Racial disparities in clinician responses to patient emotions

Abstract

Objective:

In a previous study of patients newly enrolled in HIV care, we observed that clinicians were less likely to address emotional issues expressed by African American patients compared to whites. We sought to verify and expand these findings in a larger group of patients established in HIV care.

Methods:

We used VR-CoDES to analyze transcripts from 342 audio-recorded medical visits in the United States. We used random intercept multilevel logistic regression to assess associations between patient and clinician characteristics and patterns of emotional talk.

Results:

African American patients were less likely than others to spontaneously express emotions (OR 0.50; 95 % CI 0.29 0.85). Clinicians, who were predominantly white, were more likely to respond to emotional expressions by African American patients explicitly (OR 1.56; 95 % CI 1.11–2.20) but less likely to offer neutral/passive responses that provide space for emotional conversation (OR 0.56; 95 % CI 0.37 0.84) and more likely to block discussion of the emotional issue (OR 2.20; 95 % CI 1.05–4.63). Emotional talk did not vary by patient age or gender.

Conclusion:

These results confirm our prior findings, demonstrating less open emotional communication between African American patients and their providers.

Practice Implications:

Addressing racial differences in communicating about emotions may reduce disparities in patient-clinician relationships.

1. Introduction

Effective communication is an essential component of providing high-quality care and can enhance patient satisfaction, treatment adherence, and clinical outcomes across specialties [1–3]. In HIV care, high-quality communication and patient-provider relationships heavily influence patient experience, retention in care, and continuation of antiretroviral (ARV) therapy [4–7]. In a cross-sectional study of 4694 medical visits with 1743 patients, individuals who felt their HIV providers knew them “as a person” were more likely to be prescribed and be adherent to highly active antiretroviral therapy (HAART), to keep their appointments, and to have undetectable viral loads [4].

The importance of strong relationships between patients and their clinicians suggests that racial disparities in the quality of health care may arise, at least in part, from racial barriers in patient-provider relationships. Prior studies have demonstrated that patient-provider communication differs by race, with typically lower-quality communication between minority patients and their providers [8–12], and that these differences are perceived by minority patients [13]. These findings align with existing literature that clinicians are more verbally dominant, use positive affective tone less frequently, engage in less socioemotional/psychosocial talk, and spend less time chatting with African American patients than white patients [9,14]. In one study that found African American patients spoke less frequently than white patients, the authors stated that “the reasons for this are unclear and may reflect less active engagement among black versus white patients, or alternatively less opportunity given by providers for black patients to engage in the visit [12].”

One crucial aspect of communication that has not been well studied as a potential source of racial disparities in patient-provider relationships is the response to patients’ emotional expressions [15,16]. In a previous study examining communication in initial HIV care visits, we found that clinicians addressed emotional expressions from African American patients less frequently than those from other patients [17]. In addition, in response to expressions of emotion from African American patients, clinicians were less likely to provide space for the patient to speak further (e.g., acknowledgement, empathy, silence, and back-channeling such as ‘yeah’), less likely to make exploratory statements (e.g., questions and invitations to speak), and more likely to block the conversation (e.g., shut down and postpone) than with patients from other racial groups.

Our prior study had a small sample size of 19 clinicians and 43 patients who were meeting each other for the first time, and the racial differences we observed were unexpected. We therefore sought to verify these findings in a larger pool of patients in established patient-provider relationships. We also sought to expand upon our previous study by examining different characteristics of patients’ emotional issues (i.e., topic, explicitness, origin, and repetition) and different types of clinician responses. Finally, we examined potential differences in emotional communication by clinician characteristics (i.e., age, gender, and race).

2. Methods

2.1. Study design, subjects, and setting

We analyzed data from MaRIPOHSA (Maximizing Respect and Improving Patient Outcomes in HIV and Substance Abuse), an observational cross-sectional study conducted at two urban academic medical centers. The study was approved by the Institutional Review Boards at Johns Hopkins University and the Oregon Health & Science University. Patients were eligible to participate if they were English-speaking adults established in HIV care (defined as enrolled in care for more than six months) and visiting a clinician enrolled in the study. Eligible clinicians included physicians, nurse practitioners, and physician assistants providing primary care to patients with HIV. All clinicians maintain continuity of care with their panel of patients at this HIV specialty clinic site, and patients are assigned according to whichever clinician has a new patient slot available. No patient characteristics (such as insurance type) are considered when making provider assignments.

2.2. Data collection

Data for this study came from brief surveys completed by participating providers and patients, and from audio-recordings of patient-provider visits. Research assistants recruited eligible patients in the clinic’s waiting room. All participants gave informed consent before data collection. Two recorders were set in the examination room to audio-record the visit, and a professional transcription company transcribed the recordings. We recorded one visit per enrolled patient.

2.3. Coding of audio-recorded transcripts

As in our prior study, we used the Verona Coding Definitions of Emotional Sequences (VR-CoDES) to code emotional communication. VR-CoDES has been found to be a reliable and valid tool in identifying and explaining patient expression of negative emotion and clinician response across a variety of specialties [18–22]. To calibrate the coding process, two authors (JP and MCB) coded 65 clinical encounter transcripts and resolved discrepancies by consensus, until coding was consistent between the authors. Further coding was done by a single coder (JP).

