The consequences of emotionally evocative patient behaviors on emergency nurses’ patient assessments and handoffs: An experimental study using simulated patient cases

Abstract

Background:

Engaging with human emotions is an integral but poorly understood part of the work of emergency healthcare providers. Patient factors (e.g., irritable behavior; mental illness) can evoke strong emotions, and evidence suggests that these emotions can impact care quality and patient safety. Given that nurses play a critical role in providing high quality care, efforts to identify and remedy factors that may compromise care are needed. Yet to date, few experiments have been conducted.

Objective:

To examine the effects of emotionally evocative patient behavior as well as the presence of mental illness on emergency nurses’ emotions, patient assessments, testing advocacy, and written handoffs.

Design:

Experimental vignette research.

Setting:

Online experiment distributed via email between October and December 2020.

Participants:

Convenience sample of 130 emergency nurses from seven hospitals in the Northeastern United States and one hospital in the mid-Atlantic region in the United States.

Methods:

Nurses completed four multimedia computer-simulated patient encounters in which patient behavior (irritable vs. calm) and mental illness (present vs. absent) were experimentally varied. Nurses reported their emotions and clinical assessments, recommended diagnostic tests, and provided written handoffs. Tests were coded for whether the test would result in a correct diagnosis, and handoffs were coded for negative and positive patient descriptions and the presence of specific clinical information.

Results:

Nurses experienced more negative emotions (anger, unease) and reported less engagement when assessing patients exhibiting irritable (vs. calm) behavior. Nurses also judged patients with irritable (vs. calm) behavior as more likely to exaggerate their pain and as poorer historians, and as less likely to cooperate, return to work, and recover. Nurses’ handoffs were more likely to communicate negative descriptions of patients with irritable (vs. calm) behavior and omit specific clinical information (e.g., whether tests were ordered, personal information). The presence of mental illness increased unease and sadness and resulted in nurses being less likely to recommend a necessary test for a correct diagnosis.

Conclusions:

Emergency nurses’ assessments and handoffs were impacted by patient factors, particularly irritable patient behavior. As nurses are central to the clinical team and experience regular, close contact with patients, the effects of irritable patient behavior on nursing assessments and care practices have important implications. We discuss potential approaches to address these ill effects, including reflexive practice, teamwork, and standardization of handoffs.

Tweetable abstract:

Experimental evidence links irritable patient behaviors to lower quality emergency department nurse handoffs, which may compromise patient safety @(lindamisbell) @(Nathan_Huff_1).

1. Introduction

1.1. Background

According to the Code of Ethics adopted by the American Nurses’ Association (2015), nurses have a moral imperative to treat all patients with dignity and respect while advocating for the health and safety of every patient (American Nurses’ Association, 2015) in accordance with the principle of distributive justice (Varkey, 2021). Nurses are expected to ensure timely access to safe, high-quality care across the patient journey regardless of how they feel about their patients (Bolton, 2001) and regardless of specific patient attributes or conditions (e.g., mental illness) for which nurses may possess negative attitudes (de Jacq et al., 2016). Yet, despite the ubiquity of emotional experiences associated with caring for patients, dealing with emotions remains a poorly understood part of the work of emergency healthcare providers (Croskerry et al., 2008; Croskerry et al., 2010; Committee on Diagnostic Error in Health Care and Institute of Medicine, 2015). Further, given that rates of diagnostic error are elevated in the emergency department relative to other clinical contexts (Graber et al., 2012; Hussain et al., 2019; Croskerry and Sinclair, 2001), provider emotion may play a particularly important role in patient care and safety in the emergency department (Croskerry et al., 2008; Croskerry et al., 2010; Isbell et al., 2020a; LeBlanc et al., 2015; Kozlowski et al., 2017).

A large-scale qualitative study recently conducted with emergency nurses and physicians revealed a range of patient factors (e.g., unrealistic expectations), hospital factors (e.g., staff communication), and system-level factors (e.g., overcrowding) can trigger negative emotions. Importantly, providers reported that these emotions result in compromised patient care, for example, by reducing time spent with patients, or by obtaining incomplete medical histories (Isbell et al., 2020a). Anger in particular can adversely influence clinical decision-making and behavior. Isbell and colleagues found that emergency physicians and nurses who recalled their own recent patient encounters that led them to feel angry reported engaging in behaviors that may have had a detrimental effect on patient care (e.g., delaying or failing to provide a necessary exam or treatment, ending the encounter prematurely) (Isbell et al., 2020b).

