Understanding Patient–Provider Conversations: What Are We Talking About?

Authors


  • Presented at the American College of Emergency Physicians Scientific Assembly, Denver, CO, October 2012.

  • The project was funded by the Davee Foundation—Excellence in Emergency Medicine Grant. The authors have no other disclosures or conflicts of interest to report.

Address for correspondence and reprints: Danielle M. McCarthy, MD, MS; e-mail: d-mccarthy2@northwestern.edu.

Abstract

Objectives

Effective patient–provider communication is a critical aspect of the delivery of high-quality patient care; however, research regarding the conversational dynamics of an overall emergency department (ED) visit remains unexplored. Identifying both patterns and relative frequency of utterances within these interactions will help guide future efforts to improve the communication between patients and providers within the ED setting. The objective of this study was to analyze complete audio recordings of ED visits to characterize these conversations and to determine the proportion of the conversation spent on different functional categories of communication.

Methods

Patients at an urban academic ED with four diagnoses (ankle sprain, back pain, head injury, and laceration) were recruited to have their ED visits audio recorded from the time of room placement until discharge. Patients were excluded if they were age < 18 years, were non–English-speaking, had significant history of psychiatric disease or cognitive impairment, or were medically unstable. Audio editing was performed to remove all silent downtime and non–patient–provider conversations. Audiotapes were analyzed using the Roter Interaction Analysis System (RIAS). RIAS is the most widely used medical interaction analysis system; coders assign each “utterance” (or complete thought) spoken by the patient or provider to one of 41 mutually exclusive and exhaustive categories. Descriptive statistics were calculated for all 41 categories and then grouped according to RIAS standards for “functional groupings.” The percentage of total utterances in each functional grouping is reported.

Results

Twenty-six audio recordings were analyzed. Patient participants had a mean (±SD) age of 38.8 (±16.0) years, and 30.8% were male. Intercoder reliability was good, with mean intercoder correlations of 0.76 and 0.67 for all categories of provider and patient talk, respectively. Providers accounted for the majority of the conversation in the tapes (median = 239 utterances, interquartile range [IQR] = 168 to 308) compared to patients (median = 145 utterances, IQR = 80 to 198). Providers' utterances focused most on patient education and counseling (34%), followed by patient facilitation and activation (e.g., orienting the patient to the next steps in the ED or asking if the patient understood; 30%). Approximately 15% of the provider talk was spent on data gathering, with the majority (86%) focusing on biomedical topics rather than psychosocial topics (14%). Building a relationship with the patient (e.g., social talk, jokes/laughter, showing approval, or empathetic statements) constituted 22% of providers' talk. Patients' conversation was mainly focused in two areas: information giving (47% of patient utterances: 83% biomedical, 17% psychosocial) and building a relationship (45% of patient utterances). Only 5% of patients' utterances were devoted to question asking. Patient-centeredness scores were low.

Conclusions

In this sample, both providers and patients spent a significant portion of their talk time providing information to one another, as might be expected in the fast-paced ED setting. Less expected was the result that a large percentage of both provider and patient utterances focused on relationship building, despite the lack of traditional, longitudinal provider–patient relationships.

Resumen

La Comprensión de las Conversaciones Sanitario-Paciente: ¿De Qué Estamos Hablando?

Objetivos

La comunicación efectiva entre el sanitario y el paciente es un aspecto crítico en la atención de alta calidad al paciente; sin embargo las dinámicas de conversación de una visita al servicio de urgencias (SU) permanece no explorada. La identificación tanto de los patrones como de la frecuencia relativa de comunicación en estas interacciones ayudará a guiar los esfuerzos futuros para mejorar la comunicación entre los pacientes y los sanitarios en el escenario del SU. El objetivo de este estudio fue analizar las grabaciones de audio de las visitas al SU para caracterizar estas conversaciones y determinar la proporción de la conversación ocurrida en las diferentes categorías funcionales de la comunicación.

