Effect of Telephone vs Video Interpretation on Parent Comprehension, Communication, and Utilization in the Pediatric Emergency Department: A Randomized Clinical Trial

Importance  Consistent professional interpretation improves communication with patients who have limited English proficiency. Remote modalities (telephone and video) have the potential for wide dissemination.

Objective  To test the effect of telephone vs video interpretation on communication during pediatric emergency care.

Design, Setting, and Participants  Randomized trial of telephone vs video interpretation at a free-standing, university-affiliated pediatric emergency department (ED). A convenience sample of 290 Spanish-speaking parents of pediatric ED patients with limited English proficiency were approached from February 24 through August 16, 2014, of whom 249 (85.9%) enrolled; of these, 208 (83.5%) completed the follow-up survey (91 parents in the telephone arm and 117 in the video arm). Groups did not differ significantly by consent or survey completion rate, ED factors (eg, ED crowding), child factors (eg, triage level, medical complexity), or parent factors (eg, birth country, income). Investigators were blinded to the interpretation modality during outcome ascertainment. Intention-to-treat data were analyzed August 25 to October 20, 2014.

Interventions  Telephone or video interpretation for the ED visit, randomized by day.

Main Outcomes and Measures  Parents were surveyed 1 to 7 days after the ED visit to assess communication and interpretation quality, frequency of lapses in interpreter use, and ability to name the child’s diagnosis. Two blinded reviewers compared parent-reported and medical record–abstracted diagnoses and classified parent-reported diagnoses as correct, incorrect, or vague.

Results  Among 208 parents who completed the survey, those in the video arm were more likely to name the child’s diagnosis correctly than those in the telephone arm (85 of 114 [74.6%] vs 52 of 87 [59.8%]; P = .03) and less likely to report frequent lapses in interpreter use (2 of 117 [1.7%] vs 7 of 91 [7.7%]; P = .04). No differences were found between the video and telephone arms in parent-reported quality of communication (101 of 116 [87.1%] vs 74 of 89 [83.1%]; P = .43) or interpretation (58 of 116 [50.0%] vs 42 of 89 [47.2%]; P = .69). Video interpretation was more costly (per-patient mean [SD] cost, $61 [$36] vs $31 [$20]; P < .001). Parent-reported adherence to the assigned modality was higher for the video arm (106 of 114 [93.0%] vs 68 of 86 [79.1%]; P = .004).

Conclusions and Relevance  Families with limited English proficiency who received video interpretation were more likely to correctly name the child’s diagnosis and had fewer lapses in interpreter use. Use of video interpretation shows promise for improving communication and patient care in this population.

Trial Registration  clinicaltrials.gov Identifier: NCT01986179

In 2007, more than 25 million people in the United States reported speaking English less than very well.¹ Language barriers in health care are associated with increased costs,² decreased satisfaction³ and treatment adherence,⁴ and increased risk for harm.^5-7 Professional interpretation mitigates these effects⁸ yet remains underused.^9-11 As a consequence, patients and families with limited English proficiency (LEP) may receive care that is less safe and effective relative to English-speaking families.

Technological advances have increased access to telephone- and video-based interpreters,¹² but the level of use remains low; fewer than 1 in 3 LEP patient encounters receives any professional interpretation.^9-11 Reasons for underuse are multifactorial, but cost, technical requirements, time constraints, and clinicians opting to work with nonproficient language skills contribute.^10,13-15 The methods available in a given setting and perceived barriers to access of those methods likely influence consistent use of professional interpreters, increasing the potential for ad hoc interpreter use or reliance on the clinicians’ own nonproficient language skills. Such lapses in the use of professional interpreters increase the risk for miscommunication and patient harm.^5-7,16,17 For example, ad hoc interpreters made errors twice as often as professional interpreters,¹⁶ and 77% of their errors had potential clinical consequences.¹⁷ These consequences can be serious; compared with English speakers, patients with LEP and an adverse event have been found to experience physical harm nearly twice as often⁶ and to remain hospitalized twice as long.⁷

Understanding the relationship among interpretation modality, clinician use of interpretation, and patient and family outcomes is essential for optimizing strategies to improve communication and decrease the risks associated with the failure to use professional interpreters. We aimed to determine the effect of telephone vs video interpretation on (1) parent comprehension (ability to name the child’s diagnosis), (2) parent-reported quality of communication and interpretation, and (3) frequency of lapses in the use of professional interpreters. We hypothesized that, because clinicians prefer video over telephone interpretation,¹⁸ they would use video interpretation more consistently, leading to fewer lapses in interpreter use and therefore better parent comprehension and higher-quality communication and interpretation. Secondary objectives were to compare the length of stay (LOS), duration and cost of interpreted communication, and hospital charges by assigned interpretation modality. We hypothesized that video interpretation would be used more frequently at a higher cost but that LOS and hospital charges would not differ by group, consistent with previous studies comparing professional interpretation services.^19,20