Patient factors are a particularly potent generator of intense provider emotions. Disruptive patient behaviors, like demanding services, questioning expertise, and shouting, can generate strong negative emotions (i.e., frustration, irritation, anger) (Gerrard and Riddell, 1988) and impact providers’ information processing and clinical reasoning (Croskerry et al., 2008; Croskerry et al., 2010) by promoting the use of cognitive heuristics (i.e., mental shortcuts). A wide range of heuristics have been documented in both the psychology (Tversky and Kahneman, 1974) and patient safety (Croskerry, 2009) literatures. Premature closure, for example, is common in clinical practice and occurs when a provider arrives at a conclusion about a patient’s condition quickly without carefully considering a broad array of possibilities. Such heuristics can lead to clinical errors, and research in psychology demonstrates that anger can increase reliance on heuristics (Bodenhausen et al., 1994; Parker and Isbell, 2010; Small and Lerner, 2008).

Anger and frustration may also reduce care quality and trigger diagnostic errors if providers need to allocate some of their mental resources to dealing with a patient’s behavior (Mamede et al., 2017; Schmidt et al., 2017). In a written vignette study, for example, internal medicine residents exposed to patients who engaged in either disruptive (vs. neutral) behaviors made more diagnostic errors, reported fewer clinical results, and recalled more patient behaviors (Mamede et al., 2017). Given that residents spent similar amounts of time on the patient cases, the researchers concluded that mental resources were diverted away from important aspects of the clinical case to the patient behavior. A recent review of research in social cognitive and affective sciences suggests that emotions may have wide-ranging impacts on patient care and safety (Liu et al., 2022).

Further, patients with mental illness may be especially vulnerable to the effects of negative provider emotions on care. In the study by Isbell et al. discussed earlier, the emotions that emergency providers reported experiencing during their own recent encounters with angry patients were similar to those reported in response to their recent encounters with a patient with a mental illness. Both were highly negative and associated with lower provider engagement, lower perceived care quality, and greater perceived patient safety risks compared to positive patient encounters (Isbell et al., 2020b). Importantly, these patient groups that represent a growing proportion of emergency department visits (Owens et al., 2010; Weiss et al., 2016) are considered ‘difficult’ (Jackson and Kroenke, 1999; Hahn et al., 1996), experience healthcare disparities (Druss, 2000; Sullivan et al., 2006; Sullivan et al., 2015; De Hert et al., 2011; McGinty et al., 2015; Graber et al., 2000), and face considerable stigma (Isbell et al., 2023; Dinos et al., 2004; Corrigan et al., 2014; Knaak et al., 2015), all of which may exacerbate the effects of negative provider emotions. Taken together, emotionally evocative patients (e.g., patients exhibiting irritable behavior; patients with mental illness) may trigger faulty cognitive reasoning (e.g., premature closure) leading to patient safety concerns, lower-quality care, and diagnostic errors (LeBlanc et al., 2015; Kozlowski et al., 2017; Heyhoe et al., 2016).

1.2. The role of nurses in the emergency department

Although nurses manage a wide variety of emotions while providing patient care (Bolton, 2001) and nurses make up the largest segment of the health care profession, research to date has largely focused on physician responses (Mamede et al., 2017; Schmidt et al., 2017). This is despite the fact that nurses experience close and regular engagement with patients (Committee for Assessing Progress on Implementing the Recommendations of the Institute of Medicine Report The Future of Nursing, 2016) and play a critical role in patient care and the diagnostic process (Thirsk et al., 2022). In the emergency department, nurses are responsible for triaging patients and conducting initial and ongoing assessments, setting the timeline for pain management, advocating for their care and making care recommendations, and making decisions that affect patient outcomes (American Nurses’ Association, 2015; Bergman, 2012; Manojlovich et al., 2022). Nurses’ assessments are important components of emergency care, as they determine the urgency and treatment needs of patients and inform subsequent clinical decisions. Nurses also play a critical role in evaluating and managing pain, which is the most common complaint among patients seeking care in the emergency department (Graham, 2002), although most patients report their acute pain is not adequately treated in the emergency department (Todd et al., 2007).

Additionally, nurses convey critical information to other providers through patient handoffs. An accurate and complete relay of information is crucial to patient care, and without widely implemented, standardized handoff methods (Patterson and Wears, 2010) it is largely up to nurses to choose what information is exchanged and how it is transmitted (Holly and Poletick, 2014). Emergency nurses’ emotional experiences may impact their patient assessments and advocacy, and compromise handoff quality and content, thereby influencing a crucial link in patient care (Narayan, 2019).

1.3. Objectives

The current research fill gaps in the literature on the influence of patient factors on provider emotions, clinical judgment, and behavior (Isbell et al., 2020a; Isbell et al., 2020b; Mamede et al., 2017; Schmidt et al., 2017; Liu et al., 2022). Using an experimental design and computer-simulated patient encounters to present identical patients behaving calmly or irritably with or without a mental illness, we examine how patient factors impact emergency nurses’ emotions, patient assessments, testing advocacy, and written handoffs. Specifically, we hypothesize that patients exhibiting irritable (vs. calm) behaviors will elicit more negative emotions in nurses (e.g., greater anger), will be assessed more negatively (e.g., less likely to recover), and will receive lower quality handoffs (e.g., detailed patient information will be mentioned less frequently).