Metodología

Se reclutaron los pacientes de un SU universitario urbano con cuatro diagnósticos (esguince de rodilla, lumbalgia, traumatismo craneoencefálico y heridas) para tener grabada en audio su visita al SU desde el momento de su ubicación en la sala hasta el alta. Se excluyeron los pacientes si tenían < 18 años de edad, no hablaban inglés, tenían antecedentes significativos de enfermedad psiquiátrica o deterioro cognitivo o estaban médicamente inestables. Se llevó a cabo una edición del audio para eliminar todos los tiempos de descanso en silencio y las conversaciones que no eran entre el paciente y los sanitarios. Se analizaron las grabaciones de audio usando el Roter Interaction Analysis System (RIAS). El RIAS es el sistema de análisis de interacción médica más ampliamente utilizado; los codificadores asignan a cada comunicación (o pensamiento completo) hablada por el paciente o el sanitario a una de las 41 categorías exhaustivas y mutuamente excluyentes. La estadística descriptiva se calculó para todas las 41 categorías y después se agruparon conforme a los estándares del RIAS para “grupos funcionales”. Se recogió el porcentaje de comunicaciones totales en cada grupo funcional.

Resultados

Se analizaron 26 grabaciones. Los pacientes participantes tenían una media de edad de 38,8 años (desviación estándar de 16 años), 30,8% eran hombres. La fiabilidad intercodificador fue buena, con una media de correlaciones intercodificador de 0,76 y 0,67 para todas las categorías del conversación del sanitario y del paciente, respectivamente. Los sanitarios dieron cuenta de la mayoría de las conversaciones en las cintas (mediana: 239 conversaciones, rango intercuartílico [RIC] 168 a 308) en comparación con los pacientes (mediana: 145 conversaciones, RIC 80 a 198). Las conversaciones de los sanitarios se centraron más en la educación y los consejos al paciente (34%), seguido de la activación y la facilitación al paciente (ej.: orientando al pacientes a los próximos pasos en el SU, o preguntando si el paciente entendía; 30%). Aproximadamente el 15% de la conversación del sanitario giró en torno a diferentes datos y la mayoría se centró en tópicos biomédicos (86%) más que psicosociales (14%). El generar una relación con el paciente (ej.: tema social, bromas/risas, mostrando aprobación o estado empático) constituyó el 22% de la conversación de los sanitarios. La conversación del paciente fue principalmente dirigida a dos áreas: dar información (47% de las conversaciones del paciente: 83% biomédica, 17% psicosocial) y a generar una relación (45% de las conversaciones de los pacientes). Sólo el 5% de las conversaciones de los pacientes fueron dedicadas a realizar preguntas. Las puntuaciones centradas en los pacientes fueron bajas.

Conclusiones

En esta muestra, tanto los sanitarios como los pacientes pasan una porción significativa de su tiempo de conversación dando información de uno a otro, como era de esperar en el escenario del SU. Menos esperado fue el resultado que un amplio porcentaje de conversaciones, tanto de sanitarios como de pacientes, se centraran en generar una relación, a pesar de la ausencia de relaciones longitudinales sanitario-paciente tradicionales.

Effective communication between providers and patients is critically important to the provision of safe and high-quality patient care in the emergency department (ED).[1-10] The fast-paced, complex, and stressful environment of the ED poses inherent challenges to patient–provider interactions, and as a result, communication frequently fails to meet patients' needs.[11-15] Failures in the communication process have significant implications for the patients' experience and have been found to have an effect on satisfaction, adherence, resource utilization, and health outcomes.[11, 14, 16-22]

Recent research has predominately focused on indirect evaluations of the communication process, with assessments of patients' perceptions of these interactions, as well as downstream assessments of patient knowledge following ED visits.[11, 23-25] While some work has involved direct audiotaping of patient–provider interactions in the ED, most research has primarily focused on content analysis of the opening and closing segments of the visit, without emphasis on the conversational elements.[26-30] Continuing to improve the communication process between providers and patients in the future demands an even more detailed understanding of the nature of the dialogue.

