Errors in Medical Interpretation and Their Potential Clinical Consequences

September 16, 2010 by  
Filed under Interpretation News

Errors in Medical Interpretation and Their Potential Clinical Consequences in Pediatric Encounters

Glenn Flores, MD*, M. Barton Laws, PhD||, Sandra J. Mayo, EdM||, Barry Zuckerman, MD{ddagger}, Milagros Abreu, MD*,{ddagger}, Leonardo Medina, MD{ddagger}, Eric J. Hardt, MD§

* Center for the Advancement of Urban Children, Department of Pediatrics, Medical College of Wisconsin, Milwaukee, Wisconsin
{ddagger} Departments of Pediatrics
§ Internal Medicine, Boston University School of Medicine, Boston, Massachusetts
|| Latino Health Institute, Boston, Massachusetts


Background. About 19 million people in the United States are limited in English proficiency, but little is known about the frequency and potential clinical consequences of errors in medical interpretation.

Objectives. To determine the frequency, categories, and potential clinical consequences of errors in medical interpretation.

Methods. During a 7-month period, we audiotaped and transcribed pediatric encounters in a hospital outpatient clinic in which a Spanish interpreter was used. For each transcript, we categorized each error in medical interpretation and determined whether errors had a potential clinical consequence.

Results. Thirteen encounters yielded 474 pages of transcripts. Professional hospital interpreters were present for 6 encounters; ad hoc interpreters included nurses, social workers, and an 11-year-old sibling. Three hundred ninety-six interpreter errors were noted, with a mean of 31 per encounter. The most common error type was omission (52%), followed by false fluency (16%), substitution (13%), editorialization (10%), and addition (8%). Sixty-three percent of all errors had potential clinical consequences, with a mean of 19 per encounter. Errors committed by ad hoc interpreters were significantly more likely to be errors of potential clinical consequence than those committed by hospital interpreters (77% vs 53%). Errors of clinical consequence included: 1) omitting questions about drug allergies; 2) omitting instructions on the dose, frequency, and duration of antibiotics and rehydration fluids; 3) adding that hydrocortisone cream must be applied to the entire body, instead of only to facial rash; 4) instructing a mother not to answer personal questions; 5) omitting that a child was already swabbed for a stool culture; and 6) instructing a mother to put amoxicillin in both ears for treatment of otitis media.

Conclusions. Errors in medical interpretation are common, averaging 31 per clinical encounter, and omissions are the most frequent type. Most errors have potential clinical consequences, and those committed by ad hoc interpreters are significantly more likely to have potential clinical consequences than those committed by hospital interpreters. Because errors by ad hoc interpreters are more likely to have potential clinical consequences, third-party reimbursement for trained interpreter services should be considered for patients with limited English proficiency.

Key Words: language • interpreters • medical errors • children • pediatrics • Hispanic Americans • quality

Abbreviations: LEP, limited in English proficiency • SD, standard deviation


According to the 2000 census, ~45 million people in the United States speak a language other than English at home, and ~19 million are limited in English proficiency (LEP).1 Five percent of school-aged US children (or ~2.4 million) are LEP, an 85% increase since 1979.2 Language barriers affect multiple aspects of health care for the LEP patient, including access to care, health status, and use of health services.3 Studies document that LEP patients often defer needed medical care,4 have a higher risk of leaving the hospital against medical advice,5 are less likely to have a regular health care provider,6 and are more likely to miss follow-up appointments,7 to be nonadherent with medications,7 and to be in fair/poor health.6

A medical interpreter is an essential component of effective communication between the LEP patient and the health care provider. Medical interpreters may be professional hospital interpreters employed by a health care institution, or ad hoc, untrained individuals, such as family members, friends, nonclinical hospital employees, and strangers from waiting rooms. Previous work has shown that family members8 and untrained bilingual nurses9 who provide ad hoc interpretation can commit many errors of interpretation. Not enough is known, however, about the frequency and categories of medical interpreter errors that occur in clinical encounters, whether such errors potentially have clinical consequences, and if the use of hospital rather than ad hoc interpreters produces a higher quality of medical interpretation. The goals of this study, therefore, were to: 1) determine the frequency, categories, and potential clinical consequences of errors committed by medical interpreters; and 2) compare the quality of interpretation by professional hospital versus ad hoc interpreters.


We audiotaped pediatric encounters in which a Spanish interpreter was used in the pediatric outpatient clinic of an urban Massachusetts hospital over a 7-month period. All study parents had identified themselves as LEP. A bilingual research assistant was present during the encounter only to record the interaction, and did not act as an interpreter, nor take part in subsequent production of transcripts or data analysis. A bilingual verbatim transcript was prepared from the audiotape of each encounter by a professional transcriptionist fluent in both English and Spanish. To ensure accuracy and reliability of the transcripts, each transcript was reviewed 3 times for errors, once by a bilingual physician whose first language is English (G.F.), a second time by a bilingual sociologist whose first language is English (M.B.L.), and a third time by a bilingual physician whose first language is Spanish (M.A.).