2. Methods

2.1. Case development and study design

Four multimedia patient cases were developed with emergency physicians (PS) and then adapted for this study by an expert nurse (KTG). The adapted cases were then iteratively pilot tested with four emergency nurses before being used in the study. Specifically, pilot nurses completed the study one at a time while on a video call with members of the research team. Each nurse completed the study while sharing their screen and talking aloud to provide us with a concrete understanding of how they interpreted the clinical cases and materials. They also responded to follow-up questions from our team. Following each pilot nurses’ participation and feedback the study was revised. The revised study was then completed by the next pilot nurse. After the final nurse completed the pilot process and offered no substantive feedback, no changes were implemented. The study completed by the final pilot nurse was identical to the study participants experienced. Pilot nurses were not included in the study sample.

Each of the four clinical cases included a patient presenting with physical symptoms corresponding to a diagnosis (i.e., venous sinus thrombosis, myocardial infarction, adrenal insufficiency, malaria). Diagnoses were selected because they were an appropriate difficulty to diagnose and required diverse treatments. Once the cases were finalized after piloting, we created four versions of each case by experimentally manipulating patient behavior (irritable vs. calm) and mental illness (present vs. absent).

All participants assessed all four clinical cases, and every participant received two patients displaying calm behavior and two displaying irritable behavior, each with and without a mental illness. To achieve this, we created four patient case sets including unique combinations of all four experimental conditions distributed across the four clinical cases (see Fig. 1). For example, patient case set one included the venous sinus thrombosis case in which the patient was calm and had no mental illness, the myocardial infarction case in which the patient was calm and had a mental illness, the adrenal insufficiency case in which the patient was behaving irritably and had no mental illness, and the malaria case in which the patient was behaving irritably and had a mental illness. Participants were randomly assigned to one of the four possible patient case sets using a randomizer element in Qualtrics (2011) resulting in every participant receiving all four clinical cases and some combination of all four experimental conditions. Within the patient case set to which they were assigned, participants completed their four clinical cases in a randomized order. This approach resulted in a 2 (irritable vs. calm patient behavior) × 2 (mental illness present vs. absent) × 4 (patient case set) mixed design, with the first two factors within subjects and the last factor between subjects.

Fig. 1.

Random assignment process using patient case sets.

2.2. Study participants and recruitment

Based on an a priori power analysis assuming a medium effect (f = 0.25), we aimed to recruit 132 emergency nurses to achieve 80% power. A medium effect size was assumed because while our experimental manipulation was robust, many factors are known to influence clinical reasoning. Further, a medium effect was our smallest effect of interest. We recruited a convenience sample from hospitals across the Northeastern United States (US) and one hospital in the Mid-Atlantic region of the US. Study invitations were sent via email between October 2020 and December 2020. At six hospitals, invitations were sent from emergency medicine nurse managers to all emergency medicine nurses at the hospital. At the other two hospitals, invitations were shared by nurses who were forwarded invitations from nurses working at the original six hospitals.

The email invitation described the project as a “study on patient care and decision-making among emergency nurses,” in which “you would be presented with information about four randomly selected patients, be asked to make judgements about each one, and complete some questionnaires.” The invitation also stated: “We will provide $100 compensation in appreciation of your time (about 55 min).” Importantly, we did not provide any information about our interest in emotionally evocative patient behaviors or mental illness in the study recruitment or consent materials. This was done to avoid self-selection bias, whereby participants may have only enrolled because they held a priori attitudes or beliefs about individuals with mental illness or their own ability to handle emotionally evocative patient behaviors. Although mentioning study payment may have risked eliciting demand characteristics whereby participants responded in a manner they perceived as socially desirable, its inclusion was necessary to facilitate recruitment and is common practice.

Nurses who indicated interest in participating were sent a personalized, one-time-use Qualtrics study link (Qualtrics, 2011). Nurses were given two weeks to participate. Those who did not access the study were sent reminder emails, provided a new link, and offered an additional week. Nurses who did not complete the study after three weeks were again re-invited, sent another link if interested, and given one week to participate. Given our use of a convenience sample and our recruitment method, it is not known how many nurses may have received our study invitation. However, of the 165 nurses who requested a link, 130 completed the study (78.8%). Nurses were compensated US $100.

2.3. Research ethics

The study was approved by our institutional review board (IRB#196: 2016–3291). Participants reviewed an informed consent document and provided their consent prior to completing the study. The informed consent document described our general interest in healthcare decision-making, that there were no known risks associated with this study, our data handling policies, and that payment details would be provided after study completion. Our informed consent document included no information about our interest in emotionally evocative patient behaviors or mental illness.