Additionally, it is important that we consider the role of patient-centered approaches. Patient-centered care has been defined in many ways; however, all definitions rest on four core concepts: dignity and respect, information sharing, participation, and collaboration.[31] Although there is emerging interest in patient-centered care in emergency medicine (EM), very little is known about how patient-centered approaches are integrated most effectively in the ED.[12, 32] Recent consensus-based recommendations published in Academic Emergency Medicine emphasize the importance of research that defines how principles of patient-centered care can be translated to the high-stakes and challenging setting of the ED.[12, 32] Despite this recognition, patient-centered care remains largely a topic of academic discussion, rather than an integrated part of clinical practice or research within EM.

The goal of this study was to describe the routine communication that takes place in an ED setting based on audiotape analysis, assess the areas of emphasis in these conversations, and evaluate if this communication is patient-centered using well-established tools for quantifying dialogue and patient-centeredness.

Methods

Study Design

This was a descriptive study in which audio recordings of complete patient visits were used to describe spoken interactions between EM providers and patients. Institutional review board approval was obtained for all study procedures. Written informed consent was obtained from both providers and patients.

Study Setting and Population

The study took place at an academic medical center (>85,000 annual patient visits) between May 2011 and July 2011. All emergency physicians (EPs), resident physicians, nurses, and technicians were approached to provide informed consent approximately 1 month prior to study initiation. If a provider declined to participate, patients were not recruited in clinical areas where that provider was working. If a provider had been missed in the preconsent process (due to vacations or other obligations), the provider was approached at the time of patient enrollment and written informed consent was obtained prior to entering the patient room.

This study was part of a larger study evaluating patient comprehension of discharge instructions; therefore, only patients with presenting complaints consistent with four diagnoses (ankle sprain, back pain, head injury, and laceration) were targeted. Exclusion criteria included age less than 18 years, non–English language–speaking, significant history of psychiatric disease or cognitive impairment, and medically unstable patients. Patients were recruited based on research assistant (RA) availability and were approached in triage or upon room placement.

Study Protocol

Following patient and provider consent, digital audio recorders were positioned in the patient room with “sound-grabbers” to enhance the quality of the recordings. Audio recording started at the time of patient consent and was stopped after patient discharge. The audio recorder did not leave the room with the patient (e.g., when the patient went to studies such as x-ray or computed tomography); thus only conversations that occurred within the patient room were captured. Following completion of the patient's visit, the total ED length of stay and door-to-doctor time were abstracted from the medical record.

Audiotapes were analyzed using the Roter Interaction Analysis System (RIAS). RIAS is the most widely used medical interaction analysis system, with demonstrated reliability and validity.[33, 34] RIAS is based on social exchange theories and linguistic-based techniques of communication analysis.[33] RIAS coding is performed directly from the audiotape, without transcription, and data are directly entered into RIAS software. The coders assign each utterance (or complete thought) spoken by the clinician or patient to one of 41 mutually exclusive and exhaustive categories. An utterance can range from a single word (e.g., “ok”) to a question or complete sentence. These 41 categories can then be combined to reflect larger functional groupings (see Table 1).

Table 1. Categories of the RIAS
Functional GroupingCommunication BehaviorExamples From Our Study Sample
  1. RIAS = Roter Interaction Analysis System.

Task-focused categories
Data gathering
 Open-ended questions: medical condition and therapeutic regimen

“What other medical problems do you have?”

“When were you supposed to get an x-ray for that?”

 Open-ended questions: lifestyle and social psychological“What do you do for work?”
 Closed-ended questions: medical condition and therapeutic regimen

“Did you have back pain before?”

“Will I be able to walk on it?”

 Closed-ended questions: lifestyle and social psychological

“Do you smoke cigarettes?”

“Is this going to be covered by my insurance?”