2.3.1. Patient emotional expressions

VR-CoDES defines and categorizes all patient expressions of negative emotions as concerns or cues [18]. A concern is “a clear and unambiguous expression of an unpleasant current or recent emotion where the emotion is explicitly verbalized.” A cue is “a verbal or non-verbal hint which suggests an underlying unpleasant emotion but lacks clarity.” VR-CoDES allows further divisions of cues into subcategories (Table 1). For example, cue type F represents “non-verbal clear expressions of negative or unpleasant emotions or hints to hidden emotions.” Since our methodology included transcripts from audio-recordings, we could not detect most non-verbal communication, such as facial expression and vocal tone, although it is worth noting that some indicators of emotion (e.g., silence, gasps, sighs, laughter, and crying) can be heard and are noted in the transcripts. Therefore, appropriate non-verbal indicators were coded as Cue F by our research team, despite not having video-recordings.

Table 1.

Distribution of Patient Emotional Expressions (N = 1028).

	Description		Examples	n (%)
Concern	A clear and unambiguous expression of an unpleasant current or recent emotion where the emotion is explicitly verbalized, either with a stated issue of importance for the patient or without a stated issue.		Right. That’s what I was worried about.	278 (27%)
Cue	A verbal or non-verbal hint which suggests an underlying unpleasant emotion but lacks clarity.			750 (73%)
	Cue A	Words or phrases in which the patient uses vague or unspecified words to describe his/her emotions.	D: How are you doing? P: Uh, not so good, I guess.	23 (2.2%)
	Cue B	Verbal hints to hidden concerns (emphasizing, unusual words, unusual description of symptoms, profanities, exclamations, metaphors, ambiguous words, double negatives, expressions of uncertainties and hope).	It’s a lot. It is a lot, man.	301 (29.3%)
	Cue C	Words or phrases which emphasize (verbally or non-verbally) physiological or cognitive correlates (regarding sleep, appetite, physical energy, excitement or motor slowing down, sexual desire, concentration) of unpleasant emotional states.	It really affects me. I can’t– I can’t sleep.	22 (2.1%)
	Cue D	Neutral expressions that mention issues of potential emotional importance which stand out from the narrative background and refer to stressful life events and conditions. This applies to non-verbal emphasis of the sentence, abrupt introduction of new content, pauses before or after the expression or to a patient-initiated repetition of a previous neutral expression in subsequent turns.	She’s been diagnosed now with– you know, dementia.	232 (22.6%)
	Cue E	A repetition, with very similar words, of an expression said in a previous turn by the patient.	P: I ain’t even that, that old, you know, but I feel old. P: That’s what it is. I feel real old.	129 (12.6%)
	Cue F	Non-verbal clear expressions of negative or unpleasant emotions (crying), or hints to hidden emotions (sighing, silence after clinician question, trembling voice, frowning, etc.).	P: I don’t want you to see me cry. D: No, it’s okay– P: [inaudible]	2 (0.2%)
	Cue G	A clear and unambiguous expression of an unpleasant emotion in the past (e.g., a previous mental state, a previous worry or fear).	And I think about that, and I was– I was so angry and bitter towards my cousins because of the decision they made not to come around there.	41 (4%)

VR-CoDES measures the explicitness of patient expressions of negative emotions (explicit concerns vs. implicit cues). We also assessed the origin of emotional expressions (i.e., did the clinician ask a question that elicited the negative emotion or did the patient spontaneously express it). Furthermore, we were able to distinguish the content of emotional issues (medical vs. non-medical). Medically-related topics included HIV, other illnesses, tests, treatments, symptoms, and substance use, while non-medical emotional expressions referred to general life events, living situations, and people. Finally, we evaluated the repetition of patients’ emotional expressions, coded as either an initial expression or a subsequent reference to a previously observed cue or concern.

2.3.2. Clinician response to patient emotional expression

VR-CoDES describes 17 possible clinician responses to patient emotional expressions (Table 2) [19]. These specific response types are sorted into two primary features of the response: explicitness and provision of emotional space. Explicitness refers to whether or not the clinician statement refers to the content of the emotional issue, the affect of the expression, or both. Clinician responses that provide space allow patients to talk further, and responses that reduce space discourage or deflect emotional talk.

Table 2.

Distribution of Clinician Responses to Patient Emotional Expressions (N = 1028).