2.4. Procedure

Upon consent, nurses were instructed to complete the study alone in a quiet location, and that they would evaluate four patients, imagining each patient arrived at their emergency department. The evaluation process for all cases was identical. Fig. 2 illustrates the study flow. (See recruitment materials, a sample clinical case, additional measures, and anchor labels for all measures in the Supplementary file).

Fig. 2. Study flow.

Note. Blue blocks represent patient information provided to participants. Yellow blocks represent tasks and questionnaires completed by participants. The gray block includes questionnaires unrelated to the current report (Huff et al., 2023).

2.4.1. Brief vignette and electronic health record

Participants were provided with a vignette and patient electronic health record. The vignette described the patient’s behavior (irritable vs. calm) and chief complaint. Irritable (e.g., insulting the nurse) and calm (e.g., thanking the nurse) patient behaviors were informed by qualitative research with emergency nurses (Isbell et al., 2020a). All electronic health records included basic medical information and were modeled off existing electronic health records. After reviewing online examples of electronic health records, we found many elements of electronic health records to be idiosyncratic to hospital systems. To avoid participants being relatively more or less familiar with the format of our materials, we aimed to create a template that was clear and understandable to all participants. This meant our electronic health records were less complex than real-world records. To make sure all participants would be able to use the electronic health record, expert nurses evaluated its readability and interpretability during the pilot process.

In mental illness conditions, the patient’s history included a mental illness diagnosis and appropriate medications. In no mental illness conditions, this information was excluded. The mental illness diagnoses were anxiety (venous sinus thrombosis case), schizophrenia (myocardial infarction case), bipolar disorder (adrenal insufficiency case), and depression/substance use (malaria case). Irritable (vs. calm) descriptions were also embedded in electronic health records.

2.4.2. Patient encounter video

Nurses then watched a 3–4-minute video-recorded encounter between the patient (seated on an examination table) and an out-of-frame female nurse. The videos were filmed in a medical simulation center with adult standardized patients (i.e., trained actors), including two White males (myocardial infarction and malaria cases) and two White females (venous sinus thrombosis and adrenal insufficiency cases). In irritable conditions, the patient behaved angrily and engaged in behaviors known to induce anger in nurses (e.g., demanding services, questioning the nurses’ experience) (Isbell et al., 2020a). In calm conditions, the patient behaved calmly and did not engage in such behaviors. In cases that included a mental illness, the patient mentioned their mental illness briefly at the end of their medical history. The nurses’ questions for each clinical case were specific to the patient’s complaint, but the four versions of each clinical case were identical in content other than the experimental manipulations. Common questions were included, such as “Can you describe your symptoms in more detail?” Nurses did not follow up with questions concerning mental illness to control the clinical information provided to participants.

2.4.3. Preliminary patient assessment

After viewing the encounter video, nurses provided a preliminary assessment of each patient, described in greater detail in the measures.

2.4.4. Physical exam results, testing recommendations, and testing advocacy

Nurses were next informed that the patient had been examined by a physician and were presented with physical exam results. Results were identical across experimental conditions but varied by clinical case. Next, nurses were told the physician planned to order unspecified laboratory tests or imaging studies for the patient. Nurses that indicated agreement with the physician’s decision were then able to recommend tests and indicate their willingness to advocate for testing.

2.4.5. Patient handoff, final patient assessment, and prognosis

Nurses then provided a written handoff and responded to a final set of patient measures assessing the patient and their prognosis.

2.4.6. Nurses’ emotions and engagement during cases and final demographic questionnaire

After assessing all four patients, nurses completed measures about their emotions and engagement during each case. Lastly, nurses completed several questionnaires unrelated to the current report (Huff et al., 2023), provided demographic information, and were debriefed. The debriefing form was presented to participants online, and included details regarding our research question, contact information, and further reading.

2.5. Measures

To simulate emergency department evaluations, dependent measures were included throughout each patient encounter (see Fig. 2). We developed measures for this study through rigorous and iterative dialogue with emergency medicine consultants, collaborators, and pilot participants. Specifically, all measures were inspired by prior literature and our research team’s clinical experience and were developed following discussions with expert emergency medicine physicians and nurses (including PS and KTG). Through the piloting process, pilot emergency nurses provided substantial feedback on the content and presentation of measures. Some measures have been used in prior studies with emergency physicians and were revised following feedback from those studies, and some measures were developed specifically for this research. When similar items are combined to create a composite, we describe indicators of the items’ internal consistency reliability using Cronbach’s alpha.

As this is one of the first studies on this topic, we explored the extent to which outcomes spanning the entire care process are influenced by emotion. We therefore included dependent variables related to all stages of the clinical encounter and assessment. Key measures are reported here, and additional measures are described in the Supplementary file.