Patient education and counseling
 Biomedical information-giving“We are going to give you an air cast which is a little brace that comes around the foot and has some air pillows in it to make you more comfortable.”
 Psychosocial information-giving“I actually rode my bike for the first time from my house to here and it is terrifying. People drive crazy.”
 Biomedical counseling “You are advised to keep ice on it as much as you can and to take ibuprofen 800 mg every 8 hours”
 Psychosocial counseling
Socio-emotional exchange categories
Building a relationship
 Social talk (nonmedical chitchat)“I had a friend who went to DePaul. I remember visiting him on a weekend.”
 Positive talk (agreement, jokes, approval, and laughter)Clinician → Patient “Perfect, that's what we want.”
 Negative talk (disagreements and criticisms)“Oh great, I get to go through all of the pain by myself.”
 Emotional talk (concerns, reassurance, empathy and partnership)Clinician → Patient “I know it's going to be puffy for days. You're doing everything right. For all we know, without you're doing what you're doing, it could be three times as puffy.”
Facilitation and patient activation
 Participatory facilitators (asking for patient opinion, asking for understanding, paraphrases, and back-channel responses)

“Do you think you'll need a prescription for something for pain?”

“I was told it hurts more on the outside, than on the inside, but both sides hurt some. Is that right?”

 Procedural talk (orientations and transitions)“Okay, I'll give you the air cast. I'll put it on now.”

Two researchers trained in RIAS coding (DM, BB) analyzed all audio recordings. One coder (DM) is an EP, and the other (BB) is an emergency nurse. Prior to RIAS coding, all silent downtime or patient–family member conversations were removed from the audio recording using audio editing software. All providers were coded as one team; therefore, there are not separate codes for each resident physician, attending physician, nurse, or technician. Similarly, if a patient's family member interacted directly with the care providers, that interaction was coded with the patient's codes. All downtime conversations between the patient and family members were deleted from the audio recording and were not coded. Twenty percent of the sample was double coded, and the coder's reliability was assessed at the beginning of the coding period.

In accordance with the methods described by Roter and Larson,[35] the frequencies of the physician variables in each of four larger functional categories coded (data gathering, patient education and counseling, building a relationship, and activating and partnering) were summed to determine the overall percentage of conversation spent on that category. This process was repeated for patient variables and the corresponding larger functional categories for patients (information giving, question asking, building a relationship, and facilitation and patient activation).

Data Analysis

Descriptive statistics are reported for the 41 RIAS categories and the larger functional categories. Patient-centeredness summary scores are calculated per RIAS standards. There are two patient-centeredness scores that are offered by RIAS and the scores are computed by creating a ratio of all codes related to psychosocial and socioemotional issues of the exchange (all relationship building, positive, negative, and social talk by providers and patients, partnership building, psychosocial information and counseling, all physician open-ended questions, and all patient questions) divided by the sum of all codes related to biomedical details (all physician and patient biomedical information and counseling, orientations, and physician closed-ended questions) of the visit. In contrast to the first score, which places “physician counseling on biomedical topics” in the denominator (as described above), the second score places this category in the numerator. Individual users of RIAS are permitted to choose which score they prefer depending on their own conceptualization of patient-centeredness (e.g., if they consider patient education and counseling on biomedical topics to further the patient-centered agenda or to detract from it). We have chosen to report both scores to allow the reader to choose. The face validity of these scores depends, in part, on which “philosophy” of patient-centeredness an individual reader ascribes to, and there is no “criterion standard” of patient-centeredness to determine the construct validity; however, the scores have high internal reliability.[36] Scores > 1 are considered patient-centered; in general, published scores tend to range from 0 to 5.

Results

Forty-five patients were approached for enrollment; nine patients refused. The most common reason for refusal was “not interested.” Thirty-six patients were enrolled. Seven patients were later excluded because as their care evolved, they no longer fit inclusion criteria: three cases were admitted to the hospital and four cases had final diagnoses that did not meet inclusion diagnosis criteria (rib fracture, urinary tract infection, pyelonephritis, and laceration with tendon injury). Of the remaining 29 recordings, three audiotapes were low quality and could not be analyzed, resulting in 26 recordings in the final sample. Intercoder reliability was good; the mean intercoder correlation (calculated per RIAS guidelines) for all provider categories was 0.76, and for all categories of patient talk was 0.67. Patient characteristics and final diagnoses are described in Table 2.

Table 2. Participant and Visit Characteristics
CharacteristicN (%)
  1. Aside from age, all data are reported as n (%) or median (IQR).