Overarching Response Types	Specific Response Types	Definitions	n (%)
Non-explicit - Reduces Space 73 (7%)	Ignore (code NRIg)	Ignoring the patient.	55 (5%)
	Information/Advice (code NRIa)	Giving information or advice to the patient without direct reference to the emotional expression (e.g., “Everything will be fine.”).	15 (1%)
	Shutting down (code NRSd)	Shutting the patient down (e.g., “Oh, don’t be silly!”).	3 (<1%)
Non-explicit - Provides Space 504 (49 %)	Acknowledgement (code NPAc)	Acknowledging using moderate verbal encouragement (e.g., “Are you really?”).	88 (9%)
	Back Channeling (code NPBc)	Back-channeling through minimal verbal encouragements such as “okay.”	340 (33 %)
	Active invitation (code NPAi)	Actively inviting the patient to talk further (e.g., “Would you like to tell me more?”).	26 (3%)
	Implicit Empathy (code NPIm)	Providing empathy that implies that the clinician recognized the emotion but does not specifically repeat it back (e.g., “I understand.”).	43 (4%)
	Silence (code NPSi)	Providing silence.	7 (1%)
Explicit - Reduces Space 127 (12 %)	Information-advice (code ERIa)	Giving information or advice to the patient with direct reference to the patient’s emotional expression (e.g., “I don’t think it’s an infection.”).	113 (11 %)
	Switching (code ERSw)	Explicitly directing the discussion away from emotional content by changing the frame of reference (e.g., “Talk to him tomorrow. He’s the best expert on it, I think.”).	8 (1%)
	Postponing (code ERPp)	Postponing discussion until later (e.g., “Hold on one second.”).	6 (1%)
	Active Blocking (code ERAb)	Actively blocking the patient from elaborating and refusing to talk further about it (e.g., “Worrying does not do you any good.”).	0 (0%)
Explicit - Provides Space 324 (32 %)	Content Acknowledgement (code EPCAc)	Acknowledging the circumstance giving rise to the emotion (e.g., “the operation?”).	103 (10 %)
	Content Exploration (code EPCEx)	Asking more about the circumstance (e.g., “What operation are you going to have?”).	183 (18 %)
	Affective Acknowledgement (code EPAAc)	Acknowledging the emotion itself (e.g., “worried?”).	14 (1%)
	Affective Exploration (code EPAEx)	Asking more about the emotional experience (e.g., “Why are you so worried?”).	16 (2%)
	Empathy (code EPAEm)	Expression of empathy that repeats back to the patient the emotion that is heard (e.g., “I’m sorry. I can understand why that would be really worrisome.”).	8 (1%)
Secondary Conceptual Categories			n (%)
Reduces Space	Gives information/advice (codes NRIa, ERIa)	Gave information or advice.	128 (12 %)
	Any blocking (codes NRIg, NRSd, ERSw, ERAb)	Actively tried to avoid the emotional expression.	66 (6%)
Provides Space	Neutral/passive(codes NPSi, NPBc, NPAc)	Was passive by giving silence, back-channeling, or providing non-explicit acknowledgement.	435 (42 %)
	Acknowledgement (codes NPAc, EPAAc, EPCAc)	Acknowledged the emotion or circumstance.	205 (20 %)
	Exploring (codes NPAi, EPAEx, EPCEx)	Explored the emotional issue by asking the patient for more information that referred to the emotion or circumstance.	225 (22 %)
	Explicit focus on emotion (codes EPAAc, EPAEx, EPAEm)	Focused explicitly on the emotion by acknowledging it, asking about it, or providing explicit empathy.	38 (4%)
	Any empathy(codes NPIm, EPAEm)	Expressed any empathy.	51 (5%)

^*Secondary conceptual categories are not mutually exclusive and do not add to 100 %.

2.3.3. Secondary outcomes/grouping of clinician response

We grouped qualitatively similar clinician responses, as defined by the VR-CoDES, a priori into conceptual categories that were not mutually exclusive. These categories were entirely conceptual and not empirically defined. Statements that provided emotional space to patients were grouped depending on whether or not the clinician:

• was passive by giving silence, back-channeling, or providing non-explicit acknowledgement (codes NPSi, NPBc, or NPAc).

• provided acknowledgement, either generally or explicitly, that referred to the emotion or circumstance (codes NPAc, EPAAc, or EPCAc).

• explored the emotional issue by asking the patient for more information, either generally or explicitly, that referred to the emotion or circumstance (codes NPAi, EPAEx, or EPCEx).

• focused explicitly on the emotion by acknowledging it, asking about it, or providing explicit empathy (codes EPAAc, EPAEx, or EPAEm).

• expressed any empathy, either implicitly or explicitly (codes NPIm or EPAEm). This coding system defines ‘empathy’ as a verbal response that indicates that the clinician understands the patient’s emotion.

Clinician responses that reduced space were grouped depending on whether or not the clinician:

• gave information or advice (codes NRIa or ERIa) that closed off exploration or discussion of the emotion.

• actively tried to avoid the emotional expression (codes NRIg, NRSd, ERSw, or ERAb).

2.4. Covariates

Patients and providers self-reported their age, gender, and race/ethnicity on questionnaires.

2.5. Statistical analysis

We generated descriptive statistics on clinician, patient, and visit characteristics, and the frequency of patient emotional expression and clinician response types. We used random-intercept multi-level logistic regression to assess associations between the independent variables (age, gender, and race) and outcome variables (patient emotional expression and clinician response). To account for clustering of patients within clinicians, we used a multi-level model in which patient emotional expressions were nested within each visit, and visits were nested within clinicians. Since there was only one visit per patient, we did not need to account for clustering of encounters within patients.

Most of the patients in this analysis were African American and white. As the focus of this study was on verifying previous findings specifically indicating lower-quality emotional communication for African Americans, we identified differences in emotional communication between African American patients and those from all other racial groups collectively (including whites). Because we were also interested in comparing the majority racial group (whites) to minorities, we also examined differences between white patients and those from all other racial groups collectively (including African Americans). Although most of the white vs. minority comparisons are effectively “reciprocal” findings of the African American vs. non-African American findings, we felt it was important not to restrict our analyses to African American vs. white so as to include non-African American minority patients’ experiences, even though their numbers were too small to analyze separately.