2.5.1. Preliminary patient assessment

Nurses indicated the pain level they believed the patient was likely to report, as well as the pain level that they believed the patient was experiencing on unnumbered sliding scales scored from 0 (no pain at all) to 10 (worst pain imaginable). A measure of perceived pain exaggeration was created by subtracting the second score from the first. Nurses then indicated the extent to which they believed the patient was a good historian, how honest and accurate they believed the patient was when reporting their history, and how honest and accurate the patient was when reporting their symptoms. These five historian items were completed on unnumbered sliding scales scored from 0 (not at all) to 100 (very much). The items demonstrated excellent reliability (α = 0.93) and were averaged to create a 5-item historian composite score.

2.5.2. Physical exam results, testing recommendations, and testing advocacy

After reviewing the physical exam, nurses indicated their clinical understanding using a single item: “To what extent do you know what was going on in the case?”. Nurses responded on an unnumbered sliding scale scored from 0 (I don’t know at all) to 100 (I completely understand).

Nurses were asked to indicate whether they agreed with the physician’s decision to order tests for each patient, and their responses were coded as 0 (no) or 1 (yes). Those who agreed could recommend up to ten laboratory tests or imaging studies using free response boxes. Additional free response boxes (up to 10 total) would only appear if they entered text into a prior box. Nurses were then presented with a new screen and asked to indicate the likelihood they would advocate for each of their recommended tests if the physician decided to discharge the patient without ordering it. These advocacy scores, completed individually for each test, were recorded on unnumbered sliding scales scored from 0 (not likely to recommend) to 100 (very likely to recommend).

The number of tests a nurse recommended for each case was summed, and a mean advocacy score was calculated by summing the advocacy scores for all recommended tests and dividing by the number of recommended tests. This average advocacy score indicates how willing a nurse was to advocate for the overall testing plan they recommended for the patient. One limitation of this approach is that the specific tests a nurse advocates for differ across participants (i.e., one nurse’s advocacy score may be based on a single, necessary test, whereas another nurse’s advocacy score may be based on multiple, unnecessary tests). Despite this, a mean advocacy score is useful for capturing how generally willing a nurse is to defend or advocate for a patient’s care.

An expert nurse (KTG) coded the testing recommendations for their necessity in providing timely care for the patient’s correct diagnosis. Recommended tests were coded as 0 (unnecessary) or 1 (necessary). Additionally, whether a nurse recommended at least one necessary test for a specific case was coded 0 (no) or 1 (yes). While there are many tests that may be necessary and important for caring for patients, these clinical cases were designed such that there were specific test recommendations that would be definitive in reaching the correct diagnosis (e.g., an abdominal or pelvic ultrasound was necessary for reaching the correct diagnosis in the adrenal insufficiency case). The correct diagnosis could not be arrived at without the results from one of these necessary tests, and these results would definitively reveal the correct diagnosis. Although nurses were never provided with test results, a nurse recommending at least one necessary test suggests that the nurse is “on the right track” for a case. In this sense, recommending at least one necessary test demonstrates that a nurse is attentive in their care of the patient, as it would assist in correctly diagnosing the patient.

2.5.3. Patient handoff

To provide a handoff, nurses were asked to imagine that their shift was ending and that they would give a verbal report to the nurse who is assuming care of the patient. Nurses were given a free response text box, prompted to type what they would say, and instructed to include “all information you would tell this nurse if you were actually interacting with the nurse.” While completing their handoff, nurses were able to review the patient’s electronic health record and physical exam by clicking pop-up boxes containing that information.

Handoffs were coded for their inclusion of specific patient information (e.g., personal information, vital signs, tests or additional care recommendations) and general patient descriptions (e.g., mentioning patient affect, inclusion of negative descriptions, inclusion of positive descriptions). All handoff variables were coded either 0 (attribute absent) or 1 (attribute present). For example, handoffs were coded as 1 (attribute present) for “recommending tests or additional care” if nurses included specific test recommendations (e.g., Computed Tomography), recommended a consultant or specialist before patient discharge, or indicated that test results were pending.

Examples of negative patient descriptions include calling the patient an expletive, and referring to the patient as fussy, irritable, angry, frustrated, agitated, rude, grumpy, difficult, or a poor historian. While “poor historian” may be considered a relatively benign negative description, we code it as negative in line with recent work describing how its use can “transmit judgment and bias to other care team members and should be avoided” (Healy et al., 2022). Examples of positive descriptions include referring to the patient as nice, cooperative, pleasant, or a good historian. We coded “poor” and “good” historian as a negative or positive description if the nurse used those specific terms, as well as if they used conceptually similar terms (e.g., bad historian, not the best historian, great historian, excellent historian).

An initial codebook was created, which was then iteratively refined through independent coding and comparison. A team of four research assistants independently coded handoffs, and systematically reviewed all coding across two rounds of review. In all coding rounds, all coders were blind to the experimental condition in which the handoffs they were coding were produced. Discrepancies were resolved through discussion. (See additional methodological details and final codebook in the Supplementary file).

