A Systematic Review of Interventions to Address Accent-Related Communication Problems in Healthcare

Background: Linguistic diversity in terms of speech, accent, and dialect can present a barrier to communication in healthcare. This review synthesizes current evidence on the effectiveness of interventions that target accent- or dialect-related communication problems in healthcare. Methods: Eligible studies were identified through a search of literature databases in 2018 using medical subject heading (MeSH) terms and additional keywords. Articles were screened for eligibility according to predetermined inclusion criteria. Because of the heterogeneity of the studies, effect data were analyzed using narrative synthesis. Results: Twenty-six studies (n=964) were included, reporting a range of interventions that provided accent modification training, role play or simulated patient consultation training, general language and consultation skills training, web-based training, and cultural competence training. Most studies (24 of the 26) indicated some benefits of the accent elements in the interventions based on objective or subjective measures. The key benefits were improvements in speech production, communication competence, and academic/clinical performance, as well as perceptual changes. Conclusion: Notwithstanding the low research quality and lack of standardized measures in the included studies, this review establishes an evidence base for implementing accent-related communication improvement strategies in healthcare. Large randomized controlled studies would be helpful to strengthen this evidence base.


INTRODUCTION
The quality of communication in healthcare has a significant impact on patient-clinician relationships and patient satisfaction, 1 patient safety, 2 health outcomes, 3 care team collaboration, 4 and professional liability. 5 Communication quality problems related to linguistic diversity in terms of accent and dialect are a growing issue in healthcare given the demographic change in the general population and global migration of clinicians. For instance, the United States is the largest importer of nurses, with approximately 219,000 registered nurses who were trained overseas. 6 More than one-third of the United Kingdom National Health Service medical workforce was trained overseas, 7 and onethird of the doctors working in Australia were trained overseas. 8 In addition, the growing cultural heterogeneity in patient populations resulting from immigration has been noted in the United States and Europe. 9,10 The globalization of the healthcare workforce and patient populations presents a challenge for both clinician-patient communication and clinician-clinician communication. Part of this challenge is to achieve speech clarity and comprehension when speakers with different accents or dialects participate in a verbal conversation. For understanding to occur, the speaker must deliver intelligible speech, and the listener must recognize and understand unfamiliar speech.
According to the American Speech-Language-Hearing Association, an accent is a phonetic trait from a person's original language that is carried over into a second language, while dialect is a broader term that may include pronunciation differences, semantic differences of word choice, and grammatical differences (including sentence word order) that denote regional areas or social groups. 11,12 Although accent and dialect are natural characteristics of speech, they can present a barrier to effective communication in healthcare where a variety of accents and dialects are present. 13,14 Immigrant health professionals' and patients' strong accents can reduce speech intelligibility. Also often observed during patient consultations is the use of colloquial expressions associated with a local dialect that are incomprehensible to those unfamiliar with the expressions. Research shows that accented speech slows listeners' processing time. 15 In the healthcare setting, accents and dialects have been associated with perceptions of lower clinical competency in clinician-clinician communications 16 and clinicianpatient communications, with shorter consultations, 17 with Gu, Y AND (communication OR "communication barriers" OR "clinical communication" OR "communication skills" OR "communication competency" OR "cross-cultural communication" OR "professional communication" OR "culturally and linguistically diverse" OR CALD OR bilingual OR multilingual OR "English as a second language" OR ESL OR "limited English proficiency" OR consultation OR speech OR "clear speech" OR speak OR pronunciation OR intelligibility OR comprehension OR understanding OR perception OR "speech perception" OR stereotype) AND (health OR healthcare OR medicine OR medical OR clinic OR clinical OR "health occupations" OR "foreign professional personnel" OR "foreign medical graduates" OR "international medical graduate" OR IMG OR clinician OR physicians OR doctor OR nurses OR nursing OR pharmacists OR pharmacy OR dentists OR dental OR "allied health professional" OR "occupational therapists" OR OT OR "occupational therapy" OR "physical therapists" OR PT OR "physical therapy" OR "speech-language pathologist" OR SP OR SLP OR "speech-language pathology" OR caregivers OR patients OR "healthcare consumers") AND ("accent modification" OR "accent reduction" OR "accent management" OR "accent adaptation" OR "accent processing" OR "accent identification" OR "accent detection" OR "listening exercise" OR "accent listening" OR "vocal imitation" OR lower care quality, 18 and with less trust for the healthcare provider. 19 A targeted strategy to address the communication problems in healthcare related to accent and/or dialect is accent modification training programs, often conducted among health students such as medical residents 20 and nursing students. 21,22 Another potential strategy, validated in the laboratory setting, focuses on accent adaptation through exposure to systematic variability in accented speech, via either listening exercises 23 or listening and imitating accented speech. 24 Other strategies include using tools and technologies to assist with health communication, such as giving patients a printout of their results or a written explanation to aid patient consultation with accented general practitioners. 14 However, little is known about the impact of these strategies in the healthcare setting. A 2016 systematic review of communication training programs for culturally and linguistically diverse health students concluded that objective evidence of improved skills or performance posttraining is limited. 25 In this review, we expanded the scope of the 2016 review 25 to examine interventions that address accent-and dialect-related communication problems in all healthcare contexts, including health education programs and healthcare delivery systems. The intervention subjects include culturally and linguistically diverse health students and professionals (such as immigrant healthcare professionals), professionals who are minorities or for whom English is their second language (ESL), international medical graduates, and international nursing graduates. Our goal was to synthesize the evidence on the effectiveness of interventions that target communication problems related to accent or dialect in healthcare.