3. Results

3.1. Sample characteristics

Our sample included 41 clinicians and 342 English-speaking adult patients. The average age of patients and clinicians at the time of the medical visits was 53.2 (SD 10.2) and 45.7 (SD 10.3) years, respectively (Table 3). Patients were predominantly male (63.7 %) and African American (76.6 %). Clinicians were predominantly female (65.9 %) and white (65.9 %). Clinicians provided HIV care to an average of 8 study patients (range 2 10).

Table 3.

Clinician and Patient Characteristics.

Characteristics	Clinician (N = 41)	Patient (N = 342)
Age, mean (SD)	45.7 (10.3)	53.2 (10.2)
Gender, n (%)
Male	14 (34.1 %)	218 (63.7 %)
Female	27 (65.9 %)	124 (36.3 %)
Race, n (%)
White	27 (65.9 %)	69 (20.2 %)
African American	6 (14.6 %)	262 (76.6 %)
Hispanic/Latino	2 (4.9 %)	–
Asian	4 (9.8 %)	2 (0.6 %)
Other	2 (4.9 %)	9 (2.6 %)
Number of study patients per clinician, mean (range)	8.3 (2–10)

3.2. Visit characteristics

Table 4 displays the characteristics of the 342 recorded medical visits. The average visit length was 30.4 min (range 9.5–75.0; SD 11.9). There were 228 (66.7 %) visits that contained at least one emotional expression, as categorized by VR-CoDES. There was an average of 3.0 (range 0–24; SD 3.7) emotional expressions per visit. Among those visits that contained emotional expressions, there was an average of 4.5 (range 1–24; SD 3.6) cues or concerns per visit. Visit characteristics did not significantly vary by patient age, gender, or race.

Table 4.

Visit Characteristics by Patient Demographics.

	All Patients (n = 342)	Patient Demographic Characteristics
		Age			Gender			Race/Ethnicity
		<53yo (n = 142)	≥53yo (n = 163)	p-value	Male (n = 218)	Female (n = 124)	p-value	African American (n = 262)	White (n = 69)	Other (n = 11)	p-value
Visit length in minute, mean (SD)	30.4 (11.9)	31.0 (12.6)	30.7 (11.8)	0.79a	30.7 (11.9)	29.8 (11.7)	0.52a	31.0 (11.7)	28.5 (12.6)	28.0 (11.5)	0.24c
Visits with at least 1 emotional expression, n (%)	228 (66.7 %)	96 (67.6%)	109 (66.9 %)	0.90b	144 (66.1 %)	84 (67.6%)	0.81b	173 (66.0 %)	48 (69.6%)	7 (63.6%)	0.84b
Mean (SD) emotional expressions among all visits	3.0 (3.7)	3.1 (4.0)	2.9 (3.3)	0.62a	2.8 (3.4)	3.4 (4.0)	0.11a	2.8 (3.5)	3.7 (4.4)	2.7 (3.0)	0.26c
Mean (SD) emotional expressions among visits with at least 1 emotional expression	4.5 (3.6)	4.5 (4.2)	4.3 (3.2)	0.61a	4.2 (3.4)	5.1 (4.0)	0.08b	4.3 (3.4)	5.1 (4.4)	4.3 (2.6)	0.28c

Note:

^at-test;

^bFisher’s exact test;

^cOne-way ANOVA.

3.3. Patient emotional expressions

Of the 1028 emotional expressions identified, 750 (73 %) were cues and 278 (27 %) were concerns (Table 1). The most common linguistic method patients used to express their emotional issues was through emphasizing statements containing unusual terms and phrases (Cue B, 29.3 %). Patients often provided neutral expressions that referred to potentially emotional situations that did not flow with the context of the previous dialogue (Cue D, 22.6 %).

When evaluating expressions of negative emotions by patient demographics, Cue B (i.e., emphasizing) was the most common across age and racial/ethnic groups, and among women (Table 5). Among men, explicit expression of emotional issues (Concerns, 29.9 %) was more common than any of the implicit expression types.

Table 5.

Distribution of Patient Emotional Expression and Clinician Responses by Patient Demographic Characteristics (N = 1028).