2.5.4. Final patient assessment and prognosis

After writing their handoff, nurses reported the extent to which they were optimistic the patient would cooperate with the requests of the care team, return to work soon, and recover from the chief complaint that led to this visit. All three prognosis measures were completed on unnumbered sliding scales scored from 0 (not at all optimistic) to 100 (very optimistic). Nurses also indicated how likable they found the patient using an unnumbered sliding scale scored from 0 (not at all likeable) to 100 (very likeable).

2.5.5. Nurses’ emotions and engagement during cases

After assessing all four patients, nurses were presented with a picture of each patient in a random order and reported the extent to which they felt angry, frustrated, irritated, sad, uneasy, calm, happy, pleased, engaged, and empathic while viewing each patient encounter. All items were completed in a random order on unnumbered sliding scales scored from 0 (not at all) to 100 (very much). An anger composite (α = 0.87) was created by averaging nurses’ responses to the angry, frustrated, and irritated items.

2.6. Data analysis

Data were analyzed using SPSS v.25. Missing data was minimal and addressed using listwise deletion. For continuous dependent variables, we ran mixed-design analyses of variance (ANOVAs) with patient case set included as a between-subjects factor, and patient emotion (irritable vs. calm) and mental illness (present vs. absent) as within-subjects factors. No consistent effects emerged as a function of patient case set, so this factor was omitted in subsequent analyses. Categorical dependent variables were analyzed using chisquare tests. Because mental illness diagnoses differed across clinical case, see effects of mental illness separately for each clinical case in the Supplementary file. Effect sizes presented are partial eta-squared, in which small, medium, and large effects are defined as 0.001,0.059, and 0.138 respectively (Cohen, 2013).

3. Results

3.1. Participant characteristics

One hundred and thirty emergency medicine nurses completed the study. Most participants were White, non-Hispanic women working in academic institutions. The average length of experience in emergency nursing was over 8 years. Most participants came from a large healthcare system composed of five hospitals located in the Northeastern US, while a minority of participants worked in the Mid-Atlantic region of the US. See Table 1 for participant characteristics.

Table 1.

Participant characteristics (n = 130).

Years	M	SD
Age	37.06	9.64
Nursing experience	11.64	9.19
EM experience	8.63	7.66
Gender	n	%
Female	106	81.5
Male	23	17.7
Non-binary	1	0.8
Race
White	117	90.0
Black or African American	6	4.6
Asian	4	3.1
White and Black/African American	2	1.5
White and American Indian or Alaskan Native	1	0.8
Ethnicity
Not Hispanic and/or Latino	123	94.6
Hispanic and/or Latino	7	5.4
Academic affiliation of work institution
Academic	98	75.4
Nonacademic	29	22.3
Other	3	2.3
Location of work institution
Northeastern US	112	86.2
Mid-Atlantic region of US	15	11.5
No work location provided	3	2.3

Note. EM = emergency medicine. The “other” responses for academic affiliation of work institution were: “both” (n = 1), “unsure” (n = 1), and a missing response (n = 1).

3.2. Nurse emotions, engagement, and patient assessments

After testing the effectiveness of our experimental manipulation by confirming that patients displaying irritable behaviors were perceived to be more irritable and less calm than patients displaying calm behaviors (see Methodological details and analyses in the Supplementary file), we examined nurses’ emotions, engagement, and patient assessments. Effects of patient behavior are summarized in Table 2 and all results are described below.

Table 2.

Nurses’ emotions, engagement, preliminary assessments, and final assessments of patients displaying calm and irritable behaviors.

	Calm patient behavior	Irritable patient behavior	Calm vs. irritable patient behavior
	M (SE)	M (SE)	df	F	p	${\mathrm{\eta }}_{\mathrm{p}}^2$
Nurses’ emotions and engagement
Angry (composite)	14.42 (1.45)	51.19 (2.15)	1, 129	277.03	<.001	.682
Uneasy	24.37 (2.07)	41.51 (2.39)	1, 129	46.60	<.001	.265
Sad	26.52 (2.12)	22.70 (2.02)	1, 129	3.00	.086	.023
Happy	27.13 (1.99)	14.01 (1.42)	1, 129	47.57	<.001	.269
Calm	74.99 (1.81)	54.93 (2.40)	1, 129	78.67	<.001	.379
Empathic	64.37 (1.94)	44.15 (1.97)	1, 128	86.96	<.001	.405
Engaged	72.28 (1.76)	55.94 (2.13)	1, 129	66.99	<.001	.342
Preliminary patient assessment
Good historian (composite)	74.16 (1.43)	57.42 (1.63)	1, 129	78.60	<.001	.379
Pain exaggeration	1.02 (0.13)	2.50 (0.20)	1, 127	50.63	<.001	.285
Clinical understanding	64.59 (1.55)	60.65 (1.43)	1, 127	3.89	.051	.030
Final patient assessment and prognosis
Likely to cooperate with care team	86.19 (1.27)	50.71 (1.90)	1, 129	249.96	<.001	.660
Likely to recover	83.25 (1.37)	75.03 (1.77)	1, 128	17.60	<.001	.121
Likely to return to work soon	66.82 (1.83)	56.24 (2.05)	1, 128	21.63	<.001	.145
Likeable	75.37 (1.51)	32.89 (1.60)	1, 129	423.62	<.001	.767