Box Section Ref ID

This pragmatic randomized clinical trial was conducted in the Seattle Children’s Hospital emergency department (ED), a 38-bed pediatric ED staffed by nurse practitioners, residents, fellows, and faculty physicians. In 2014, the ED had 38 954 visits; 20% involved families with LEP. Language need was assessed at triage by asking for the parent-preferred language for medical care. This study was approved by the institutional review board of the Seattle Children’s Hospital. The full study protocol can be found in Supplement 1. All participating parents provided written informed consent.

Before the study, professional interpretation was available in person and via video and telephone modalities. During peak census hours, a Spanish interpreter was present in the ED. All ED rooms were equipped with dual-handset, speaker-enabled telephones with 1-touch dialing to telephone interpretation. Five mobile consoles for video interpretation were also available throughout the ED. Families received 1 or more modalities during a visit based on clinician preference and availability. Hospital policy prohibits clinicians from using nonnative language skills for medical communication unless certified as proficient by the hospital through the Clinician Cultural and Linguistic Assessment, a validated, telephone-based test.²¹

Randomization assigned a remote interpretation modality (telephone or video) to be used throughout the ED each day at the beginning of daily recruitment (Figure 1). Randomization by day occurred via sealed, opaque envelopes in blocks of 6, such that each set of 6 envelopes contained an even number of each modality to ensure an even distribution across days of the week. After opening the envelope, the clinical research assistant (one of whom was B.S.) hung signs announcing the assigned modality on every ED patient room. Faculty and staff were asked to use the assigned modality with Spanish-speaking families with LEP until the clinical research assistant could approach the family, unless in-person interpretation was clinically required. If a family declined to participate or was ineligible, clinicians could use any interpretation modality they chose.

Patients and their parents were recruited from February 24 through August 16, 2014, 5 to 6 days per week (including 1 weekend day each week) from 2 to 10 pm. Recruitment concluded after reaching the goal sample size.

Eligible families included Spanish-speaking parents and their child (younger than 18 years) presenting to the Seattle Children’s Hospital ED for care and requesting Spanish interpretation. Only Spanish-speaking families were enrolled because they constitute two-thirds of LEP families seen, and validated study instruments were only available in Spanish.

Families were ineligible if the child had a life-threatening emergency (triage level 1), the chief complaint indicated a psychiatric disorder or suspected abuse, or the family was assigned a room before the interpretation modality of the day was announced. Families were also ineligible if their ED clinician deemed that care required in-person interpretation. Once enrolled, however, families were retained in the study even if a clinician subsequently used a nonstudy interpretation modality, consistent with our intention-to-treat design. If siblings presented for care together, only the youngest child was eligible to avoid selection bias.

Contact information was collected at enrollment. A telephone survey, administered by a Spanish-speaking clinical research assistant 1 to 7 days after discharge, asked the parent about parent characteristics (level of English proficiency using US Census Bureau categories,¹ highest educational level, family income, and previous experience with the child’s current condition), the quality of communication and interpretation received, how clinicians communicated during the ED visit, and the child’s discharge diagnosis. Measures are described below. If a parent could not be reached within 7 days or 10 telephone calls, their survey responses were considered missing.

The ED visit information was captured at enrollment from the electronic medical record, including triage acuity level, ranging from 1 (life-threatening emergency) to 5 (anticipated need for limited to no workup or intervention) as assigned by the triage nurse²²; day of the week; ED crowding using the routinely calculated National Emergency Department Overcrowding Score (range, 0-200, with higher scores indicating greater degrees of ED crowding relative to available rooms and staff); and the names of the treating clinicians. Those names were compared with a list of certified Spanish-proficient clinicians to determine which patients received some care by clinicians proficient in the parent’s language. Few Spanish-proficient faculty or staff were available, and because all ED patients also received care from multiple nonproficient nurses and other professionals, families receiving some care from a Spanish-proficient clinician were included in the study.

Hospital administrative data provided patient age, insurance type, and sex. Medical complexity was determined with the Pediatric Medical Complexity Algorithm, which uses as many as 3 years of codes from the International Classification of Diseases, Ninth Revision to classify children as having illness that is nonchronic, noncomplex chronic, or complex chronic.²³

Patient utilization data included disposition (admission vs discharge), ED LOS (electronically captured time from registration to admission or discharge), and hospital charges. Telephone and video interpretation invoices documented the duration and cost of remote interpretations. In several cases, parents reported receiving a remote modality but no billing record was found; these data were treated as missing. The child’s ED discharge diagnosis was abstracted from the ED clinician note by an investigator (K.C.L.) who was blinded to study group assignment.