	Patient Demographic Characteristic
	Age			Gender			Race/Ethnicity
	<53yo (n = 435)	≥53yo (n = 465)	Male (n = 603)	Female (n = 425)	African American (n = 746)	White (n = 252)	Other (n = 30)
Patient emotional expression type Medical Non-medical	208 (47.8 %)227 (52.2 %)	253 (54.4 %)212 (45.6 %)	322 (53.4 %)281 (46.6 %)	219 (51.5 %)206 (48.5 %)	384 (51.5 %)362 (48.5 %)	144 (57.1 %)108 (42.9 %)	13 (43.3 %)17 (56.7 %)
Concern	105 (24.1 %)	126 (27.1 %)	180 (29.9 %)	98 (23.1 %)	193 (25.9 %)	78 (31.0 %)	7 (23.3 %)
Cue	330 (75.9 %)	339 (72.9 %)	423 (70.2 %)	327 (76.9 %)	553 (74.1 %)	174 (69.1 %)	23 (76.7 %)
Cue A	9 (2.1 %)	11 (2.4 %)	12 (2.0 %)	11 (2.6 %)	20 (2.7 %)	2 (0.8 %)	1 (3.3 %)
Cue B	118 (27.1 %)	143 (30.8 %)	177 (29.4 %)	124 (29.2 %)	205 (27.5 %)	83 (32.9 %)	13 (43.3 %)
Cue C	9 (2.1 %)	12 (2.6 %)	13 (2.2 %)	9 (2.1 %)	18 (2.4 %)	4 (1.6 %)	0 (0.0 %)
Cue D	116 (26.7 %)	90 (19.4 %)	128 (21.2 %)	104 (24.5 %)	169 (22.7 %)	57 (22.6 %)	6 (20.0 %)
Cue E	63 (14.5 %)	63 (13.6 %)	70 (11.6 %)	59 (13.9 %)	115 (15.4 %)	12 (4.8 %)	2 (6.7 %)
Cue F	2 (0.5 %)	0 (0.0 %)	0 (0.0 %)	2 (0.5 %)	2 (0.3 %)	0 (0.0 %)	0 (0.0 %)
Cue G	13 (3.0 %)	20 (4.3 %)	23 (3.8 %)	18 (4.2 %)	24 (3.2 %)	16 (6.4 %)	1 (3.3 %)
SubsequentInitial	232 (53.3 %)203 (46.7 %)	207 (44.5 %)258 (55.5 %)	291 (48.3 %)312 (51.7 %)	213 (50.1 %)212 (49.9 %)	359 (48.1 %)387 (51.9 %)	133 (52.8 %)119 (47.2 %)	12 (40.0 %)18 (60.0 %)
Doctor-ElicitedPatient-initiated	101 (23.3 %)332 (76.7 %)	107 (23.3 %)353 (76.7 %)	142 (23.7 %)457 (76.3 %)	90 (21.4 %)330 (78.6 %)	188 (25.4 %)551 (74.6 %)	36 (14.4 %)214 (85.6 %)	8 (26.7 %)22 (73.3 %)
Primary clinician response categories
ExplicitNon-explicit	199 (45.8 %)236 (54.3 %)	207 (44.5 %)258 (55.5 %)	259 (43.0 %)344 (57.1 %)	192 (45.2 %)233 (54.8 %)	348 (46.7 %)398 (53.4 %)	92 (36.5 %)160 (63.5 %)	11 (36.7 %)19 (63.3 %)
Provide spaceReduce space	360 (82.8 %)75 (17.2 %)	358 (77.0 %)107 (23.0 %)	488 (80.9 %)115 (19.1 %)	340 (80.0 %)85 (20.0 %)	586 (78.6 %)160 (21.4 %)	215 (85.3 %)37 (14.7 %)	27 (90.0 %)3 (10.0 %)
Secondary clinician response categories
Neutral/passive (NPSi, NPBc, NPAc)	180 (41.4 %)	185 (39.8 %)	260 (43.1 %)	175 (41.2 %)	287 (38.5 %)	133 (52.8 %)	15 (50.0 %)
Acknowledgement (NPAc, EPAAc, EPCAc)	94 (21.6 %)	92 (19.8 %)	132 (21.9 %)	73 (17.2 %)	150 (20.1 %)	46 (18.3 %)	9 (30.0 %)
Exploring (NPAi, EPAEx, EPCEx)	107 (24.6 %)	98 (21.1 %)	121 (20.1 %)	104 (24.5 %)	174 (23.3 %)	48 (19.1 %)	3 (10.0 %)
Explicit response to emotion (EPAAc, EPAEx, EPAEm)	19 (4.4 %)	16 (3.4 %)	19 (3.2 %)	19 (4.5 %)	30 (4.0 %)	8 (3.2 %)	0 (0.0 %)
Any empathy (NPIm, EPAEm)	18 (4.1 %)	26 (5.6 %)	32 (5.3 %)	19 (4.5 %)	36 (4.8 %)	13 (5.2 %)	2 (6.7 %)
Gives information/advice (NRIa, ERIa)	44 (10.1 %)	69 (14.8 %)	73 (12.1 %)	55 (12.9 %)	99 (13.3 %)	28 (11.1 %)	1 (3.3 %)
Any blocking (NRIg, NRSd, ERSw, ERAb)	29 (6.7 %)	35 (7.5 %)	37 (6.1 %)	29 (6.8 %)	56 (7.5 %)	8 (3.2 %)	2 (6.7 %)

Emotional expressions were more frequently initiated by patients (77 %) than elicited by clinicians (23 %). Patients’ negative emotions were slightly more often medically-related (53 %) than non-medically related (47 %). Patients raised 524 distinct emotional issues. Of these, 289 (55 %) were mentioned only once, while 235 (45 %) came up multiple times. Of the total 1028 emotional expressions, 504 (49 %) were repeated expressions of previously mentioned issues.

3.4. Clinician responses to emotional expressions

Most clinician responses provided space for patients to elaborate on their emotions (Table 2; 81 %). Of these 828 responses that provided space, 504 (61 %) were non-explicit (i.e., did not directly refer to the content nor emotion), and 324 (39 %) were explicit. Among the 200 responses that reduced emotional space for patients, 73 (36.5 %) were non-explicit, while 127 (63.5 %) were explicit. Back-channeling, a non-explicit response that provides space for emotional expression, was the most frequent clinician response type (33 %).