Note. Statistics presented include mean (M), standard error (SE), degrees of freedom (df), F statistic (F), p value (p), and partial eta-squared (${\mathrm{\eta }}_{\mathrm{p}}^2$).

Sliding scales range from 0 (not at all) to 100 (very much), except for pain exaggeration which ranges from — 10 (maximum underreporting of pain) to 10 (maximum exaggeration of pain).

Angry (composite): angry, frustrated, irritated; α = 0.87. Good historian (composite): 5 items; α = 0.93.

3.2.1. Nurses’ emotions and engagement during cases

When assessing patients behaving irritably (vs. calmly), nurses reported feeling significantly greater anger and unease, and significantly less happy, calm, empathic, and engaged. Also, nurses reported experiencing greater unease (36.73 vs. 29.15; p = .001; ${\mathrm{\eta }}_{\mathrm{p}}^2=0.085$) and sadness (27.10 vs. 22.12; p = .015; ${\mathrm{\eta }}_{\mathrm{p}}^2=0.045$) in response to a patient with a mental illness (vs. without).

3.2.2. Preliminary patient assessment

Nurses judged patients exhibiting irritable (vs. calm) behaviors to be worse historians as captured by the 5 -item historian composite score. Nurses also rated patients with irritable behaviors as exaggerating their pain to a greater extent than those behaving calmly. Further, after receiving identical physical exam results, nurses reported diminished confidence in their understanding of patients’ cases when patients displayed irritable (vs. calm) behaviors.

3.2.3. Final patient assessment and prognosis

Nurses expected patients with irritable (vs. calm) behaviors to be less likely to cooperate with the care team, to recover from their chief complaint, and to return to work soon. Nurses also evaluated patients displaying irritable behaviors as less likable than those displaying calm behaviors. Further, compared to patients without a mental illness, nurses perceived patients with a mental illness as less likely to return to work soon (59.33 vs. 63.74; p = .046; ${\mathrm{\eta }}_{\mathrm{p}}^2=0.031$), but counter to expectations, more likable (55.84 vs. 52.41; p = .046; ${\mathrm{\eta }}_{\mathrm{p}}^2=0.031$).

All other effects of patients’ emotionally evocative behavior, mental illness, and the interaction between patient behavior and mental illness on nurses’ emotions, engagement, preliminary assessments, and final assessments were not statistically significant, all p values>.092.

3.3. Testing recommendations and testing advocacy

Across all cases, most nurses agreed with the physician’s decision to order tests (464/520; 89.2%). On average, nurses suggested 3.87 tests per case (SD = 2.45), and largely advocated for these tests (M = 76.40, SD = 28.80). Patient’s emotionally evocative behaviors and the presence of mental illness did not influence nurses’ agreement to test the patient (χ²(3) = 1.76, p = .623), the number of tests recommended by nurses (all p values>.138), or nurses’ mean advocacy scores (all p values>.374). However, the proportion of nurses who suggested at least one necessary test did vary by condition, χ²(3) = 12.58, p = .006. Nurses were less likely to suggest a necessary test when the patient had a mental illness (104/260; 40.0%) than when the patient did not (144/260; 55.4%), χ²(1) = 12.33, p < .001. No statistically significant differences emerged as a function of patient emotion, χ²(1) = 0.12, p = .725.

3.4. Patient handoffs

Table 3 presents the proportions of nurses who mentioned specific patient information and global patient descriptions in their handoffs based on patient emotion and mental illness condition. Handoffs for patients behaving in a calm (vs. irritable) manner were more likely to include the patient’s personal information and tests or additional care recommended. For example, handoffs for patients who behaved calmly included, “…22-year-old female with migraine…”, “…chest x-ray, ekg, laboratory tests pending…”, “…no prior medical history, no known allergies…”, whereas this information was less likely for patients exhibiting irritable behavior.

Table 3.

Nurse handoff frequencies by experimental condition.