METHODS
To identify scientific evaluations of accent-or dialectrelated interventions in healthcare, we systematically searched and selected relevant studies 26,27 and included both quantitative and qualitative studies in this review, as findings from all types of studies may contribute to understanding the effectiveness of interventions. 28 Before undertaking the review, we developed a review protocol according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guideline. 29 The protocol defined the literature search strategy (Table 1), study eligibility criteria (Table 2), study selection procedure, quality of evidence criteria, and data collection and analysis methods.

Literature Search Strategy
Eligible studies published between 1990 and 2018 were identified through a search of literature databases using medical subject heading (MeSH) terms and additional search terms relevant to the research topic. The databases searched included MEDLINE (via PubMed), Cochrane Central Register of Controlled Trials (via Ovid), Embase (via Ovid), Cumulative Index to Nursing and Allied Health Literature (via EBSCOhost), PsycINFO (via EBSCOhost), Education Resources Information Center (via EBSCOhost), Linguistics Database (via ProQuest), and ScienceDirect. To retrieve further possible articles, we searched the reference lists in the examined studies and in relevant systematic reviews, conducted a hand search of relevant authors and journals, and searched Google Scholar for grey literature such as unpublished studies and dissertations, conference proceedings, and university websites.

Study Selection Procedure
The authors independently reviewed the titles and/or abstracts of all retrieved articles for eligibility according to the study eligibility criteria ( Table 2). We discussed and reconciled discrepancies in screening decisions on 5 abstracts; none of the 5 was included in full-text screening. We then retrieved and reviewed the full text of the remaining studies using the same procedure: independent reviews by the authors followed by discussions until the authors agreed upon the selection decisions.

