Medical Team Practices and Interpreter Alterations on Family-Centered Rounds Free

This is a secondary analysis of an observational study of family-centered rounds on hospital day 1 from August 2020 through June 2021 on a general pediatric service.¹⁵  The goal of the primary study was to measure racial and ethnic differences in communication on family-centered rounds. Black and Latino(a/x) families were oversampled, with about half of Latino(a/x) families using an LOE for health care. The encounters analyzed for this study represent all family-centered rounds encounters with a certified in-person professional Spanish interpreter that occurred in a larger study of patient-physician communication. We excluded children undergoing evaluation for neglect or abuse as well as those admitted with a primary psychiatric diagnosis. All enrolled caregivers consented to audio-record family-centered rounds and completed a survey immediately following rounds. Detailed methods of caregiver recruitment and eligibility have been previously published. This study was approved by our institutional review board and all caregivers underwent written consent before participating. We compensated caregivers $20 for their participation in the study.

On the postrounds survey, caregivers reported their age, gender, highest educational level, financial security, child’s insurance, child’s age, and whether their child had a previous hospitalization (excluding birth admission). General pediatric attending physicians participating in this study completed a survey after all audio recordings were complete and reported their gender, race/ethnicity, and years in practice.

All audio recordings were transcribed verbatim by a third-party bilingual professional transcription service. In addition, the professional transcription service back-translated Spanish segments of the encounter into English. Transcriptions were imported into Nvivo for identification of the frequency and categories of interpreter alterations. Initially, we used 4 categories to classify interpreter alterations based on prior literature:^21,22  (1) omission: the interpreter did not interpret a word/phrase uttered by the clinician, parent, or child; (2) addition: the interpreter added a word/phrase to the interpretation that was not uttered by the clinician, parent, or child; (3) substitution: the interpreter substituted a word/phrase for a different word/phrase uttered by the clinician, parent, or child; and (4) editorialization: the interpreter provided his or her own personal views as the interpretation of a word/phrase uttered by the clinician, parent, or child. In preliminary testing, the editorialization code was challenging to identify and rarely occurred independently from other codes. As a result, we omitted this code. Two Spanish-proficient physicians (J.M.R. and V.P.) applied codes of interpreter alterations to the original language. Back-translations were referenced for discrepancies as interpretation may not be a word-for-word process. This phase of coding required time-intensive line-by-line translation and code application. To facilitate timely completion, reliability was assessed iteratively. Acceptable reliability was achieved after the independent coding of 4 transcripts (Cohen’s κ range, 0.78–0.93), after which codes were applied by a single team member.

In a second stage of coding, 2 investigators (V.P. and M.L.) deductively created a second coding schema to characterize alterations as having a positive or negative impact on communication. In the absence of an existing analysis framework, we developed our codebook based on qualitative discussion of the transcripts and review of prior literature.^24,26  We defined 4 positive and 4 negative types of interpreter alterations: (1) gain of information: interpreter alteration improves communication of biomedical information; (2) gain of social support: interpreter alteration increases caregiver psychosocial support; (3) gain of partnership: interpreter alteration increases caregiver empowerment, eliciting caregiver opinion, or shared-decision making; (4) gain of shared understanding: interpreter alteration improves shared understanding between team and patient/family; (5) loss of information: interpreter alteration decreases the communication of biomedical information; (6) loss of social support: interpreter alteration decreases caregiver psychosocial support; (7) loss of partnership: interpreter alteration decreases caregiver empowerment, eliciting caregiver opinion, or shared-decision making; and (8) loss of shared understanding: interpreter alteration decreases shared understanding between team and patient/family (for full definitions and examples, see Supplemental Table 5). Both investigators independently coded all 10 transcripts in this second stage of coding. Investigators met after coding to resolve any discrepancies.

