The literature on measure translation tends to hold, overtly or covertly, a questionable assumption about the possibility of exact translation and almost completely ignores issues of within language variation. Equally, psychometric methods used to assess cross-cultural validity after translation focus on large sample tests of cross-sectional measurement invariance. Such invariance is often not found and is of dubious pertinence to change/outcome measures usually used in psychotherapy research. We present a sequential process of three substudies using quantitative and qualitative procedures to explore whether an outcome measure needs to be changed when used across language variation. Qualitative data confirmed that an item was not ideal in the new context. However, quantitative exploration showed that, although statistically significant and affected by gender and item order, the impact of changing the item in the overall score was small, allowing retention of the existing Spanish translation. We argue that the myth of perfect translation and over-reliance on large-sample psychometric testing pursuing measurement invariance limit exploration of language effects. We recommend that these be used in the companion of user-based, sequential, mixed-method exploration to support the development of a richer field of understanding of outcomes and change self-report measures across languages and cultures and both across and within languages.
Quantifying the impact of partial measurement invariance in diagnostic research: An application to addiction research