Criteria for the Quality of Evidence
To assess the validity of included studies, we examined each individual study in terms of the level of evidence, 30 the quality of evidence, 31 the overall research quality (considering the research method, reported measures, and study rigor), and the risk of bias. 32 Taking into account the wide range of research designs and outcome measures used in accent studies, we defined the following criteria based on existing quality appraisal tools and clinical expertise to evaluate the quality of accent intervention studies: r High quality: High-level evidence (eg, randomized controlled trial [RCT], cohort, and case-control studies) with low risk of bias and statistical significance tested of objective measures related to communication quality (including flow-on effects such as patient outcome) r Moderate quality: RCT, cohort, or controlled studies with low risk of bias but with no statistical testing of objective measures (eg, reporting only subjective measures such as qualitative feedback) r Low quality: Uncontrolled studies (eg, case series) or poorquality cohort control studies with significant limitations in scientific rigor, result generalizability, or reproducibility

Data Collection and Analysis Procedures
Data from the eligible studies were extracted and recorded in Microsoft Excel. Multiple reports of one study were collated into a single data point. Extracted data items from each study included the following: r Country where the study was conducted r Study design r Study quality evaluated as described in the Criteria for the Quality of Evidence section r Type and number of intervention participants (health student, healthcare professional, or patient) r Content, structure, duration, and delivery form of the accent or dialect intervention (only the intervention elements relevant to accent or dialect topics were included) r Objective (standardized assessment tools such as established language tests and validated questionnaire surveys and nonstandardized measures based on a theoretical framework or on consistent rubrics such as intelligibility test grades assessed by the trainer) and subjective (qualitative feedback and self-rated competence without a rubric) outcome measures (pretest and posttest, if available) relevant to the short-and long-term effects of the intervention r Study results We planned both a meta-analysis and a narrative synthesis of the data. However, we found substantial heterogeneity among the studies and a range of nonstandardized measures used to assess the intervention effects. As a result, meta-analysis was infeasible, and the narrative synthesis method was used to analyze the data.

RESULTS
We reviewed a total of 729 abstracts; 57 studies met the inclusion criteria and underwent full-text review. Among these, 31 studies were excluded (Figure). Of the 26 studies included in this review, 15 were conducted in the United States, 9 in Australia, and 2 in the United Kingdom. One of the 26 studies was a controlled before-after study, 22 were case series, and 3 were case studies.

Characteristics of Accent Interventions
The studies included in this review described several accent intervention strategies: r Accent reduction training, often led by a speech-language pathologist (SLP) 21,22,[33][34][35][36][37][38] r Role play or simulated patient consultation training 7,39-48 r General language and consultation skill training [49][50][51][52]  r Internet-based education 53,54 r Cultural competence training 55 While each study involved communication skills training, all had distinctive program content, structure, and duration (ranging from 1 session to 2 semesters). Only 3 studies mentioned that the interventions were based on established accent interventions; one study 33 was based on the Compton Pronouncing English as a Second Language Program, 56 another study 37 used the Comprehensive Assessment of Accentedness and Intelligibility model and method, 57 and the third study had expanded the trainer's own accent reduction course regularly taught at a university. 38,58 A range of linguistic elements related to accent or dialect were taught in the included studies: segmental production (consonants and vowels), phonological and prosodic rules, colloquial/idiom/dialect language, vocabulary and medical terminology, grammatical features, nonverbal/body language/pragmatic features to supplement speech understanding, auditory discrimination, and listening skills. Many studies targeted specific populations such as undergraduate and postgraduate culturally and linguistically diverse health students, international medical graduates, and culturally and linguistically diverse/ESL practicing health professionals. In the delivery of accent interventions, the majority of included studies (n=17, 65%) engaged a language expert, such as an SLP or English teacher. Among these 17 programs, 9 (often accent modification training programs) employed a language expert as the sole instructor, 21,33,34,[36][37][38]45,49,50 while 8 used a multidisciplinary teaching team that included both a language teacher and a clinical teacher. 7,22,[41][42][43][44]47,48 Study details are provided in Table 3.