We coded 2 medical team behaviors: how long they spoke before pausing for interpretation and if they used medical jargon. We coded team behaviors by interpreter segment, which was defined as the content from when a speaker began to when there was either a pause for interpretation or a change in speaker. The group of clinicians were viewed as 1 speaker; thus, if there was a lengthy side conversation among the medical team or a clinician added information to what another had said, this was counted as a single segment. The goal of analyzing by interpreter segment was to capture all language that was spoken before interpretation and thus measure if that content was altered in the interpretation process. A word processing program was used to count the number of words and sentences in a segment. We only applied these counts to medical team segments. We used a previously published coding schema of medical jargon to identify different types of jargon in the interpreter segment.^27,28  Jargon was coded as a binary, “present” or “not present,” in our first stage of coding (Cohen’s κ 0.96).

We performed descriptive statistics to report child, caregiver, and attending demographics as well as to identify the frequency of interpreter alterations. To evaluate the association between medical team behaviors and frequency of interpreter alterations we used χ² tests. We looked at 2 specific medical team behaviors: number of sentences and presence of medical jargon. For these analyses, number of sentences was categorized into 1, 2, or 3 or more sentences.

We recorded, transcribed, and coded 10 interpreter-supported family-centered rounds encounters of 10 unique children and their families. In 9 of the 10 encounters, the child had their mother with them during rounds, and in 1 encounter, both the mother and father were present (Table 1). Most caregivers were <45 years old (n = 9) and had a high school level education (n = 6). Four bilingual adolescent children (ages 13–18 years) communicated with the medical team in English during the encounter. There were 7 different attending physicians for the 10 encounters (3 attending physicians had 2 encounters). Of note, none of the attending physicians in this study were Latino(a/x), and none of the clinicians (residents, attending physicians, or nurses) communicated with the family in Spanish. All interpreters in this study were certified medical interpreters per our institution standards. We did not collect further demographics on the interpreters in this study.

In the 10 encounters, there were 529 interpreter segments (Table 2). Of those 529 segments, the primary speaker was a student physician or physician (medical student, resident, or attending) 71% of the time, the caregiver (mother or father) 19% of the time, the patient 9% of the time, and the nurse 1% of the time. Interpreter alterations were common in up to 72% of interpreter segments. Omissions were the most common alteration type (242/529, 46%) followed by substitutions (177/529, 34%) and additions (71/529, 13%). In 17% (90/529) of interpreter segments, there was more than 1 type of alteration that occurred (ie, an omission and a substitution).

A potentially negative alteration occurred in 29% (155/529) of total segments or in 41% (155/382) of segments with an alteration. The most frequent type of negative alteration was loss of information (113/529, 21%), which included omissions of biomedical information, omission of anticipatory guidance (eg, “we may check another one [bilirubin level] after that”), and inaccurate interpretation of information (eg, substituting left for right kidney). Less commonly, negative alterations resulted in a loss of partnership/empowerment (25/529, 5%), social support (23/529, 4%), or shared understanding (7/529, 1%) (Table 3). During our analyses, we noted an increase in negative alterations when conversations between bilingual children (ages 13–18 years) and the team were not interpreted for the caregiver. For example, comparing encounters when the child conversed in English with the team to those where the child did not, the average percent of segments with a loss of information was 27% (64/232) versus 16% (49/297) (χ²P = .002).

In 9% (45/529) of segments, there was a potentially positive alteration. Gain of information was the most common positive alteration type (24/529, 5%) and included simplifying or clarifying medical jargon, avoiding direct translation of idioms, advocating for the team to pause for interpretation, and clarifying ambiguous language (eg, interpreting “pumping to get breastfeeding that way” as “pumping your milk to feed your baby through the bottle”). Shared understanding was the next most observed positive alteration (12/529, 2%) and included actions such as checking for caregiver agreement or understanding during interpretation (eg, “is that correct?”). Gains in social support (5/529, 1%) and partnership/empowerment (7/529, 1%) were infrequently observed.

