Compared with other international community-based studies [e.g. [25, 26]], our study found high rates of trauma exposure on both clinician-administered and self-report measures in adolescents, with the majority (86% on the KSADS and 91% on the CATS) reporting exposure to at least one traumatic event in their lifetime. These rates are consistent with previous South African studies [e.g. ].
Consistency in reporting of traumatic events was low between the measures and participants were more likely to endorse a trauma on the CATS than on the K-SADS. This may be attributable to the fact that more vicarious traumatisation as compared to directly experienced or witnessed traumas is asked about in the CATS, or to the relative privacy of the self-report format- participants may have felt more comfortable in admitting to traumatic experiences on a self-report scale which may be perceived as less intrusive .
19% of adolescents in the sample were diagnosed with PTSD on the K-SADS. This rate is comparable with the PTSD rate found in a larger sample of adolescents who were sampled in the same geographical region . The rate of 19% is, however, higher than that documented in a previous survey of which this sample constituted a sub-sample  The passage of time (i.e. more than a year between assessments) may be one reason for the higher rate of PTSD in the sub-sample. Most other South African community-based studies in adolescents (with the exception of a study by Ensink et al. , that have used self-report measures of assessment, have documented lower rates of PTSD than was found in this study.
The differing rates of PTSD between the K-SADS and the CATS (using a cut-off 27 on the CATS), suggests that this cut-off may be too high in our setting. The ROC analysis yielded a cut-off of 15 on the CATS. This cut-off maximizes both the number of true positives and true negatives and may be more appropriate. Using a cut-off of 15, 22 participants (38%) were diagnosed with PTSD. While there still remained significant differences in the rates of PTSD using this cut-off, the level of diagnostic agreement was higher than with a cut-off of 27. Our findings are consistent with studies that have demonstrated that self-report measures [e.g. [29, 30]] yield higher rates of psychiatric diagnoses than clinician-based interviews [e.g. [25, 27]]. Moreover, significant differences in CATS severity scores between participants with and without PTSD, suggests that the CATS discriminates well between those with and without the disorder.
Further, significant differences were found between mean CATS scores for Criterion B (intrusive), C (avoidance), and D (hyperarousal) symptoms between participants meeting DSM-IV criteria for these clusters on the K-SADS, and those not meeting criteria. The two symptoms that were most frequently endorsed on both the K-SADS and the CATS (recurrent thoughts/ images of event and efforts to avoid thoughts of the event) are also among the symptoms most frequently reported in other studies [11, 12], suggesting that careful inquiry of these symptoms is important. However, the level of agreement for specific symptoms appeared to be suboptimal: overall, participants who reported symptoms on the K-SADS did not necessarily report the same symptoms on the CATS. That said, participants with PTSD were more consistent in their reporting than those without PTSD. Nevertheless, the lack of significant differences in the numbers of symptoms reported between the measures suggests that these measures may be comparable in eliciting the average number of symptoms experienced post-trauma. The CATS appeared to discriminate well between those with and without PTSD on five of twelve items (recurrent thoughts about the event, exaggerated startle response, difficulty concentrating, avoidance of physical reminders of the event, and nightmares), suggesting that these symptoms may be more sensitive indicators of PTSD.
General Implications of Findings
The K-SADS and CATS yield different information about the level and type of trauma exposure, therefore researchers and clinicians should be cautious when substituting one for the other. The K-SADS is likely to yield more detailed information on witnessing traumatic events, while the CATS is likely to yield more information on vicarious trauma exposure. Adolescents are also more likely to endorse a trauma on the CATS than they are on the K-SADS. The significantly larger proportion of adolescents with scores indicative of PTSD on the CATS, compared to the K-SADS, indicates that the CATS may be better utilized as a PTSD screening device (as suggested by its author), with a cut-off threshold of 15 instead of the original threshold of 27, in the South African context. This will identify over one third of all participants with PTSD while making few false positive identifications. This will, however, require replication in a larger South African cohort. For an actual PTSD diagnosis, a clinician-based diagnostic interview may be more appropriate even though it is likely to be more time consuming.
Several limitations are worth mentioning. First, the K-SADS was not administered to both parents and learners as it is intended to be, thus participants' responses were not verified by collateral information from parents and legal guardians. Second, the sample comprised predominantly female adolescents of mixed race. Even though this constitutes the majority ethnic group in the province, the small sample and truncated age limits the generalizability to the larger population. Further, socio-demographic variables (e.g. social class, family income and race) were not accounted for in the analysis. Third, cultural influences may favour certain symptoms of trauma over others  and it has been noted that there is a need to identify other post-traumatic expressions of distress, such as somatization [32, 33]. Both the K-SADS and the CATS do not attempt to capture these experiences. However, PTSD has been widely documented in traumatized cohorts from different ethnocultural backgrounds and those from non-Western cultures who meet PTSD diagnostic criteria often show a similar clinical course and response to treatment . Fourth, we used the DSM-IV concept of trauma to compare these instruments and some authors, for example Summerfield , have highlighted some of the difficulties with the concept of trauma as defined in the DSM. It may be that events counted and endorsed as traumas were too broad to ascertain their level of agreement on the K-SADS and the CATS. Fifth, while we attempted to compare traumatic events and symptoms across instruments, it must be noted that these instruments are not necessarily suited to direct comparison. For example, the two instruments measure different traumatic events, automatically placing a cap on the level of agreement.
In view of the high levels of violence in South African youth, identification of those children and adolescents with PTSD is important and necessary to allow for appropriate interventions. Owing to limited resources, administration of diagnostic clinical interviews to all youth is not feasible. Self-report scales, even though they do not replace clinical interviews, may be useful in identifying those youth in the community who are most at risk. This may help to facilitate more targeted and efficient treatments. While this study has limitations, some tentative conclusions can nevertheless be drawn. High rates of trauma exposure and PTSD characterize South African children and adolescents. Self-report scales may be better utilized as screening instruments rather than as diagnostic tools. To establish more efficient ways of diagnosing PTSD and other post-traumatic sequelae in the South African setting, future studies (using self-rating scales and brief PTSD diagnostic measures) should be conducted in larger samples, more representative of the South African population. In particular, we need to establish and verify more suitable cut-off values on these instruments to enable the identification of those children and adolescents who are at higher risk for PTSD and other disorders.