Mean (±SD) age, yr38.8 (±16.0)
Male sex8.0 (30.8)
Diagnosis
Ankle sprain6.0 (23.1)
Back pain9.0 (34.6)
Head injury2.0 (7.7)
Laceration9.0 (34.6)
Total ED length of stay, minutes125.5 (80–166)
Door to physician, minutes
Length of audio-recorded visit after down time removed in minutes15.5 (12–19)
Patient-centeredness score 10.44 (0.36–0.51)
Patient-centeredness score 21.37 (0.91–2.12)

Overall Patterns of Communication

Analysis of the data revealed that the distribution of approximately half of the RIAS categories was normal; the other half of the data was skewed. The task-focused, higher frequency categories were more likely to have normal distributions, whereas the socioemotional categories were more likely to have skewed distributions. Medians and interquartile ranges (IQRs) are reported. Provider codes are reported in Table 3; patient codes are reported in Table 4.

Table 3. Provider Communication Patterns
CategoryMedianIQR%
  1. IQR = interquartile range.

  2. a

    Not included in percentages.

Data gathering  14.85
Biomedical questions28.515–4112.80
Closed-ended15.39–23.75 
Open-ended11.07.5–20 
Lifestyle-psychosocial questions5.02–72.05
Closed-ended2.00.75–5.25 
Open-ended1.50–3.25 
Patient education and counseling  33.93
Biomedical topics68.541–101.2531.17
Medical condition16.09–34.75 
Therapeutic regimen25.015.25–39.25 
Counseling20.09.25–32.75 
Other 0.00–1 
Lifestyle-psychosocial topics5.01–102.76
Lifestyle2.01–4.25 
Psychosocial0.00–0 
Counseling1.50–5 
Building a relationship  21.60
Social talk7.03–9.254.14
Positive talk22.013–4011.73
Agreements16.05.5–20.5 
Approvals3.51–5.5 
Compliments0.00–1 
Laughter/jokes4.01.75–7 
Negative talk0.00–00.06
Disagreements0.00–0 
Criticisms0.00–0 
Emotional talk11.04.75–205.67
Empathy1.00–3.25 
Legitimizing statements0.00–1 
Concern1.00–3 
Reassurance5.02–13 
Partnership0.00–0 
Self-disclosure0.00–0 
Facilitation and patient activation  29.62
Participatory facilitators30.516.25–4814.90
Back channels7.01.5–15.5 
Paraphrase8.03.75–16 
Ask for opinion1.00–2.25 
Asks if understood9.54–18.25 
Asks for reassurance0.00–0 
Asks for permission0.00–1 
Procedural talk32.526.25–39.2514.72
Orientations29.520.75–33.25 
Transitions3.01.75–6.25 
Uninterpretablea1.50–2.25 
Table 4. Patient Communication Patterns
CategoryMedianIQR%
  1. IQR = interquartile range.

  2. a

    Not included in percentages.

Information giving  47.45
Biomedical topics43.023–78.539.42
Medical condition37.018.75–75 
Therapeutic regimen5.03–8 
Other0.00–1 
Lifestyle-psychosocial topics10.55–14.258.03
Lifestyle10.54.75–14 
Psychosocial0.00–1 
Question asking   5.21
All questions6.03–9.755.21
Biomedical questions 5.02.75–9 
Lifestyle-psychosocial questions1.00–2 
Building a relationship  45.50
Social talk3.52–6.55.18
Positive talk43.525.75–56.7531.21
Agreements28.021.75–36 
Approvals3.52–7 
Compliments0.00–0 
Laughter/jokes4.51.75–13.25 
Negative talk0.00–10.79
Disagreements0.00–1 
Criticisms0.00–0 
Emotional talk8.03–168.32
Legitimizing statements0.00–0 
Concern6.02.75–10.5 
Reassurance1.00–3 
Facilitation and patient activation  1.84
Participatory facilitators1.00–31.38
Asks for service0.00–1 
Paraphrase0.00–1 
Asks if understood0.00–0 
Asks for reassurance0.00–0 
Procedural talk0.50–10.46
Orientations0.00–1 
Transitions0.00–0.25 
Uninterpretablea2.00–4 

