Lessons learned from the translation of the Internalised Stigma of Mental Illness (ISMI) scale into isiXhosa for use with South African Xhosa people with schizophrenia

Abstract

Internalised stigma is highly prevalent among people with mental illness. This is concerning because internalised stigma is often associated with negative consequences affecting individuals’ personal, familial, social, and overall wellbeing, employment opportunities and recovery. Currently, there is no psychometrically validated instrument to measure internalised stigma among Xhosa people in their home language. Our study aimed to translate the Internalised Stigma of Mental Illness (ISMI) scale into isiXhosa. Following WHO guidelines, the ISMI scale was translated using a five-stage translation design which included (i) forward-translation, (ii) back-translation, (iii) committee approach, (iv) quantitative piloting, and (v) qualitative piloting using cognitive interviews. The ISMI isiXhosa version (ISMI-X) underwent psychometric testing to establish utility, within-scale validity, convergent, divergent, and content validity (assessed using frequency of endorsements and cognitive interviewing) with n = 65 Xhosa people with schizophrenia. The resultant ISMI-X scale demonstrated good psychometric utility, internal consistency for the overall scale (α = .90) and most subscales (α > .70, except the Stigma Resistance subscale where α = .57), convergent validity between the ISMI Discrimination Experiences subscale and the Discrimination and Stigma (DISC) scale's Treated Unfairly subscale ( r = .34, p = .03) and divergent validity between the ISMI Stigma Resistance and DISC Treated Unfairly subscales ( r = .13, p = .49). But more importantly the study provides valuable insights into strengths and limitations of the present translation design. Specifically, validation methods such as assessing frequency of endorsements of scale items and using cognitive interviewing to establish conceptual clarity and relevance of items may be useful in small piloting sample sizes.