When using our conceptual groupings of provider responses to patients’ emotions, common response types were passive (42 %), exploring (22 %), acknowledging (20 %), and information-giving (12 %). Verbal empathy (5%) and an explicit focus on patient emotions (4%) were uncommon.

3.5. Patient characteristics and emotional communication

In our multivariate analysis, we found no associations between patient age and gender and the characteristics of their emotional expressions, nor how their clinicians responded to them (Table 6). Patient race, however, was associated with both patient emotional expressions and clinician responses. In terms of patient emotional expressions, African American patients had more than three times greater odds than non-African American patients of conveying emotion by repeating an otherwise neutral expression (Cue E, OR 3.40; 95 % CI 1.57–7.36). Further, African American patients were more likely than others to have emotional issues elicited by the provider rather than spontaneously expressing them (OR 1.99; 95 % CI 1.17, 3.39). In contrast, white patients were less likely (OR 0.43; 95 % CI 0.24 0.75) than patients from other racial groups to have clinicians elicit emotional talk (i.e., were more likely to bring attention to their emotional issues spontaneously).

Table 6.

Patient Demographic Characteristics Associated with Emotional Expressions and Clinician Responses.

	Patient Demographics
	Age OR (95% CI)a	Female vs. Male OR (95 % CI)a	African American vs. all other race/ethnicities OR (95 % CI)a	White vs. all other race/ethnicities OR (95 % CI)a
Patient emotional expression characteristics
Medical vs. Non-medical	1.01 (0.97, 1.05)	1.07 (0.49, 2.34)	0.91 (0.36, 2.28)	1.38 (0.54, 3.51)
Concern vs. Cue	1.00 (0.97, 1.03)	0.84 (0.47, 1.51)	0.99 (0.50, 2.00)	1.17 (0.58, 2.39)
Cue A	0.99 (0.94, 1.04)	1.32 (0.51, 3.44)	2.78 (0.70, 11.03)	0.28 (0.06, 1.33)
Cue B	1.01 (0.99, 1.03)	1.03 (0.66, 1.58)	0.75 (0.46, 1.23)	1.17 (0.70, 1.96)
Cue C	1.02 (0.95, 1.10)	0.90 (0.19, 4.25)	1.65 (0.28, 9.64)	0.78 (0.13, 4.74)
Cue D	0.99 (0.96, 1.02)	1.08 (0.59, 1.97)	0.83 (0.41, 1.68)	1.16 (0.56, 2.41)
Cue E	1.00 (0.97, 1.03)	1.05 (0.59, 1.88)	3.40 (1.57, 7.36)	0.28 (0.12, 0.65)
Cue F	0.98 (0.87, 1.11)	–	–	–
Cue G	1.01 (0.95, 1.07)	1.51 (0.48, 4.69)	0.40 (0.12, 1.30)	2.79 (0.83, 9.37)
Subsequent vs. Initial	0.99 (0.97, 1.00)	1.08 (0.80, 1.44)	0.88 (0.63, 1.23)	1.21 (0.85, 1.72)
Doctor vs. Patient-initiated	1.01 (0.99, 1.03)	0.74 (0.47, 1.19)	1.99 (1.17, 3.39)	0.43 (0.24, 0.75)
Primary clinician response categories
Explicit vs. non-explicit	1.00 (0.98, 1.01)	1.05 (0.77, 1.44)	1.56 (1.11, 2.20)	0.65 (0.45, 0.93)
Provide vs. reduce space	0.99 (0.97, 1.01)	0.97 (0.63, 1.50)	0.64 (0.39, 1.07)	1.41 (0.82, 2.41)
Secondary clinician response categories
Neutral/passive(NPSi, NPBc, NPAc)	0.99 (0.98, 1.01)	0.97 (0.67, 1.39)	0.56 (0.37, 0.84)	1.76 (1.14, 2.71)
Acknowledgement(NPAc, EPAAc, EPCAc)	1.00 (0.98, 1.02)	0.75 (0.51, 1.10)	1.01 (0.65, 1.59)	0.84 (0.53, 1.35)
Exploring(NPAi, EPAEx, EPCEx)	0.99 (0.98, 1.01)	1.27 (0.91, 1.78)	1.41 (0.95, 2.08)	0.78 (0.52, 1.17)
Explicit response to emotion(EPAAc, EPAEx, EPAEm)	0.99 (0.96, 1.02)	1.45 (0.74, 2.83)	1.47 (0.64, 3.35)	0.80 (0.35, 1.83)
Any empathy(NPIm, EPAEm)	1.02 (0.99, 1.06)	1.09 (0.56, 2.14)	0.89 (0.42, 1.92)	1.05 (0.47, 2.32)
Gives information/advic(NRIa, ERIa)	1.01 (0.98, 1.03)	1.07 (0.62, 1.83)	1.29 (0.70, 2.39)	0.92 (0.49, 1.73)
Any blocking(NRIg, NRSd, ERSw, ERAb)	1.01 (0.98, 1.04)	1.09 (0.60, 1.98)	2.20 (1.05, 4.63)	0.40 (0.18, 0.90)

^aRandom intercept multilevel logistic regression models (accounts for clustering of emotional expressions within encounters and of patients within clinicians).

