Patterns and Predictors of Professional Interpreter Use in the Pediatric Emergency Department

This study of observed communication behavior with Spanish-speaking LEP families was embedded within a larger randomized controlled trial (RCT) comparing telephone and video interpretation in a pediatric ED. In the larger study, described elsewhere, we enrolled 249 Spanish-speaking LEP patients in the ED.¹⁸  After enrollment in the main study, parents were asked for permission to video record the encounter. For consenting families, a video camera was mounted in the room, and door signs stated that recording was occurring. Recording typically began partway through the visit.

Eligibility criteria for the main study and this substudy were the same. Eligible patients were <18 years of age and presented to the ED during recruitment hours with a family identified at triage as preferring Spanish for medical care between March and August 2014. Patients were ineligible if they had a life-threatening emergency¹⁹ ; if the presenting complaint was primarily behavioral, psychiatric, or social (eg, concern for abuse); or if a treating doctor or nurse practitioner (henceforth, provider) determined that in-person interpretation was needed.

Patients were randomly assigned to telephone or video interpretation by day, so there was a default interpretation modality to use with Spanish-speaking families randomly selected via sealed envelope each recruitment day. We used this approach because (to not delay patient care) study recruitment often happened partway through the visit after some communication had already occurred. Health care professionals were asked to use the default interpretation modality with Spanish-speaking families until eligibility and participation could be assessed, unless in-person interpretation was required.

A bilingual research assistant called parents 2 to 7 days after the visit to complete a survey assessing communication outcomes¹⁸  and family background information. Patient age, insurance type, sex, and medical complexity^20,21  were obtained from hospital administrative data. This study was approved by the Seattle Children’s Hospital Institutional Review Board.

Fifty videotaped encounters were reviewed and coded by at least 1 of 3 investigators (K.C.L., J.G., and J.S.). The coding scheme was developed with the first 5 videos and then refined with the subsequent 5. Questions were flagged and reviewed by other investigators. A random 10% sample of videos underwent double coding. κ statistics were calculated for interrater agreement, with values of 0.89 for method of communication, 0.92 for speaker type, and 0.72 to 1.00 for content categories described below.

Each interaction between a health care professional and the patient or family was deemed a communication event if it included communication or would reasonably have done so if there were no language barrier (eg, checking vital signs). Interactions without medical staff present were excluded. Each communication event was coded for duration, speaker, content, and interpretation type. When the speakers or interpretation type changed, we considered the change to mark the beginning of a new communication event. For example, if a provider gathered the history in English and then switched to telephone interpretation to discuss the plan, we coded 2 separate events.

For each communication event, start and stop times were recorded to calculate duration. For events involving telephone or video interpretation, we recorded interpretation delay time as the time between health care professional initiation of the connection to the interpreter and initiation of an interpreted conversation.

Coded communication content included history taking, plan explanation or discussion, medical updates (including assessing response to treatment), medication administration, procedures, vital signs, other medical interactions, or nonmedical interactions (eg, registration). Many communication events included multiple types of content, in which case all elements were recorded. Some events consisted of a physical examination without any other communication (generally performed by the attending physician); these were categorized separately. When an examination was performed in conjunction with communication, the event was coded on the basis of the communication content.

The primary person speaking with the patient or family was classified as follows: attending ED physician, ED-based resident or nurse practitioner, consulting physician (eg, general surgeon), nurse, other clinical staff (eg, respiratory therapist), or nonmedical staff (eg, registrar). A change in the primary speaker was coded as a new communication event.

The method of communication or interpretation with LEP families was recorded for each communication event as follows: professional in-person interpreter, professional video interpreter, professional telephone interpreter, ad hoc interpreter (family member, friend, sibling, patient, 1 parent for the other, or noninterpreter staff member), and communication primarily in English, Spanish, or a mix of English and Spanish. Speakers recorded using Spanish were compared with the list of individuals certified to use Spanish for medical discussions, as required by hospital policy; events conducted by certified speakers were reclassified as bilingual communications.

Communication events that were exclusively nonmedical were dropped from analysis. Bilingual communications by certified speakers were also dropped because no language barrier existed.

To assess our primary outcome, the communication method was grouped as using professional interpretation (in-person, telephone, or video interpreter) or communicating without professional interpretation (ad hoc interpreter, English, or Spanish by an uncertified speaker). Analyses were conducted in 2 ways: at the level of the patient (n = 50) and at the level of the individual communication event (N = 312). Descriptive statistics were compiled at both levels.

For patient-level analyses, we conducted unadjusted analysis because this study was nested within an RCT, and we expected potential confounders to be evenly distributed between groups. At the patient level, we examined the percentage of communication time that was professionally interpreted, which was calculated as the communication event time with a professional interpreter divided by the total communication time (in seconds).

