Methods: Participants were recruited via snowball sampling and included 331 young Somali refugees, aged 13 to 35 years, who were living in Nairobi (M =25.06, SD=19.35; 55% female). All somatization items were developed and translated into Somali via the translation-back-translation method with the assistance of local community health counselors and physicians. The final item pool consisted of 13-items. Seven were locally developed to capture culturally-relevant somatic idioms of distress; six indicators were adapted from established Western measures of depression, anxiety, and posttraumatic stress disorder (PTSD) symptoms (HSCL-25; SCL-90; PCL-17), which were altered to include language preferences and response styles common to Somali refugees living in Kenya. Confirmatory factor analysis (CFA) was conducted in Mplus to compare eight theoretically-informed models of somatization (1-4 factor models, including simple, correlated, and higher order structures). Invariance tests were conducted to evaluate the quality of the best-fitting model across gender and age groups (13-24; 25-35 years). Construct validity of the best-fitting model was then tested using measures of trauma and traumatic stress.
Findings: Results of the CFA indicated that, across all models tested, the items adapted from established measures were poor indicators of somatization in the current sample. After poor indicators were excluded, we found that a unidimensional model consisting of the seven locally-developed somatization items best fit our data (χ2(14)=24.70, p >.05; CFI=1.0, TLI=1.00; RMSEA<.01, 90% CI [ .001, .05]; SRMR=.04). The items significantly loaded on the factor, with coefficients ranging from 0.63 to 0.71. Configural, metric, and scalar invariance were established, suggesting measurement equivalence across gender and age groups. Evidence of construct validity was found with significant relationships in expected directions with traumatic events and PTSD.
Conclusions and Implications: Results indicate that the 7-item somatization scale can be utilized to assess somatoform symptoms as common responses to trauma sequelae in the Somali refugee community. Although somatization is pervasive among refugees, most Western scales fail to encompass forms of somatic symptoms unique to certain refugee groups. Our study implies that omission of locally-identified somatic items could lead to significant underestimation of somatization and might distort our understanding of commonly-experienced mental health issues in the refugee community. Future studies should aim to replicate and build on these findings to expand and/or refine this locally-developed scale in various Somali communities.