Synthesis of Study Results
Participant Characteristics. A total of 964 participants were involved in 23 studies; the other 3 studies did not specify class size/participant number but were included in this review because they reported relevant intervention descriptions, evaluation methods, and results. More than three-fourths of the study participants were health Addressing Accent-Related Communication Problems        Gu, Y students (n=737, 76%) majoring in nursing, medicine, pharmacy, speech-language pathology, or other health sciences. The second-largest group consisted of 143 (15%) prelicensed health professionals, such as international medical graduates, international nursing graduates, and international medical researchers. In addition, 84 participants (9%) were practicing health professionals, including licensed clinicians, administrators, and clerks.

Intervention Effects
Beneficial Effects. All but 2 studies demonstrated some specific benefits from accent interventions; however, the 2 inconclusive studies reported student trainees' satisfaction with the intervention. 41,52 Identified benefits included improvements in speech production and communication competence, improvements in academic or clinical performance, and positive perceptual changes, as summarized in Table 4. The majority of the studies (n=20, 77%) indicated program satisfaction based on trainee or trainer feedback.
Adverse Effects. Adverse effects or unintended consequences were seldom discussed in the included studies. No objective measures of adverse events were reported, but 3 studies provided qualitative comments on the topic. In a study with culturally and linguistically diverse nurses, the length of the accent reduction training program (10 weeks) was reported to have fatigued some participants; the authors also observed that videotaping encounters with standardized patients added anxiety for some participants. 58 In a language-training program for international medical graduates, disruption of normal work patterns in small departments was mentioned. 49 The third study reporting adverse effects implemented a web-based training program for nursing students, and the author reported "two students felt marginalized and suggested that more emphasis be placed on creating opportunities for them to communicate with local students." 53 Medium-Term Effects. Most studies reported only shortterm effect data collected immediately or almost immediately after completion of the intervention. The longest follow-up period was 2 semesters after the completion of a 2-semester supplementary support program for nursing students. 22 In another study, trainees were interviewed 4 to 8 months after a 13-week training. 33 In another study, a trainee questionnaire survey was administered 3 months after a 2semester training. 46 These 3 studies all reported positive trainee feedback on the intervention effects.

Quality of the Included Studies
As shown in Table 3, all included studies were of low research quality and often had small sample sizes and few objective outcome measures, indicating a lack of generalizability and reproducibility. Only one study-a controlled before-after study-had a control group. 38 Most studies (n=22, 85%) were case series (also known as uncontrolled longitudinal studies), which were before-after studies with no comparison group as a control. The remaining 3 studies 35,37,55 had a case study design.
A range of objective and subjective outcome measures was used in the included studies. Notably, in every one of the 26 studies, a different set of evaluation measures was chosen or developed. Eighteen studies (69%) tested some objective measure (pretest and/or posttest); 8 (31%) reported only qualitative data. Objective posttest outcome measures used to assess intervention effects included study-specific indicators such as knowledge tests based on the content of the training program 50,54 and standardized/ established instruments such as the following: r Australian Medical Council Clinical Examination: 59 Strengths include being a practical clinical test and representative of the performance and knowledge expected in clinical situations.
r Comprehensive Assessment of Accentedness and Intelligibility: 57 Tool for diagnosis and evaluation of a wide variety of speech, auditory, and nonverbal features related to accents/dialects that improves upon the Compton Phonological Assessment of Foreign Accent, 60 the Psycholinguistic Aspects of Foreign Accent, 61  Focuses on self-perceived comfort/apprehension with communication.
r Self-Esteem Scale: 64 Validated instrument used to measure self-esteem.