The encounters analyzed for this study represent all pediatric visits with Spanish interpreters that occurred in a larger study of patient-physician communication, which consisted of a convenience sample of 153 audiotaped visits in the pediatric outpatient clinics of an urban Massachusetts hospital. Of the 153 participants in this larger study, 110 of the children and their families were Latino. Among these 110 Latino participants, there were 74 mothers/adult caregivers who were LEP, for which 38 visits occurred in Spanish with Spanish-speaking clinicians, 13 visits included a Spanish interpreter, and 25 occurred in English without an interpreter. Although this larger study used a convenience sample, the sample was obtained to reflect a reasonable spectrum of outpatient pediatric visits experienced by Latino families, and has no obvious selection biases other than respondent refusal, which was rare (only 2 potential subjects refused to participate). Participants from the larger study were sampled to capture visits from the full range of daily office hours and all 5 clinic days (Monday-Friday) during the work week. Pediatric encounters included walk-in, sick, and routine health care maintenance visits at the pediatric primary care clinic, and initial and follow-up visits at the outpatient lead and failure-to-thrive clinics. Both pediatricians and pediatric nurse practitioners provided care to study patients, and patient care was in no way altered by the study, except for the presence of the research assistant and tape recorder. The patients and their families, clinicians, and interpreters were told only that this was a study of patient-physician communication, and they were not aware that errors of medical interpretation would be analyzed.

Personnel who provided medical interpretation were classified as: 1) hospital interpreters, professional interpreters (ie, those receiving financial compensation) employed by the study hospital’s department of interpreter services; and 2) ad hoc interpreters, who could include family members, friends, nonclinical hospital employees, strangers from waiting rooms, and hospital clinical staff (including nurses and social workers) who had received no formal medical interpreter training or screening. During the period when the study was conducted, all Spanish hospital interpreters who had been hired had undergone some level of screening and evaluation for language proficiency in Spanish and English. There was, however, no ongoing training or formal performance evaluation in the hospital for interpreters. Low-intensity, voluntary formal interpreter training was sporadically available at various community sites, but it was not known what proportion of interpreters took advantage of these voluntary community opportunities.

For each audiotaped encounter, analysis consisted of identification of the frequency and categories of interpreter errors. An “interpreter error” was defined as any misinterpretation of an utterance that occurred in the clinical encounter, including those committed by the designated medical interpreter, as well as those made by health care providers (such as when a physician with limited Spanish proficiency made errors in Spanish while talking to the mother after the designated interpreter had departed). Errors by health care providers were classified as interpreter errors because the study focus was on errors of interpretation made by any staff member acting as a medical interpreter during a clinical encounter, and we found that certain providers often would attempt to interpret when the designated medical interpreter departed or was temporarily unavailable.

Five categories were used to classify interpreter errors, based on 4 categories used in previous work,10,11 supplemented by an additional category (false fluency). These categories are as follows:

Omission: The interpreter did not interpret a word/phrase uttered by the clinician, parent, or child.

Addition: The interpreter added a word/phrase to the interpretation that was not uttered by the clinician, parent, or child.

Substitution: The interpreter substituted a word/phrase for a different word/phrase uttered by the clinician, parent, or child.

Editorialization: The interpreter provided his or her own personal views as the interpretation of a word/phrase uttered by the clinician, parent, or child.

False Fluency: The interpreter used an incorrect word/phrase, or word/phrase that does not exist in that particular language.

In addition to being classified into 1 of these 5 categories, an interpreter error was also considered to have potential clinical consequences if it altered or potentially altered 1 or more of the following: 1) the history of present illness; 2) the past medical history; 3) diagnostic or therapeutic interventions; 4) parental understanding of the child’s medical condition; or 5) plans for future medical visits (including follow-up visits and specialty referrals).

Medical jargon, idiomatic expressions, and contextual clarifications may occasionally require medical interpreters to not interpret a phrase word-for-word. Thus, any deviations from word-for-word interpretation in transcripts that were attributable to jargon, idioms, or contextual clarifications were not classified as interpreter errors. Because medical interpreters may also act as a cultural broker or advocate, any utterances that could be interpreted as cultural explanations or patient or family advocacy were not classified as interpreter errors. A separate analysis of the relationship of the number of verbal exchanges, the interlocutor, and the quality of the interpretation will be reported elsewhere in a separate paper.