In our analyses of medical team behavior, we found that the number of sentences used before the team paused for interpretation was associated with all types of interpreter alterations (Table 4). Notably, when teams spoke more sentences before pausing for interpretation, the frequency of omissions increased from 35% if they spoke 1 sentence, to 52% if they spoke 2 sentences, to 81% if they spoke 3 or more sentences (P = .001). Presence of medical jargon was also associated with interpreter alterations, specifically omissions (P = .002) and substitutions (P = .012).

Importantly, medical team behaviors were associated with potentially negative alterations. As noted previously, when segments got long, phrases and concepts were more likely to be omitted in the interpretation, which included losses in both biomedical and psychosocial communication. In some instances, when a long segment started with a supportive statement (eg, “that is a great question”) followed by biomedical information, we observed that the supportive statement was omitted in preference of interpreting the biomedical information.

The relationship between medical jargon and negative alterations was mixed. Although we found a higher frequency of loss of information when jargon was used (15% vs 47%, P < .001), there was a lower frequency of loss of partnership (7% vs 1%, P = .023). This is perhaps explained by where partnership opportunities often occurred during rounds—at the start or end—rather than in the middle of rounds when more technical information was being presented. For example, when teams encouraged caregivers to actively participate at the start of rounds (eg, “interrupt me and correct anything I get wrong”), this invitation was translated into a less empowering statement (eg, “She will pause for the interpreter…and see if you have a question”). These introductions to rounds rarely included jargon but were opportunities for loss of partnership.

Last, medical team behaviors were also associated with positive alterations, which likely reflected efforts by interpreters to improve communication. For example, both number of sentences and use of jargon were associated with gains in shared understanding, which was often the interpreter checking for accuracy and understanding after long and complex segments. Interestingly, we found that the use of jargon was associated with a positive gain of information in addition to the negative loss of information. This likely reflects that when jargon was used, there was an increase in several types of alterations including: a gain of information such as (1) simplifying jargon into lay terms (eg, adding that ondansetron is an antinausea medication) or a loss of information such as (2) incorrectly defining jargon (eg, saying that a child was going to get an EEG instead of an echocardiogram), or (3) omitting medical jargon (eg assuming that jargon-rich segments were intended for team communication end excluded the caregiver).

This study highlights opportunities to improve communication with families during interpreter-supported family-centered rounds given that interpreter alterations were common, had potentially negative consequences on communication, and were influenced by medical team behaviors. Several practical implications can be elicited from this study to inform future interventions including building evidence for frequent pauses for interpretation, offering suggestions to improve caregiver psychosocial support and empowerment, improving partnership with interpreters as cultural brokers and patient advocates, and identifying challenging clinical scenarios such as when 1 caregiver is monolingual and the child is bilingual.

Recommendations exist to guide clinicians on how to effectively work with a medical interpreter.²⁹  However, operational and time constraints contribute to clinicians deviating from best practices when communicating with families who use an LOE.³⁰  In this study, we found that when clinicians did not pause every 1–2 sentences to allow for interpretation as recommended, there was an increase in interpreter alterations, including those we labeled as potentially negative. This likely reflects the challenge of including all communication during interpretation when multiple thoughts are being conveyed. Further, longer team segments were associated with both a loss of biomedical information and communication aimed to support, empower, and partner with caregivers. Hence, not keeping ideas succinct for interpretation may lead to lower caregiver understanding, therapeutic alliance, and rapport.