We tested for an association between assigned interpreter modality (telephone versus video using an intention-to-treat approach) and the percentage of communication using professional interpretation by using Student’s t test, assuming unequal variance. We then repeated this analysis after restricting the communication events to providers (ie, doctors and nurse practitioners) because we hypothesized that providers might be more likely to use interpretation at baseline and so might be more sensitive to the assigned modality.

For communication event–level analyses, we used hierarchical multivariate random-effects logistic models, clustered on the individual patient. Use of this approach allowed the consideration of factors occurring at the level of the patient (eg, assigned modality) and the individual communication event (eg, health care professional type, communication content, and communication duration). We examined the adjusted odds of professional interpreter use for any communication event on the basis of assigned interpreter modality. We then created a similar model, stratified by speaker type (eg, provider versus nurse).

Of 249 participants in the main RCT, 160 were approached for consent for video recording, of whom 57 consented (35.6%), yielding 50 videos that captured at least 1 medical communication (see Fig 1). Patients who were video recorded resembled those in the main study, with the exception of being more often assigned to telephone interpretation, more often boys, and more frequently admitted to the hospital (Table 1). Participating parents were similar to those in the main study, with >80% reporting speaking English not well or not at all and >60% with less than a high school degree. Participant characteristics were evenly balanced between study arms, except that more parents assigned to telephone, compared to video, reported previous experience with the child’s current medical condition (42% telephone arm versus 11% video arm; P = .03).

On average, using video recording, we captured 6.4 discrete medical communications with telephone-assigned families and 6.0 discrete communications with video-assigned families, yielding 312 medical communications across the 50 families (Table 2). Professional interpreter use was captured on video at least once for 86% and 95% of families assigned to telephone and video, respectively.

Among individual communication events, over half of medical communications were conducted in English (51% for telephone arm and 56% for video arm), and about one-third were conducted with the assigned interpreter modality (36% for telephone and 33% for video). Thirty percent of individual communications were part of a longer communication, with a switch in primary speaker (eg, resident to attending) or communication method (eg, starting with an interpreter then disconnecting and finishing in English). Each part of the longer communication had to contain medical information to be counted as a discrete communication event.

Interpretation initiation delay, defined as the time from when a health care professional started the process to connect to a telephone or video interpreter (eg, by dialing the phone number or tapping the video icon) to when they were able to start communicating with the parent, was more than twice as long for telephone compared to video interpretation (mean, 2.1 minutes for telephone versus 1.0 minute for video; P < .001).

Although most LEP families received professional interpretation during their visit, in a more granular analysis, we found professional interpretation was used for half of the total medical communication time (41% for telephone arm and 50% for video arm; P = .16; Table 3). In unadjusted analysis, during communication with providers, 52% of communication time was interpreted in the telephone arm, and 64% was interpreted in the video arm (P = .08).

Overall, professional interpretation was used more often during communication events with providers compared to those with nurses (51% vs 16%; P < .001) and for detailed histories (89%) and explaining the plan (60%) compared to procedures (11%) or medication administration (8%; P < .001; Table 4). In multivariate hierarchical models, we found no overall association between assigned study arm and odds of professional interpreter use (Table 5). However, among provider-led communication events, assignment to video interpretation was associated with a 2.7-fold increased odds of professional interpreter use, after controlling for communication content, duration, and provider role (95% confidence interval [CI]: 1.1–7.0). There was no association between assigned modality and interpreter use among nurses.

In this study of Spanish-speaking LEP families seeking emergency care, professional interpretation was used at least once for most visits. Nevertheless, over half of medical communication minutes occurred without professional interpretation. Interpreter use was particularly low for high-risk activities like medication administration and procedures. Professional interpretation was more often used by providers, compared with nurses, and for longer communications and those involving history taking or education. After controlling for role, communication content, and duration, providers whose patients were assigned to video interpretation had an almost threefold increased odds of using professional interpretation for any communication compared with patients who were assigned to telephone. There was no association between assigned interpreter modality and interpreter use among nurses or in the overall sample, including medical communication by nurses, providers, and other staff.

Most LEP families in our study received some professional interpretation (96%), which compares favorably to studies revealing professional interpretation in less than one-third to one-half of LEP patient encounters.^8,10,22,23  We found lower rates of ad hoc interpreter use than what has been found in many previous studies,^6,9  reflective of a decade of institutional commitment to improving professional interpreter use.^24–26  However, in more granular analysis, we found that over half of medical communication still occurred without professional interpretation despite readily accessible services. These findings are striking for 2 reasons. First, all health care professionals knew they were being video recorded for a study about interpreter use, so typical use may be even lower than what we found. Second, some interpreter use was observed for nearly all participants. In most studies, interpreter use is classified as a dichotomous exposure (ie, any interpreter use in the visit), so these encounters would have been counted as using professional interpretation when most included substantial medical communication without it. This suggests that, at least when considering remote methods of interpretation (in which a new connection to an interpreter must be established for each communication event), even patients classified as receiving interpretation may not have received it for important discussions.