To measure communication quality, we used the Consumer Assessment of Healthcare Providers and Systems Child Visit Survey 2.0 communication composite (5 items), which has been positively associated with treatment adherence and better overall care (eAppendix in Supplement 2).^24,25 These items have response options of no; yes, somewhat; or yes, definitely. The survey items were selected because they are widely used and have undergone extensive cognitive testing for item understandability and validation for use among English speakers and Spanish speakers with LEP.^25-31

Interpretation quality was measured using a 4-item composite and 3 additional questions adapted from questions developed for a previous trial of interpretation methods that included Spanish speakers (eAppendix in Supplement 2).³² Composite item responses were based on a 4-point scale, with scores ranging from poor (1) to excellent (4). The parent was also asked to name or to describe the child’s ED discharge diagnosis using medical or nonmedical terms.³³ Responses were recorded verbatim. Parents were asked to report on the frequency (never, sometimes, frequently, or always) with which clinicians and nurses used the following methods to discuss medical information: professional in-person interpretation, telephone interpretation, video interpretation, use of a family member or a friend to interpret (ad hoc interpretation), speaking in English without an interpreter, and speaking in Spanish without an interpreter.

To analyze the quality of communication and interpretation, top-box scoring was used,³⁴ which is useful for measures that exhibit ceiling effects. In top-box scoring, participants responding with the most positive response (eg, yes, definitely) to all composite items were classified as reporting high-quality communication or interpretation; all others were categorized as reporting low-quality communication. For each composite, if 2 or more items were missing, the composite was scored as missing.

Parent-reported diagnoses were compared with medical record–abstracted discharge diagnoses by 2 investigators (K.C.L. and C.K.G.) who were blinded to group assignment. Using a previously described method,³³ the investigators independently classified each pair of diagnoses as correct, incorrect, or vague/incomplete based on whether a follow-up clinician would reasonably know the diagnosis from the parent-provided information. For example, for a child with bronchiolitis, “a virus in the lungs” was considered correct; “a virus,” vague; and “asthma,” incorrect. Disagreements were reviewed until consensus was achieved. The weighted κ statistic for interrater reliability was 0.51, indicating moderate agreement.³⁵ For analysis, responses were dichotomized as correct vs incorrect, vague, or incomplete.

Parental report of ad hoc interpretation, English without an interpreter, and Spanish without an interpreter were coded as lapses in the use of professional interpreters. For the composite, the response of never to all 3 categories was classified as never; frequently or always in any category was classified as frequently; and the remaining responses were grouped as sometimes. For patients with a Spanish-proficient clinician (31 of 248 [12.5%]), reported Spanish use was not considered in composite coding because we could not know whether the Spanish was used by the certified clinician or another noncertified clinician. In 3 cases, parents reported always receiving more than 1 interpretation modality, and these responses were considered to represent poor question comprehension and therefore excluded from composite calculation.

A subset of study encounters (n = 50) were video recorded. We validated parental ability to report interpretation modality by comparing the parent report with observed use for the 29 video-recorded encounters for which a parent reported always receiving a particular modality (14 for the telephone arm and 15 for the video arm). Parent report of always using a particular modality (eg, video) was deemed correct if no other modality use was observed (eg, telephone or in-person interpretation).

Data were analyzed from August 25 through October 20, 2014. All analyses were conducted as intention to treat based on the assigned interpretation modality. Patient, parent, and ED characteristics were examined using χ² analysis and the unpaired 2-tailed t test by assigned study group to evaluate the effectiveness of randomization. All outcomes were evaluated by assigned interpretation modality in bivariate analysis, using the χ² test for categorical outcomes and the unpaired 2-tailed t test for continuous outcomes.

In a sensitivity analysis, multivariate logistic and linear regressions were used to evaluate relationships between study outcomes and assigned modality, controlling for baseline characteristics that were found to be unbalanced between groups after randomization. Based on the a priori analytic plan, characteristics that differed between groups with P < .10 were included in the sensitivity analyses. A priori power calculations determined that 208 completed surveys would provide more than 80% power to detect small to medium effects in communication²⁸ and interpretation quality,³² as well as medium to large effects in diagnosis comprehension³³ and lapses in professional interpretation.³⁶

During 25 weeks of enrollment, 336 Spanish-speaking families were screened, 290 (86.3%) were eligible, and, of these, 249 (85.9%) consented to participate (Figure 1). The consent rate did not vary by assigned modality, but eligibility did, with significantly more families deemed ineligible on the days assigned to telephone compared with video interpretation (31 of 153 [20.3%] vs 15 of 183 [8.2%]; P = .001). The most common reason for ineligibility was a clinician report of clinical indications requiring in-person interpretation. Loss to follow-up was similar between groups (16 of 107 [15.0%] for the telephone arm vs 25 of 142 [17.6%] for the video arm; P = .58).