In responding to patient emotions, clinicians were more likely with African American patients (OR 1.56; 95 % CI 1.11–2.20), compared to patients from other racial groups (including whites), to explicitly refer to patients’ emotional expressions. Clinicians were also less likely with African Americans, compared to others, to offer neutral or passive responses (e.g., back-channeling) that facilitate further emotional expression (OR 0.56; 95 % CI 0.37 0.84), and more likely to block further conversation about the emotional issue (OR 2.20; 95 % CI 1.05–4.63).

Conversely, clinicians were less likely with white patients, compared to others (including African Americans) to provide an explicit response to patient emotions (OR 0.65; 95 % CI 0.45 0.93). White patients were also more likely than others to be provided neutral/passive responses (OR 1.76; 95 % CI 1.14–2.71) and were less likely to be blocked (OR 0.40; 95 % CI 0.18 0.90).

3.6. Clinician characteristics and emotional communication

Clinician race was not associated with patient emotional expressions or clinician responses (Table 7). Patients were more likely to express emotion by repeating a neutral statement (Cue E) with older (OR 1.04; 95 % CI 1.01–1.07) and female clinicians (OR 2.13; 95 % CI 1.01–4.51). Older clinicians were more likely to respond to patients’ emotions by giving information and advice (OR 1.03; 95 % CI 1.00–1.05).

Table 7.

Clinician Demographic Characteristics Associated with Emotional Expressions and Clinician Responses.

	Clinician Demographics
	Age OR (95 % CI)a	Female vs. Male OR (95 % CI)a	White vs. all other race/ethnicities OR (95 % CI)a
Patient emotional expression characteristics
Medical vs. Non-medical	0.99 (0.95, 1.03)	1.07 (0.45, 2.52)	0.99 (0.42, 2.31)
Concern vs. Cue	0.97 (0.94, 1.00)	0.57 (0.28, 1.18)	1.18 (0.55, 2.51)
Cue A	0.97 (0.92, 1.03)	0.95 (0.32, 2.84)	1.05 (0.36, 3.11)
Cue B	1.00 (0.98, 1.02)	1.34 (0.81, 2.21)	1.24 (0.76, 2.02)
Cue C	1.03 (0.95, 1.12)	2.29 (0.40, 13.18)	0.77 (0.15, 3.90)
Cue D	1.02 (0.99, 1.05)	0.75 (0.38, 1.47)	0.68 (0.35, 1.33)
Cue E	1.04 (1.01, 1.07)	2.13 (1.01, 4.51)	0.97 (0.45, 2.06)
Cue F	1.05 (0.92, 1.21)	–	0.50 (0.03, 8.09)
Cue G	0.99 (0.93, 1.05)	1.33 (0.37, 4.76)	1.07 (0.32, 3.60)
Subsequent vs. Initial	1.00 (0.98, 1.02)	0.91 (0.65, 1.26)	1.09 (0.78, 1.52)
Doctor vs. Patient-initiated	1.00 (0.98, 1.02)	0.89 (0.55, 1.44)	1.36 (0.84, 2.21)
Primary clinician response categories
Explicit vs. non-explicit	1.01 (0.99, 1.02)	1.26 (0.91, 1.75)	0.96 (0.69, 1.32)
Provide vs. reduce space	0.98 (0.96, 1.00)	0.88 (0.52, 1.48)	0.90 (0.54, 1.51)
Secondary clinician response categories
Neutral/passive(NPSi, NPBc, NPAc)	0.98 (0.96, 1.00)	0.94 (0.61, 1.47)	0.88 (0.57, 1.36)
Acknowledgement(NPAc, EPAAc, EPCAc)	1.00 (0.98, 1.02)	0.85 (0.54, 1.34)	0.97 (0.61, 1.53)
Exploring(NPAi, EPAEx, EPCEx)	1.00 (0.99, 1.02)	1.36 (0.94, 1.96)	1.00 (0.70, 1.43)
Explicit response to emotion(EPAAc, EPAEx, EPAEm)	0.96 (0.93, 1.00)	1.34 (0.63, 2.87)	0.69 (0.35, 1.35)
Any empathy(NPIm, EPAEm)	1.01 (0.97, 1.05)	0.55 (0.23, 1.31)	1.17 (0.46, 2.97)
Gives information/advice(NRIa, ERIa)	1.03 (1.00, 1.05)	1.19 (0.65, 2.16)	0.99 (0.55, 1.76)
Any blocking(NRIg, NRSd, ERSw, ERAb)	1.01 (0.98, 1.04)	1.15 (0.60, 2.19)	1.22 (0.64, 2.33)

^aRandom intercept multilevel logistic regression models (accounts for clustering of emotional expressions within encounters and of patients within clinicians).

4. Discussion and conclusion

4.1. Discussion

Our study examined the influence of patient and clinician characteristics on emotional communication. As in our prior study, we found significant racial differences in both patients’ expression of emotion and clinicians’ responses to patients’ emotional expressions. African American patients were less forthcoming and direct in expressing their emotions than other (predominantly white) patients, and clinicians were less likely to respond to African American patients with “permissive” communication behaviors that allow patients to continue expressing their emotions. Most notably, our results confirm the most troubling finding of our prior study, which showed that clinicians were more likely to avoid or actively block further conversation about emotional issues among African American patients.