Providers accounted for the majority of the conversation in the tapes (median = 239 utterances) compared to patients (median = 145 utterances). Across all tapes, this difference resulted in a talk ratio of 1.67 provider utterances for every one patient utterance. Providers' utterances primarily focused on patient education and counseling (33.9%), followed by patient facilitation and activation (e.g., orienting the patient to the next steps in the ED or asking if the patient understood, 29.6%). Data gathering comprised 14.8% of the provider talk, with the majority of data gathering (86.2%) focusing on biomedical topics rather than psychosocial topics (13.8%). Providers focused 21.6% of their talk on building a relationship (e.g., social talk, jokes/laughter, showing approval, or empathetic statements). Providers often interspersed counseling and information giving throughout their visit; for example, while suturing a laceration, a physician would explain the signs of infection to a patient, or while dispensing a medication, a nurse would discuss the home dosing and side effects.

Patients' conversation was primarily focused in two areas: information giving (47.4% of patient utterances: 83.1% biomedical, 16.9% psychosocial) and building a relationship (45.5% of patient utterances). Only 5.2% of patients' utterances were devoted to question asking.

The median patient-centeredness score using patient-centeredness score 1 (biomedical counseling in the denominator) was 0.44 (IQR = 0.36 to 0.51) and using score 2 (biomedical counseling in the numerator) was 1.37 (IQR = 0.91 to 2.12). Although this study had limited discriminatory ability due to low sample size, post hoc analysis was performed to assess any influence of operational metrics (length of stay, time spent talking to providers) or personal characteristics (age, sex, diagnosis) on the patient-centeredness scores. No differences were found in these post hoc analyses (data not shown).

Discussion

Our data show that there is provider verbal dominance in this set of low-acuity ED encounters and that most of the spoken exchange is focused on providers gathering and patients giving biomedical information. However, this is not the only focus. Emergency providers and their patients are engaging in a multifaceted exchange that is not solely focused on giving and receiving medical information, but also emphasizes relationship building.

Through the use of RIAS methodology, we are able to add to the previous literature that has focused on more specific aspects of ED visits (e.g., discharge) or certain types of conversations (e.g., smoking cessation, depression, domestic violence screening).[27-30] Although such studies are valuable for a thorough understanding of the individual conversations in question, they fall short of providing a global view of the conversational dynamics as experienced by the patient throughout a complete visit.

Through capturing the complete visit we were able to qualitatively compare characteristics of our sample to those of previous samples that have been analyzed by RIAS. Roter and colleagues[37] previously described five distinct types of communication seen in primary care visits. Interestingly, the patterns seen in our sample do not fit with any of the patterns described. The overall provider talk to patient talk ratio of 1.67:1 identified in this study equals the 1.67:1 ratio noted by Roter et al. as reflective of the “consumerist” encounter (e.g., the physician as a consultant who primarily answers questions rather than asks them). However, the content of the conversations is more reflective of Roter's “narrow biomedical” encounters, which are “highly physician controlled” and spend a minimal amount of time on lifestyle and psychosocial topics.[37] Yet, the physicians in our sample spend a larger proportion of time on patient facilitation and orientation than any of the models Roter and colleagues describe. Additionally, patients in our sample seem to spend a larger proportion of their time engaging in positive and social talk. Such differences may be attributable to the lack of a preexisting relationship between EPs and their patients, compared to the primary care sample; perhaps the patients feels the need to develop a personal connection with the EP, however short lived that connection may be. It is not yet clear if the different practice setting and styles of EPs compared to primary care physicians may cause these observed differences.

Other interesting patterns in our data worth comment are expected patterns in the data (a focus on patient orientation and interspersing information giving throughout the visit) and unexpected patterns (a focus on relationship building). We anticipated that a portion of the visit would be spent on orienting the patient to the procedures and processes of the ED, which are often unfamiliar to patients. We also anticipated that provider counseling about topics traditionally considered “discharge topics,” (e.g., patient home care, follow-up, and reasons to return to the ED) would not always take place during the “closing” of the visit. However, less expected was the observed focus on relationship-building. Although the likelihood of the patient and provider having future interactions is slight, providers still spent nearly a quarter of their conversation on topics coded by RIAS as relationship building. It seems from these data that ED providers and patients are striving to establish a relationship for the brief duration of their interaction, despite the lack of a traditional, longitudinal relationship.

