Bridging language barriers in developing valid health policy research tools: insights from the translation and validation process of the SHEMESH questionnaire

Abstract

BackgroundThe use of research tools developed and validated in one cultural and linguistic context to another often faces challenges. One major challenge is poor performance of the tool in the new context. This potentially impact the legitimacy of health policy research conducted with informal adaptations of existing tools which have not been subjected to formal validation. Best practices exist to guide researchers in adapting and validating research tools effectively. We present here, as an extended example, our validation of the SHEMESH questionnaire ('Organizational Readiness to Change Assessment'; In Hebrew: 'SHE'elon Muchanut Ergunit le'SHinuy'), a Hebrew-language version of the Organizational Readiness to Change Assessment (ORCA). SHEMESH is tailored to support implementation science projects, whose aim is to promote a more rapid and complete adoption of evidence-based health policies and practices.MethodsThe SHEMESH included originally eleven questions from the Evidence (item 1–4) and Context (items 5–11) domains. We validated SHEMESH through the following steps: 1. Professional translation to Hebrew and discussion of the translation by multidisciplinary committee; 2. Back-translation into English by a different translator to detect discrepancies; 3. Eleven cognitive interviews with psychiatric emergency department physicians and nurses; and 4. Pilot testing and psychometric analyses, including Cronbach’s alpha for subscales and factor analyses.ResultsFollowing translation and cognitive interviews, SHEMESH was administered to 222 psychiatrists and nurses. Pearson correlation showed significant and strong correlations of items 1–4 to the Evidence construct and items 6–11 to the Context construct. Item 5 did not correlate with the other items, and therefore was removed from the other psychometric procedures and eventually from the SHEMESH. Factor analysis with the remaining 10 items yielded two factors, which together explained a total of 69.7% of variance. Cronbach's Alpha scores for the two subscales were high (Evidence, 0.887, and Context, 0.852).ConclusionsThis multi-step validation process of the SHEMESH questionnaire may serve as a comprehensive guideline for others who are willing to adapt research tools that were developed in other languages. Practically, SHEMESH has been validated for use in implementation science research projects in Israel.