In our prior study, which included a smaller sample of 19 clinicians and 43 patients, we did not find racial differences in patients’ expression of emotion. However, in that study, we did not examine differences in the subtypes of emotional cues as we did in the current study. Our finding that African American patients were more likely to express emotion using the subtle mechanism of repeating an ostensibly neutral comment (Cue E) sheds light on possible explanations for racial differences in emotional dialogue, as discussed below.

Differences in the results of our prior study compared to our current findings may be attributable to the different contexts in which the two studies were conducted. Our prior study examined initial interactions with patients who were new to HIV care, whereas our current study enrolled established patients. In initial encounters, providers are generally unfamiliar with individual patients, but there may be greater unfamiliarity with African American patients due to the social and cultural distance imposed by racial differences [23–25]. In that context (prior study), we found that clinicians were less likely to actively explore emotional issues with African American patients than with white patients. At follow-up visits (current study), where one would expect the “distance” between patients and providers to have diminished, we found that white patients were more likely than African Americans to bring up emotional issues with clinicians spontaneously. Clinicians were more likely to respond to African American patients’ emotions by explicitly calling them out while responding more often to white patients’ emotions with neutral or passive communication intended to encourage the patient to continue talking. These findings suggest that over time, familiarity and comfort with emotional conversations between patients and providers may increase more for white patients than for African Americans.

The most concerning of our study findings, which was also observed in our prior study, was that in response to patients’ emotional expression, clinicians were more likely with African American patients to actively avoid or shut down conversation about emotional issues. Our analysis did not reveal underlying explanations for this finding, but several possibilities are worth considering. First, this finding may reflect racial bias, either conscious or unconscious, on the part of clinicians. Prior research has demonstrated that unconscious racial bias among clinicians is associated with less patient-centered communication and lower ratings of care among African American patients [26]. Another possibility is that there may be cross-cultural differences in the expression of emotion, such that providers, who in our study were predominantly white, are less apt to recognize subtle emotional expression among African American patients and, therefore, more likely to miss emotional cues and move on to other topics. Whatever the underlying cause, this pattern of communication may result in a dysfunctional cycle, where African American patients, having had emotional issues ignored or closed off by clinicians, may be less forthcoming and direct about their emotions, which in turn might further impair recognition and acknowledgement of emotion by clinicians, which then might make patients even more reticent to open up, and so on.

In addition to our analysis of patient race and verbal behavior, we assessed other patient and clinician characteristics. We did not observe significant differences in the expression or handling of patients’ emotional expressions based on patient age or gender. In our prior study, there were no significant differences by patient gender, but older patients expressed emotion slightly more with cues vs. concerns. That finding was not verified in the current study. We also found no clear pattern of association between clinician characteristics and emotional communication.

There are several limitations to this study. First, we analyzed transcripts from audio-recorded visits and, therefore, could not capture non-verbal behaviors that might have been observed in videotaped encounters. If patients from different backgrounds differentially conveyed emotion non-verbally, or if clinicians responded non-verbally to a greater or lesser extent with African American compared to white patients, our results may have over- or under-estimated actual disparities in the expression and handling of patient emotions. Second, we examined numerous outcomes in our analyses, increasing the possibility that some findings may have been statistically significant by chance. Our finding, however, of racial differences in clinicians’ blocking emotional dialogue was observed in our prior study as well as in this study with a much larger group of clinicians and patients, suggesting that this is a real disparity. Third, due to our small pool of non-white clinicians, we could not reliably examine the impact of clinician race nor patient-clinician concordance on emotional talk. Fourth, our study took place in the context of HIV care, where emotional communication may be more or less prevalent and essential than in other care settings. It would be useful for future studies to evaluate the association of patient race and emotional communication in other clinical contexts. Finally, although all patients in our study were established in HIV care for at least six months, we did not have data on the actual duration of relationships, limiting our ability to determine if emotional communication changes over time. The fact that the racial differences we observed were similar to those seen in our previous study, which examined only first visits, suggests that those associations were not likely to have been confounded by relationship duration.

4.2. Conclusion

African American patients established in HIV care were less likely to express their emotions spontaneously to clinicians compared to patients of other races. Clinicians were less likely to respond to African American patients’ emotions by providing “permissive” responses intended to allow patients to continue expressing their emotions. They were also more likely to avoid and block emotional conversations with African American patients. These deficits in emotional communication may contribute to mistrust between patients and providers [27], which may, in turn, contribute to well-documented racial disparities in the overall quality of health care [28].

4.3. Practice implications

Attending to patients’ emotional issues is a crucial aspect of patient-centered care and is associated with higher self-efficacy, greater patient satisfaction, decreased anxiety, and improved clinical outcomes [29,30]. Our findings of racial disparities in clinician responses to patients’ emotions should serve as a call to clinicians, and to those who provide communication skills training to clinicians, to raise awareness of potential racial and cultural differences in emotional expression, and potential biases or “blind spots” in discerning emotional cues and responding to them appropriately. Better recognizing and addressing patients’ emotional issues holds the potential to bridge the social distance that may exist in encounters with African American and other minority patients, and thereby strengthen patient-provider relationships in a way that helps reduce racial disparities in health care delivery and outcomes.