Evaluation of the PCL-C in a Sample of Somali Refugee Youth Living in Kenya: A Multidimensional Rasch Analysis

Background: Approximately 1.1 million Somalis live in refugee camps worldwide; more than half are in urban Kenya.  Unique challenges faced by this population heighten their risk for common mental disorders, including posttraumatic stress disorder (PTSD). To develop and implement effective, culturally appropriate interventions for PTSD with this population requires adaptation of rigorous assessment instruments. The current study evaluated the suitability of the PTSD Checklist-Civilian Version (PCL-C) for assessing PTSD with a group of Somali youth exposed to urban violence in Kenya.  Specifically, we tested multiple competing latent factor structures of the PCL-C using multidimensional item response theory (MIRT).

Methods: Data for the current study were collected as part of an intervention implemented three months after escalation of violence and police exploitation in Kenya. Two hundred fifty Somali youth (56.8% female, mean age = 20 years) were recruited using snowball-sampling procedures. On average, participants had spent 7.48 (SD=4.91) years in Kenya. Along with the PCL-C, we administered a locally-developed traumatic events questionnaire and the Hopkins Symptom Checklist-25 to assess depression and anxiety symptoms. All measures were translated into Somali with the assistance of local clinic staff. We assessed the psychometric properties of the PCL-C using a multidimensional extension of the Rasch model. Seven competing models of PTSD were tested. Models were evaluated under both rating scale and partial credit models. After identifying the best-fitting model, we ran unidimensional analyses with the PCL-C.  Convergent validity of the best-fitting PCL-C model was assessed using canonical correlation between the PCL-C factors and each of two sets of variables hypothesized to be related to PTSD—traumatic events and common mental disorders (i.e., anxiety, depression, somatization).

Results: The highest response scale category was infrequently used in all analyses; therefore, response category options were collapsed to a 0-3 scale. Using the revised response scale, of the seven models we tested, MIRT results indicated that the four-factor Emotional Numbing model of PTSD (King et al., 1998; Intrusion, Avoidance, Numbing, Hyperarousal) best fit data from our sample. This model had the lowest deviance and AIC in three out of four conditions. There was no evidence of differential item functioning by sex. Support for convergent validity was found. Results of canonical correlations were supportive of the theoretically expected relationships between PCL-C subscales and both common mental disorders and trauma-related variables.

Conclusions and Implications:  Our findings support previous research linking PTSD with cumulative trauma and emotional distress among Somali populations. Although replication is necessary, our finding, regarding the superior fit of the Emotional Numbing model of PTSD, is important, as it has the potential to guide treatment modalities in this population. Future research should test the factor-structure of DSM-5 PTSD criteria with Somali refugee youth while including a cultural formulation interview. In addition, future research should examine response category use to identify the optimal rating scale for use in this population. Overall, our findings indicate the cross-cultural applicability of the PCL-C for a large population of Somali refugee youth displaced in a low resource setting.