A new method for evaluating the quality of medical interpretation.

OBJECTIVES: To develop and implement a method to evaluate the quality of medical interpretation in a pediatric outpatient setting and explore the patterns and correlates of errors and failures in translation.

DESIGN, SETTING, AND PATIENTS: Observational study of a convenience sample of 13 Spanish-speaking families attending pediatric outpatient clinics at an urban teaching hospital, their English-speaking providers, and interpreters. Visits were audiotaped and transcribed. The transcripts were divided into segments consisting of continuous sections of dialogue in a single language, plus any translation of that dialogue, and segments were coded for characteristics of the translation, word count, and the identity of the speakers.

MAIN OUTCOME MEASURES: Translation quality on an ordinal scale derived from a set of nominal codes.

RESULTS: We found that 66.1% of segments in which translation should have occurred were translated with substantial errors or omissions or not translated at all. In 29.8% of segments, the interpreter engaged in speech unrelated to interpretation. Quality of interpretation was inversely associated with the word count per segment and, independently, whether the interpreter engaged in speech acts which did not consist of interpretation, which we call "role exchange." We give several examples and qualitative discussion of "role exchange" and show that it not necessarily associated with mistranslation but may have egregious consequences.

CONCLUSIONS: Interpreters who lack appropriate training fail to interpret accurately. Engaging in speech behaviors other than interpreting is associated with a higher rate of errors.