Study groups were balanced with regard to ED, parent, and child characteristics (Table 1). The only difference approaching significance was the parental report of not speaking any English (16 of 89 [18.0%] in the telephone arm vs 36 of 115 [31.3%] in the video arm; P = .07). Denominators vary by outcome, both because some relied on survey data while others were derived from administrative data and because of survey questions that respondents skipped.

Parent-reported adherence to the study modality varied significantly by assigned arm, with greater use of the assigned modality on video interpretation days (106 of 114 [93.0%] vs 68 of 86 [79.1%]; P = .004) and greater use of the nonassigned remote modality on telephone interpretation days (30 of 86 [34.9%] vs 18 of 115 [15.7%]; P = .002) (Figure 2). In our internal validation comparing the parent report with the video-recorded encounters, parents who reported always receiving a particular modality were correct in 25 of 29 cases (86.2%).

Parents assigned to the video arm correctly named their child’s diagnosis significantly more often than parents assigned to the telephone arm (85 of 114 [74.6%] vs 52 of 87 [59.8%]; P = .03) (Table 2 and Table 3). More than one-third of parents in each group (43 of 117 [36.8%] in the video arm and 34 of 91 [37.4%] in the telephone arm) reported never experiencing a lapse in professional interpretation, with no significant between-group difference (P = .93). Parents in the telephone arm were significantly more likely to report frequent lapses in professional interpretation although overall levels were low (7 of 91 [7.7%] vs 2 of 117 [1.7%]; P = .04). Groups did not differ with respect to parent-reported quality of communication or interpretation. The ED LOS for admitted and discharged patients, ED charges, and minutes of remote interpretation received did not differ between groups. Mean (SD) charges per patient for video interpretation were twice those for telephone interpretation ($61 [$36] vs $31 [$20]; P < .001). In the multivariate sensitivity analyses when controlling for the only unbalanced baseline characteristic (parent-reported English proficiency), the relationships between the study arm and outcomes remained unchanged (eTable in Supplement 2).

In this pragmatic randomized clinical trial comparing telephone and video interpretation, video interpretation led to better parental ability to name their child’s diagnosis and a lower likelihood of frequent lapses in professional interpreter use. Study groups had comparable parent-reported communication and interpretation quality, ED LOS, charges, and duration of remote interpretation. These findings suggest that video interpretation may improve parental comprehension and decrease opportunities for medical errors, ultimately improving health care outcomes. From the perspectives of family experience and ED flow, however, both remote modalities worked well.

Our finding that video interpretation led to greater diagnosis comprehension suggests that communication may be more effective when the interpreter, family, and clinician can see one another, allowing the interpreter to understand and relay nonverbal aspects of the encounter and better approximating the experience of in-person interpretation. However, comprehension was low in both groups, demonstrating the need for ongoing improvements to communication beyond the selection of the interpretation modality. The diagnosis concordance rates in this study were similar to those we described among LEP parents of hospitalized children³³ but lower than those reported in a previous ED-based study,³⁷ possibly owing to methodologic differences in determining concordance.

The lack of a difference by study modality in parent-reported communication and interpretation quality may result in part from the known tendency among Hispanic respondents to prefer extreme Likert scale responses, which leads to ceiling effects.³⁸ However, the use of top-box scoring should mitigate those effects. Thus, the lack of difference may reflect that telephone and video interpretation work equally well from the families’ perspectives and that professional interpretation itself is the crucial element.

In nearly two-thirds of study encounters (63.2% in the video arm and 62.6% in the telephone arm), parents reported at least some lapses in professional interpretation. While the practice of getting by without professional interpretation is common,^15,39 the high rates we observed are unexpected in a study of interpreter modalities and are cause for concern. Video interpretation led to fewer lapses in interpreter use, which may explain the better parent comprehension and may have implications for patient safety given the link between language barriers and serious safety events.^5-7 The reduced lapses with video interpretation may be owing to clinician preference for this modality.¹⁸

Cost remains an important consideration when addressing language needs in health care settings. We found no difference by assigned modality in ED LOS, charges, or minutes of remote interpretation. Video interpretation was approximately twice as expensive; nevertheless, the relatively small difference between groups ($30 per patient) may be a good investment to improve parent comprehension and lower the risk for communication-related adverse events.