Patient-centeredness scores from our sample differ depending on which formulation is used. Most previously published RIAS papers employ patient-centeredness score 1. The values seen in the literature in other clinical settings exceed the scores calculated in our sample. For example, three previous studies of primary care physicians have reported scores above 1.0, including 2.42 (Helitzer et al.),[38] 1.3 (Cooper et al.),[39] and 1.1 (Paasche-Orlow and Roter).[40] The discrepancy between reported scores in other practice settings and the scores in our sample is not particularly surprising, in light of the components of the score. Most practicing EPs would acknowledge that more time is spent on biomedical topics and less time is devoted to furthering the socioemotional and psychosocial elements of the exchange. Although the result is not surprising, it underscores the recent Academic Emergency Medicine consensus conference recommendations of Govindarajan and colleagues[12] that research is needed to define how existing metrics of patient-centeredness translate into the context of emergency care. For example, in the setting of emergency care it may be patient-centered to spend more time educating our patients on biomedical topics; however, if the commonly used metrics (e.g. patient-centeredness score 1) do not recognize such education as being patient-centered, then the patient-centered nature of the exchange is not being captured adequately. Research is needed to not only operationally define the key components of patient-centered care for EM, but also to develop measurement tools or instruments that can assess patients' perceptions of their interactions and, in turn, guide future strategies to improve this aspect of patient care.

Limitations

The sample size is small (26 patients), and participants were recruited as a convenience sample from a single ED. Our patients also had a limited set of four low-acuity diagnoses. The low acuity of the diagnoses may have affected the nature of the conversation, making it more transactional than the interactions surrounding other, more complex diagnoses. We recognize that the findings from 26 visits with four diagnoses are difficult to generalize to our own ED, which sees >85,000 annual visits and hundreds of diagnoses, or to other EDs around the country. Nonetheless, given the lack of previous analysis in the ED setting using this methodology, we believe that our findings are novel and represent a first look into the conversation dynamics in an ED visit.

Other limitations include a possible Hawthorne effect from both the patients and the providers being aware of the recorder in the room. If such awareness existed, it may have led to increased patient education. However, this limitation could not be avoided as, unlike an office practice with a predictable group of return patients, the ED patients in our sample could not be consented beforehand. The providers were consented in advance; however, they may have noticed the audio-recording device in the room or noted the presence of an RA setting up the equipment. Previous studies using audio recording demonstrate that it does not significantly alter physician behavior; thus we that believe this limitation was unlikely to alter significantly the content of the recorded conversations.[41]

Additional endpoints, such as patient satisfaction, knowledge, and medical outcomes are not available for our study sample; therefore, no conclusions can be drawn about the effect of patient-centeredness in this group. Previous reviews of the literature in patient-centeredness have found no consistent relationship between patient-centered encounters and patient outcomes or patient satisfaction.[42] Finally, there are limitations to the RIAS coding framework. The qualitative nature of the work means that the two coders may have brought biases to the application of the coding. Additionally, the RIAS system has been criticized at times because the utterances are defined linguistically rather than by content and because there is some ambiguity between closely related categories, leaving room for coder interpretation of the more socioemotional categories.[43] Although the RIAS system has limitations, these limitations should not be more pronounced in the ED setting than in other clinical settings, and it remains the most widely used medical interaction analysis system.

Conclusions

In this sample, both providers and patients spent significant portions of their time providing information to one another, as might be expected in the fast-paced ED setting. Less expected was the result that a large percentage of both provider and patient utterances focused on relationship building, despite the lack of traditional, longitudinal provider–patient relationships. Compared to studies in other clinical settings, the patient-centeredness scores in this sample were lower than previously reported scores. Before drawing a strong conclusion from this small sample, similar studies are needed in the ED setting to assess spoken interactions in a broader range of conditions and further define patient-centeredness for the unique ED setting.

The authors acknowledge Jennifer Stancati and Francisco Acosta.

Ancillary