Role of Language or Communication Experts
Most studies included in this review involved language/ communication experts-an SLP or English teacher-in the delivery of accent interventions. These experts led the training either as a sole instructor 21,33,34,[36][37][38]45,49,50 or as part of a multidisciplinary teaching team along with a clinical teacher. 7,22,[41][42][43][44]47,48 The latter collaborative teaching style was unanimously praised and recommended in all of the studies that used this style. ESL teachers, with understanding of linguistic complexity, are able to teach aspects of language in detail with a specific curriculum, goals, and lesson plans. SLPs, with expertise and experience in speech and language therapy, are also ideally suited to conduct regional or foreign accent improvement programs. 66

DISCUSSION Effects and Need for Accent Interventions
Despite the lack of standardized outcome measures in the studies included in this review, the current evidence suggests that accent interventions in the healthcare context can have direct benefits on communication quality: improvement in speech production, clarity, intelligibility, listening comprehension, knowledge of colloquialism and phonological rules, and reduction of phonological errors. Indirect benefits include improved academic and clinical performance and positive perceptual changes, such as Gu, Y Improved collegiality Yahes and Dunn, 1996 48 Guhde, 2003 35 Boughton et al, 2010 40 Carr and DeKemel-Ichikawa, 2012 21 Harvey et al, 2013 42 Improved attitude towards people with accent/dialect Satter et al, 2005 55 Blackburn, 2012 50 Increased job satisfaction Yahes and Dunn, 1996 48 improved collegiality and job satisfaction. The demonstrated range of beneficial effects provides a strong case for implementing accent-related communication improvement strategies in healthcare to support the diverse healthcare workforce serving increasingly diverse populations. Long-term effects and adverse effects were seldom measured in the included studies. Only 3 studies reported qualitative comments that identified problems with participant fatigue and anxiety, 58 disruption of normal clinical work, 49 and feeling marginalized. 53 We recommend that future studies further explore such issues and comprehensively assess the intervention impact on the participants. Documenting such unintended consequences and long-term effects is important because accent interventions are largely unregulated and free of the jurisdiction of any particular clinical practice area. In other words, no specialized license or accreditation requirements and no guiding ethical code of conduct apply to this area of practice. Accent modification services, for instance, are offered by a variety of practitioners, including licensed professionals (SLPs), ESL teachers, and nonprofessionals. Overidentification and underidentification of participants are possible in accent modification referrals. Exclusionary and discriminatory marginalization of a culturally and linguistically diverse individual, simply on the basis of an accent that differs from the local norm, is a problem of linguistic profiling. 67 We recommend that future studies of accent interventions document the participant selection criteria and plan for rigorous assessment of both beneficial and adverse effects in the short and long term.

Lack of Standardized Outcome Measurements
Every included study used a different set of evaluation measures, ranging from objective measures of speech production, to clinical examinations with standardized patients, to qualitative feedback. The lack of control groups and the lack of comparison from pretest to posttest using consistent objective instruments limit the strength of the current evidence base. The substantial heterogeneity in effect measures reflects a lack of standardized assessment tools in this field and a lack of methodological rigor to support generalizability and reproducibility.
The objective assessment tools that were used in some of the studies have strengths and weaknesses that require a careful examination for adoption in accent interventions. Future research that evaluates accent interventions should consider using standardized measurement tools to pro-vide comparable findings. Moreover, none of the included studies reported findings on patient outcomes or the costeffectiveness of the accent interventions. Future research may explore these important topics to guide practice. We recommend a comprehensive assessment of accent intervention effects to measure not only the effects on communication and intelligibility improvement but also flow-on effects, including patient satisfaction and health outcomes.