This study was strengthened by its pragmatic design, which assessed the effectiveness of remote interpretation in real-world circumstances. The randomization to the study condition by day approximated the conditions in most clinical settings, where a single remote modality is available and clinicians decide how to communicate with families within those conditions. The study also included a wide range of nurses and clinicians with varying experiences with interpreter modalities and patients and families with diverse conditions and health care experiences. The study design and liberal inclusion criteria enhance the generalizability of the findings. Our use of diagnosis comprehension as an objective outcome of communication reduced the risk for response bias owing to social desirability.

This study has some limitations. As a single-institution study, results may have been influenced by institutional culture or hospital facilities (eg, the telephone and video equipment). Interpreter modalities were provided by different vendors (InDemand Interpreting [video] and Pacific Interpreters [telephone]), with separate professional interpreter networks. However, both vendors employ more than 100 Spanish interpreters and use similar hiring processes and criteria (based on a review of the companies’ websites and written communication from Susan Morgan [Pacific Interpreters], September 23, 2015, and Andrew Drake [InDemand], February 23, 2015), which decreases the chance of systematic differences. Differences in clinician-reported eligibility on telephone compared with video days may reflect clinician dislike for telephone interpretation, leading clinicians on telephone days to declare a patient ineligible so that they could use an alternate method. However, study groups were well balanced; thus, eligibility differences were unlikely to have affected our results. We could not blind clinical staff or families to the assigned study group, but investigators were blinded during the outcome ascertainment. Finally, we gathered some demographic information, such as English proficiency, from parent survey responses. Multiple caregivers are often present during pediatric ED visits, and characteristics of caregivers not surveyed may have been unbalanced between groups. However, the child’s primary caregiver was surveyed most often, and his or her personal experiences and attributes are most likely to affect the child’s care.

Assignment to video interpretation led to improved diagnosis comprehension among parents with LEP of children treated in a pediatric ED. Families assigned to video interpretation were also less likely to report frequent lapses in professional interpreter use. Parent-reported communication and interpretation quality, ED LOS, and charges did not differ by interpretation modality, suggesting that both work well for families and for patient flow. Although video interpretation was more expensive, its use led to better parent comprehension and may decrease the risk for communication-related safety events. Expanding access to video interpretation may improve the quality and safety of care for patients and families with LEP.

Corresponding Author: K. Casey Lion, MD, MPH, Center for Child Health, Behavior, and Development, Seattle Children’s Research Institute, PO Box 5371, Mail Stop CW8-6, Seattle, WA 98145 (casey.lion@seattlechildrens.org).

Accepted for Publication: July 29, 2015.

Published Online: October 26, 2015. doi:10.1001/jamapediatrics.2015.2630.

Author Contributions: Dr Lion had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Lion, Brown, Ebel, Klein, Hencz, Fernandez, Mangione-Smith.

Acquisition, analysis, or interpretation of data: Lion, Brown, Ebel, Klein, Strelitz, Gutman, Mangione-Smith.

Drafting of the manuscript: Lion, Mangione-Smith.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Lion.

Obtained funding: Lion.

Administrative, technical, or material support: Brown, Ebel, Hencz, Fernandez.

Study supervision: Lion, Ebel, Mangione-Smith.

Conflict of Interest Disclosures: Ms Hencz and Mr Fernandez are part of the Department of Interpreter Services within the Center for Diversity and Health Equity but are entirely independent of the grant-making functions of the Center. No other disclosures were reported.

Funding/Support: This study was supported by the Center for Child Health, Behavior, and Development at the Seattle Children’s Research Institute and by the Center for Diversity and Health Equity at Seattle Children’s Hospital; by grant 1K23 HD078507 from the National Institute of Child Health and Human Development (principal investigator, Dr Lion); and by the Department of Interpreter Services and the Emergency Department research team, Seattle Children’s Hospital.

Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Contributions: Jesse Gritton, MPH, Kristin Follmer, BS, and Kelsey Pullar, MPH, Department of Emergency Services, Seattle Children’s Hospital, Shannon Granillo, BA, Center for Clinical and Translational Research, Seattle Children’s Research Institute, and Megan Higgins, BA, and Lurdes Ramos Cayao, BA, Center for Child Health, Behavior, and Development, Seattle Children’s Research Institute, assisted with data collection. Wren Haaland, MPH, Center for Child Health, Behavior, and Development, Seattle Children’s Research Institute, assisted with survey database construction. These contributors were paid for their roles. We thank the Emergency Department faculty, staff, and residents and the parents and children who participated in our study.