Our findings of variability in interpreter use based on communication content and duration are consistent with results from previous studies of provider decision-making around interpretation.^13–16  Providers are less likely to use interpretation for discussing things they hope will be straightforward¹⁷  or if they believe a patient or parent understands them, regardless of stated preference for interpretation.^13,27  Time constraints and perceived benefits factor into decision-making so that providers may forgo interpretation for anticipated short communications or updates.^13–15  Variable interpreter use over the course of a single ED visit suggests a new decision about use is made for each communication event rather than a decision being made about a family for the entire visit. This finding has important implications for intervention because barriers to timely interpreter access are repeatedly weighed against perceived benefit and hassle.

The association between interpreter use and speaker role is a novel finding. After controlling for communication content, we found that nurses and consulting physicians (ie, not ED-based) had significantly lower odds of interpreter use compared with those of ED-based physicians and nurse practitioners. Potential explanations include less familiarity with interpreter access in the ED or lower importance being placed on information exchange by those individuals. For example, consulting physicians may rely on history documented by ED providers, and nurses may view their primary job as task-oriented (eg, vitals and medication administration) rather than communication-oriented. This latter hypothesis is consistent with the pattern of interpreter use by communication content, in which interpreter use was most frequent when the health care professional needed something from the family (eg, detailed history) and least frequent when the health care professional wished to accomplish a task (eg, administering medications). Unfortunately, these patterns led to infrequent interpreter use for medication administration and procedures, 2 high-risk activities in which a parent’s ability to ask questions and seek clarification may prevent serious medical errors and patient harm. Examples could include a parent’s ability to intervene before a child is given a medication they are allergic to or before a peripheral intravenous line is placed near an arteriovenous fistula used for hemodialysis access.

Assignment to video interpretation was associated with increased professional interpreter use among providers. These observed results are consistent with parent-reported findings from the main RCT’s survey, in which parents assigned to video were less likely to report frequent lapses in interpreter use,¹⁸  and provide objective evidence of greater interpreter use when video was available. Health care professionals and LEP parents have been found to prefer video over telephone interpretation.^28,29  Increasing video interpreter availability, in places where telephone interpretation has been the norm, may increase the use of professional interpretation and improve communication with families.³⁰  In addition, the start-up delay for telephone interpretation was twice as long as for video. Given that health care professionals are deciding whether to use interpretation before every communication event, even small additional barriers can amount to important decreases in use. Efforts to improve interpreter use should prioritize making access as quick and easy as possible.

This study had several limitations. We enrolled families at a single institution with excellent interpreter services, so results may not be generalizable. We had low enrollment from the overall RCT; families who agreed to video recording may differ from those who declined, although the video-recorded group appeared similar to the overall study group on assessed measures. Captured provider behavior was also likely affected by the overall study design: awareness of recording likely improved interpreter use, whereas assignment to a particular method may have dampened it. Nonetheless, because assignment to a particular modality was not enforced, health care professionals retained discretion as to how they would communicate with each family (with only about one-third of providers using the assigned modality during communication events, as detailed in Table 2), producing results that shed new light on how interpreter use decisions are made in practice. Additionally, patterns of remote interpreter use, in which a new connection must be established for each communication, may not reflect patterns for in-person interpreter use, in which that may not be the case. Video recording was generally initiated part way through the visit, often missing the initial history, when interpreter use was highest. Finally, participants were not blinded to study objectives or video recording, so results may reflect a best-case scenario of interpreter use.

In this study of video-recorded pediatric ED visits with Spanish-speaking LEP parents, we found low rates of professional interpreter use over the course of the ED visit. Interpretation was used infrequently by nurses and consulting physicians and for communication that was not part of the initial history or discussion of the treatment plan. We found particularly low interpreter use for high-risk activities, such as medication administration and procedures.

Providers assigned to use video interpretation, rather than telephone, had significantly greater odds of professional interpreter use for medical communication. Additional interventions will be needed to address low use for high-risk activities that are not traditionally thought of as communication but which are likely made safer with communication. Efforts to change the culture around interpreter use, to focus on ensuring that patients and parents can communicate throughout the encounter, rather than only when a provider feels it is needed, will be essential for creating a safer and more equitable health care system.

We thank Juan Kraus (1958–2019) for his invaluable contributions to this study and for his tireless work to improve language access and care for LEP families locally, regionally, and nationally. He was the ever-present, unassuming force that made this study, and many others, possible, and he will be greatly missed.

CI

confidence interval

ED

emergency department

LEP

limited English proficiency

RCT

randomized controlled trial