichester: John Wiley. Dyregrov, A., Kuterovac, G., & Barath, A. (1996). Factor analysis of the impact of event scale with children in war. Scandinavian Journal of Psychology, 36, 339–350. Dyregrov, A., & Raundalen, M. (1992). The impact of the Gulf war on the children of Iraq. Paper presented at the International Society for Traumatic Stress Studies World Conference, ‘Trauma and Tragedy’ Amsterdam, June. Dyregrov, A., & Yule, W. (1995). Screening measures: the development of the UNICEF screening battery. Paper at The Fourth European Conference on Traumatic Stress (May), Paris. Foa, E. B., Riggs, D. S., & Gershuny, B. S. (1995). Arousal, numbing, and intrusion: symptom structure of PTSD following assault. American Journal of Psychiatry, 152, 116–120. Goenjian, A. K., Yehuda, R., Pynoos, R. S., & Steinberg, A. M. (1996). Basal cortisol, dexamethasone suppression of cortisol, and MHPG in adolescents after the 1988 earthquake in Armenia. American Journal of Psychiatry, 153, 929– 934. Gupta, L., Dyregrov, A., Gjestad, R., & Mukanoheli, X. (1996). Trauma, exposure, and psychological reactions to genocide among Rwandan refugees. Paper presented at the 12th Annual Convention of the International Society for Traumatic Stress Studies (November), San Francisco, CA. Horowitz, M. J., Wilner, N., & Alvarez, W. (1979). Impact of event scale: a measure of subjective stress. Psychosomatic Medicine, 41, 209–218. Kinzie, J. D., Sack, W. H., Angell, R. H., Manson, S., & Rath, B. (1986). The psychiatric effects of massive trauma on Cambodian children: I. The children. Journal of the American Academy of Child and Adolescent Psychiatry, 25, 370– 376. 322 P. Smith et al. / Personality and Individual Differences 34 (2003) 315–322 Kuterovac, G., Dyregrov, A., & Stuvland, R. (1994). Children in war: a silent majority under stress. British Journal of Medical Psychology, 67, 363–375. Sack, W. H., Seeley, J. R., & Clarke, G. N. (1997). Does PTSD transcend cultural barriers? Journal of the American Academy of Child and Adolescent Psychiatry, 36, 49–54. Sack, W. H., Seeley, J. R., Him, C., & Clarke, G. N. (1998). Psychometric properties of the Impact of Events Scale in traumatised Cambodian refugee youth. Personality and Individual Differences, 25, 57–67. Saigh, P. A. (1991). The development of posttraumatic stress disorder following four different types of traumatization. Behavior Research and Therapy, 29, 213–216. Smith, P., Perrin, S., Yule, W., Hacam, B., & Stuvland, R. (2002). War exposure among children from Bosnia-Hercegovina: psychological adjustment in a community sample. Journal of Traumatic Stress, 15(2), 147–156. SPSS. (1993). Base system user’s guide: release 6.0. Chicago, IL: SPSS Inc. Stallard, P., Velleman, R., & Baldwin, S. (1999). Psychological screening of children for post traumatic stress disorder. Journal of Child Psychology and Psychiatry, 40(7), 1075–1082. Summerfield, D. (1997). The social, cultural, and political dimensions of contemporary war. Medicine, Conflict, and Survival, 13(1), 3–25. Thabet, A. A. M., & Vostanis, P. (1999). Post traumatic stress reactions in children of war. Journal of Child Psychology and Psychiatry, 40(3), 385–391. UNICEF. (1994). Mostar: situational analysis. Sarajevo: Unicef. Watson, C. G., Kucala, T., Juba, M., Manifold, V., Anderson, P. E. D., & Anderson, D. (1991). Factor analysis of the DSM-III post traumatic stress disorder criteria. Journal of Clinical Psychology, 47, 205–214. Welner, Z., Reich, W., Herjanic, B., Jung, K. G., & Amado, H. (1987). Reliability, validity, and parent-child agreement studies of the Diagnostic Interview for Children and Adolescents (DICA). Journal of the American Academy of Child and Adolescent Psychiatry, 26, 649–653. Yule, W. (1998). Anxiety, depression, and post traumatic stress disorder in children. In I. Sclare (Ed.), The NFER child portfolio. Windsor: NFER-Nelson. Yule, W., ten Bruggencate, S., & Joseph, S. (1994). Principal components analysis of the Impact of Event Scale in adolescents who survived a shipping disaster. Personality and Individual Differences, 16, 685–691.