Variability of Accent Intervention Elements
The accent interventions included in this review were all training programs for healthcare professionals or students, most of whom had a culturally and linguistically diverse/ESL background. The heterogeneity of the intervention design is evident in the great variety of program elements in terms of training content, structure, and duration. The linguistic topics in the training content varied from program to program, ranging from vowel and consonant production and stress and prosody training to higher-order language elements such as grammar, colloquialisms, idioms, small talk, formulaic expressions, paralinguistic features, medical terminology, and listening comprehension.
A needs assessment, often via trainee survey or focus group session, was conducted in a few studies before the accent intervention was implemented. Only one intervention, a case study of a one-on-one training program, 37 was tailored to address specific speech areas based on structured pretraining diagnostics using the Comprehensive Assessment of Accentedness and Intelligibility. 57 A few other studies recorded pretest speech performance based on tutor-graded consultation skills or established instruments such as the Compton Phonological Assessment of Foreign Accent 60 and Proficiency in Oral English Communication -Screen. 62 Future research may explore the feasibility and impact of using such instruments with medium or large groups of trainees to guide the customization of accent training programs to meet individual trainees' needs.
The inconsistency of the content and procedure among accent interventions highlights a need to develop evidencebased, best-practice guidelines in this field. Laboratory research in speech sciences suggests training in vowel production, [68][69][70] consonant production, 71,72 suprasegmental features, 73-75 auditory discrimination, 76 and listening comprehension. 23,24,77 Rigorous evaluation research in the healthcare setting to assess the impact of accent interventions that is based on these laboratory findings may strengthen the evidence base and contribute to the standardization of practice.
None of the included studies reported on the scalability of their accent interventions, but the 2 fully automated web-based training programs 53,54 appeared promising, with both indicating benefits according to trainee feedback. No other technology-based interventions that particularly address the accent/dialect problem in healthcare were retrieved in the search. However, general communication assistance tools are effective in this setting. For example, a paper-based visual communication tool to assist clinician-patient interaction during counseling was found effective. 78 Giving patients a printout of their results or a written explanation to assist during encounters with accented physicians has been suggested. 14 Such visual assistive tools (paper-based or technology-based) might be useful in addressing communication barriers related to linguistic diversity and might be worth testing further. Future research should also gather evidence on the cost-effectiveness of accent interventions whether technology-based or not.

Implication for Accent Intervention Practice
Accent interventions in healthcare should aim for improvement in speech production in terms of clarity and intelligibility, listening comprehension ability, knowledge of colloquialism and phonological rules, and reduction of phonological errors. Further goals of accent intervention may include better academic or clinical performance and positive perceptual change.
Accent training, ideally delivered by an SLP, may plan for curriculum content on vowel and consonant production, suprasegmental features, phonological variation, auditory discrimination, listening comprehension, vocabulary variability, nonverbal cues, and pragmatic cues (ie, social use of language). Decisions regarding other accent intervention elements may also require careful consideration in terms of the target population, intervention structure, duration, and delivery media, as well as diagnostics and assessment of trainee performance, ideally using objective tools. The rationale for each decision should be documented and reviewed posttraining to assist with the intervention evaluation. The evaluation should also identify the program impact on patient outcomes and health system outcomes, including adverse or unintended consequences in the short and long terms.

Study Limitations
The topic of this review was limited to spoken language diversity in the healthcare setting, especially related to accent and dialect. We did not consider other aspects of communication differences, such as nonverbal linguistic diversity and other dimensions of cultural competence in healthcare. We also excluded translation, interpretation, and transcribing services provided by humans or technology. By including only the interventions that had an element to address accent-or dialect-related communication problems, we aimed to add specific knowledge regarding the efficacy of accent-targeted interventions in healthcare.
The clear heterogeneity of the accent interventions and outcome measures prevented any meaningful metaanalysis of the current data. No randomized controlled trials on accent intervention in the health setting were identified. Instead of reporting risk of bias in each study, we enhanced the review rigor via robust study inclusion criteria and study quality assessment criteria based on the literature and clinical expertise. All included studies had low quality, highlighting the need to strengthen the evidence base in the accent intervention evaluation field. We recommend implementing large randomized controlled studies that not only measure the intervention effects with standardized instruments but also collect data on cost-effectiveness.

CONCLUSION
Despite the low research quality and lack of standardized measures in the included studies, this review establishes a sufficient evidence base to advocate for implementing accent-related communication improvement strategies in healthcare. The key benefits associated with accent interventions include improvements in speech production, communication competence, and academic/clinical performance, as well as perceptual changes. Large randomized controlled studies would be helpful to strengthen the evidence base regarding accent interventions in healthcare.