Additionally, we found that supportive and partnership-building communication was sometimes lost during interpretation when embedded within biomedical content. This practice of empathetic statements encased in medical talk or “buried empathetic statements” is a common practice among clinicians (in both interpreted and noninterpreted conversations) and diminishes the potential for therapeutic alliance.^31,32  In interpreted conversations, our study suggests buried empathetic statements may not just be hidden but completely absent in the interpretation. This is consent with prior studies, which found that biomedical content may be preferentially interpreted over psychosocial information.²⁴  Thus, 1 recommendation to come from this work would be to pause for interpretation after a supportive or empowering statement. We should also note the overall low rates of alterations with potential for loss of social support and partnership statements in this study. We suspect the low rates of these type of alterations may reflect that majority of rounds communication is biomedical with little psychosocial talk. As a result, any loss of psychosocial communication between the team and family may have an outsized impact on the conversation. Nurse participation on interpreted rounds was also infrequent, with the nurse being the primary speaker in only 1% of interpreted segments. Nurse empowerment is another opportunity for improvement given their critical role in safety and quality for children in the hospital.

Cultural and linguistic preferences may have contributed to potentially negative and positive interpreter alterations. For example, interpreters often omitted potentially empowering language such as, “interrupt,” “correct,” or “stop me” at any time, which may reflect that these behaviors may not resonate with sociocultural norms (eg, respeto or deference) in Hispanic/Latino populations.³³  Given that families that use an LOE are less likely to question clinicians in the face of concern for medical error, these alterations can have a negative impact on patient safety and child outcomes.³⁴  These findings emphasize the need to partner with interpreters as cultural conduits. Practices such as welcoming the interpreter to rounds, sharing encounter goals before the encounter, inviting them to pause the encounter if they notice cultural miscommunication, and debriefing afterward are tools to empower interpreters as cultural mediators and patient advocates.³⁵  These strategies can be accomplished in a few sentences at the start and end of an encounter without significantly adding time.

Last, our study highlighted the particular challenge to pediatricians when a child converses in English and a caregiver does not. Several of the encounters with the highest rates of potentially negative alterations were with bilingual children in which large portions of the conversation occurred in English between the medical team and child without interpretation for the caregiver. This aligns with previous studies demonstrating that adolescents participate more in bilingual than monolingual visits.³⁶  Although providers may assume that an adolescent is able to advocate for themselves, literature suggests most do not participate meaningfully in conversations with healthcare providers and may feel anxious or fearful during family centered rounds.^37,38  Additionally, this practice places the burden of interpretation on the child, rather than the professional interpreter, which likely leads to greater errors in interpretation.²¹ 

This study has several important limitations. First, it only examined rounds that occurred a single center and therefore may not be generalizable to other interpreter-supported family-centered rounds. Second, all of these rounds had a certified in-person professional interpreter. Prior research suggests that errors with clinical consequences are more likely to occur with ad hoc interpreters and therefore our study likely underestimates the prevalence of alterations in health care settings.²¹  In addition, different interpreter modalities such as phone or video interpretation were not included in this study and may have different frequency and types of interpreter alterations.^30,39,40  Third, our codes of losses and gains may not capture all the different types of communication being lost or gained by interpreter alterations, including nonverbal communication. Fourth, our study was limited to Spanish interpretation. The specific language and culture of interpreters and caregivers in our study may have affected the types, prevalence, and effects of alterations. Fifth, we only observed 10 encounters that may not be a large enough sample size to reach thematic sufficiency. Sixth, we did not capture level of training or past education working with interpreters for the members of the medical team. Future studies should collect this information to identify if further training is needed at the undergraduate/graduate medical education as well as faculty levels. Last, it is possible that some of the alterations examined in this study were in fact alterations or errors made by the professional transcriptionist or research team rather than the clinical interpreter. Given that the clinical interpreter had to interpret in real time and the research team was able to pause and listen again to content, we suspect greater accuracy in our research team. However, there may have content that was hard to hear, was interpreted differently based on regional variations in dialect, or was communicated nonverbally and not accounted for in recording transcriptions.

To achieve health equity for patients and families that use an LOE, greater attention is needed to how interpreters are incorporated into family-centered rounds. Our study suggests several practical behavioral changes that inpatient teams could trial to improve communication. Future work should evaluate the impact of these modifications on patient and family experience, trust, and health outcomes.