The United States experienced a 191% increase in immigrants from 1970 to 2000, resulting in the growth of individuals with Limited English Proficiency (LEP). The HHS Office of Minority Health defines an LEP person as someone who can “read, write or speak English less than very well”. At present, 32 million LEP individuals face significant linguistic barriers in accessing and receiving quality medical care. A lack of interpreter services can result in adverse outcomes. Numerous studies have demonstrated that LEP patients are more likely to undergo testing in Emergency Rooms, to report less satisfaction with a medical visit and to experience difficulty in understanding prescriptions. In response, health care organizations often provide interpreters. However, there is great diversity in the methods of interpretation and the training of interpreters. The accuracy of provided interpreter services has not been well studied across modes of interpretation or across differing topics of interpretation. If interpretations contain errors, inequities in access to, and quality of, health care could continue for LEP patients. This study aims to fill the gap in understanding the accuracy of interpreter services by creating a model to quantify errors made during interpretation across different modes and topics.
Four different Spanish-English physician-patient initial history-taking dialogues, ranging in length from 1366 words to 1500 words (6 to 9 minutes), were scripted. Four conditions were addressed: tuberculosis, diabetes, depression and menopause. Patients and physicians were recruited as volunteers to act the scripts. Each script was interpreted via four different modes of interpretation. The training and location of the interpreter, as well as the timing of the interpretation define each mode:
- Trained proximate consecutive. A trained interpreter is present during the medical encounter;
- Customary proximate consecutive (Ad hoc). Personnel who are untrained in medical interpreting, such as medical support staff or family members, are utilized;
- Trained remote simultaneous medical interpretation (RSMI). Both provider and patient wear a wireless headset that is connected to a trained offsite interpreter who provides simultaneous interpretation;
- Customary remote consecutive is the use of a telephone interpreter service. In this study, trained interpreters based at Gouverneur Hospital performed this service via telephone.
Each interpretation was tape recorded by a member of the study group, and was transcribed by a bilingual medical transcriptionist. An error analysis model was devised to measure linguistic errors and medical errors in interpretation. Linguistic errors were categorized as additions, omissions or substitutions made by the interpreter to the scripted dialogue. Those errors that altered the meaning of the sentence were categorized as meaningful. Those linguistic errors that were medically related were noted as medical errors. Depending on their severity, medical errors were further categorized as being of mild, moderate, high clinical significance, or as life threatening.
Two bilingual reviewers, blinded to the interpreting mode, analyzed the transcribed, translated scripts for linguistic errors. A study member who is bilingual and trained as a linguist served as a consultant as needed. Three reviewers of varying clinical expertise (one medical student, one fellow and one attending) categorized medical errors. Categorization of errors was made after discussion among the reviewers. Mild discordance was noted among the three medical reviewers. All of the reviewers in the study remain blinded to the mode of interpretation.
At the present time, only the results for the analysis of the tuberculosis script are available. There is a statistically significant difference in the proportion of meaningful linguistic errors among the scripts. Of six possible comparisons among four scripts, three pairings show a statistically significant difference (p<.01) in the proportion of meaningful linguistic errors/total linguistic errors. After discussion, the three reviewers had a concordance rate of categorizing medical errors of 99%. While differences do exist in the number of medical errors across the modes of interpretation, the differences are not statistically significant. However, of note, in two of the four scripts, interpreters misstated the names and dosages of the tuberculosis medications.
- Additional research is needed to elucidate the relationship of interpretation mode to errors in interpretation.
- To more accurately determine error rates with potential clinical impact, error analyses should be conducted on scenarios in which treatments are prescribed.
- Interpreters should be rigorously trained on medical terminology, and should receive continuing medical education to ensure that they remain abreast of new technologies and medications.
- Error analysis model should be employed in other languages and cultures.
Francesca Gany, MD, MS, Center for Immigrant Health, NYU School of Medicine