The validity of the analytic method for identification and classification of interpreter errors was assessed as follows: 2 transcripts (cases 26 and 153) were first subjected to preliminary error analysis using simple definitions of each error type and category. The 2 transcripts were scored by 3 observers, a bilingual physician whose first language is English (G.F.) and 2 bilingual physicians (M.A. and L.M.) whose first language is Spanish. To avoid the introduction of bias, the latter 2 observers were blinded to the study goals and hypotheses. Each of the observers was assessed as being highly fluent in their second language based on years of experience providing primary care to Spanish-speaking patients in a Pediatric Latino Clinic (G.F.), 7 years as a research associate on studies of English-speaking populations in the United States (M.A.), and years of teaching high school to English-speaking students in the Massachusetts school system (L.M.). Interobserver variability for the 3 observers was assessed using agreement matrices and by calculating the percentage of agreement in 2 separate analyses, 1 for overall interpreter errors, and the second only for errors of potential clinical consequence. The Kappa Index was also determined for errors of clinical consequence. It was not possible to derive a Kappa Index for overall errors, as transcripts could not be accurately scored for 1 of the 4 cells (cell d): when neither observer identified an error, there was no reliable way to determine whether one should count by words, phrases, transcript lines, or utterances.

The preliminary error analysis of the 2 test transcripts revealed a mean percentage of agreement (± standard deviation [SD]) among the 3 observers on the overall errors of 60% ± 19, with a range of 31% to 82%. Disagreements were primarily attributable to either overlooked errors or unintended differences in the line numbering of the transcripts analyzed by different observers. After line numbering corrections, refinements, and meeting for consensus purposes, there was complete agreement among the 3 observers on the number and type of overall interpreter errors. The mean percentage of agreement (± SD) among the 3 observers on errors of potential clinical consequence in the preliminary analysis was 83% ± 12, with a range of 72% to 97%. The mean {kappa} (± SD) for errors of potential clinical consequence in the preliminary analysis was 0.57 ± 0.3 (considered a moderate strength of agreement by the guidelines of Landis and Koch12), with a range of 0.21 to 0.97 (from fair to almost perfect agreement by the Landis and Koch guidelines12). Because the mean percentage of agreement and {kappa} were considered unacceptably low, the error categories and types were further refined. After refinement, there was mean agreement of 99% ± 1.7 (range: 97%–100%) and a mean {kappa} of 0.99 ± 0.03 (range: 0.94–1.0 [almost perfect by the Landis and Koch guidelines12 for both the mean and range]) regarding interpreter errors of potential clinical consequence on the 2 test transcripts. The remaining 11 transcripts were analyzed by the first author, using the refined error categories, types, and analytic approaches.

To analyze the statistical significance of differences between hospital and ad hoc interpreters in the proportion of errors made, the Yates-corrected {chi}2 test was used, with P < .05 considered statistically significant.

Institutional review board approval was obtained from the participating institution to conduct this study, and written informed consent was obtained from each participating parent.


9 Comments on "Errors in Medical Interpretation and Their Potential Clinical Consequences"

  1. david on Wed, 17th Nov 2010 11:48 pm 

    give me now

  2. newegg promo code on Sun, 21st Nov 2010 6:41 am 

    Nice to see you back. And again by having an interesting post.


    newegg promo code

  3. Medicine Ball Exercises on Sat, 4th Dec 2010 10:11 am 

    Beneficial info and excellent design you got here! I want to thank you for sharing your ideas and putting the time into the stuff you publish! Great work!

  4. Motorcycle Fairing on Sun, 5th Dec 2010 1:36 am 

    Hi there

    Looking forward to your next post

  5. citskaxia on Mon, 6th Dec 2010 5:26 pm 

    Hi, very interesting post, greetings from Greece!

  6. financial aid for college on Wed, 8th Dec 2010 12:31 am 

    Terrific work! This is the type of information that should be shared around the web. Shame on the search engines for not positioning this post higher!

  7. Jamie Iomo on Wed, 8th Dec 2010 5:11 pm 

    Good evening

    This post was interesting, how long did it take you to write?

  8. student scholarship on Thu, 9th Dec 2010 4:24 pm 

    Great site. A lot of useful information here. I’m sending it to some friends!

  9. darArrilefarl on Fri, 7th Jan 2011 7:24 pm 

    Imaginative, they worthy to be taught that filing lawsuits is not the personality to object piracy. Exceedingly than, it’s to review something rally than piracy. Like soundless of use. It’s thoroughly a a barrel easier to train iTunes than to search the Internet with jeopardy of malware and then crappy character, but if people are expected to relief after loads and splotch all together seeing that ages, it’s not affluent to work. They definitely not radix a low-lying on but in days of yore people beget software and Network sites that amount to it ridiculously tame to infringer, and up the quality. If that happens, then there trusted be no stopping piracy. But they’re too watchful and appalled of losing. Risks give creation to to be spellbound!