	Calm patient behavior			Irritable patient behavior			Grand Total	Calm vs. irritable patient behavior
	No mental illness	Mental illness	Total	No mental illness	Mental illness	Total
Did the nurse mention...	n (%)	n (%)	n (%)	n (%)	n (%)	n (%)	n (%)	χ 2	p
Specific patient information
Personal information	115 (88.5)	116 (89.2)	231 (88.8)	109 (83.8)	104 (80.0)	213 (81.9)	444 (85.4)	4.99	.025
Vital signs	93 (71.5)	81 (62.3)	174 (66.9)	78 (60.0)	78 (60.0)	156 (60.0)	330 (63.5)	2.69	.101
Tests or additional care recommended	71 (54.6)	81 (62.3)	152 (58.5)	68 (52.3)	58 (44.6)	126 (48.5)	278 (53.5)	5.23	.022
Global patient descriptions
Patient affective state	32 (24.6)	38 (29.2)	70 (26.9)	81 (62.3)	75 (57.7)	156 (60.0)	226 (43.5)	57.88	<.001
Negative patient descriptions	12 (9.2)	7 (5.4)	19 (7.3)	79 (60.8)	68 (52.3)	147 (56.5)	166 (31.9)	144.98	<.001
Positive patient descriptions	15 (11.5)	18 (13.8)	33 (12.7)	0 (0.0)	1 (0.8)	1 (0.8)	34 (6.5)	30.24a	<.001

Note. Statistics presented include frequency count (n), percentage (%), Chi-square test statistic (χ²), and p value (p). χ² analyses compare calm and irritable patient behavior conditions (bolded total columns).

^aYates’ correction was employed for the χ² for positive patient descriptions due to n < 5 in the irritable behavior condition.

Conversely, handoffs for patients with irritable behaviors were more likely to include references to patient affect and negative patient descriptions. For example, handoffs for these patients included, “…well appearing but fussy/irritable…”, “…irritable, weakness, poor historian…”, “…will probably be a bit difficult…” “…she is sick but a bad historian…”, whereas this was less likely to appear in handoffs for patients behaving calmly. (See sample handoffs in the Supplementary file).

Of note, handoff length (i.e., total number of typed characters) did not vary as a function of patient’s irritable behavior (p = .459). Further, whether a patient had a mental illness or not did not influence any handoff variables, all p values > .159.

4. Discussion

Across four patient cases, irritable patient behavior and mental illness negatively impacted emergency nurses’ clinical assessments. Whereas a mental illness reduced the likelihood of a nurse recommending necessary tests, irritable patient behavior decreased the overall quality of nurse handoffs, evoked negative emotions, and diminished nurses’ optimism in a positive prognosis. Effect sizes for irritable patient behavior were large and impacted many outcomes; effect sizes for mental illness were moderate and impacted fewer outcomes. By using a computer-simulated experimental procedure, our data isolate patient factors and rule out other co-occurring explanatory variables.

4.1. Adverse effects of irritable patient behavior and mental illness on nursing care

Despite having received identical clinical information, nurses believed patients displaying irritable behaviors were less likely to return to work soon and less likely to recover than patients who displayed calm behavior. Finding that patient behaviors directly shape nurses’ prognostic beliefs has important implications for patients. If a nurse believes a patient is less likely to recover, nurses may subtly communicate this belief in ways that influence the patient’s actual recovery. For example, recent work on such “interpersonal expectancy effects” (Rosenthal, 2002) indicates that a provider’s expectation of a treatment’s success can influence a patient’s subjective pain (Chen et al., 2019). In this way, when nurses have biased attitudes or expectations, these may act as a self-fulfilling prophecy, compromising care quality and safety, particularly for patients who display irritable behavior.

Nurses’ care practices were also influenced by emotionally evocative patient behavior. While handoff length did not differ, handoffs for patients exhibiting irritable (vs. calm) behavior were more likely to incorporate negative patient descriptions and omit personal patient information and recommendations for tests and additional care. These omissions are worrisome, as group interviews with emergency nurses suggest that such information is essential for effective handoffs (Klim et al., 2013), and information transmission failures can have important downstream consequences. Between 2004 and 2013, approximately 65% of sentinel events (i.e., events that result in death, permanent harm, or severe temporary harm) reported to The Joint Commission were traced to communication breakdowns (Sentinel Event Data Root Causes by Event Type, 2004–2013). Research also links inadequate communication to medication administration errors (Keers et al., 2013), surgical error incidents (Gawande et al., 2003), and medical malpractice cases (Malpractice risks in communication failures, 2015).

Handoffs that include negative patient descriptions can also transmit biased information to other providers. Qualitative research has examined how stigmatizing language manifests in patient medical records (Park et al., 2021), and two recent studies found that stigmatizing language exists in online patient records, particularly in Black patients’ records (Himmelstein et al., 2022; Sun et al., 2022). Exposure to stigmatizing language in a medical note can increase negative attitudes towards patients and promote less aggressive pain management among physicians in training (O’Conor et al., 2018). Even seemingly benign language (“poor historian”, “noncompliant”) can transmit bias and adversely influence patient care (Healy et al., 2022; Fisher, 2016; Cox and Fritz, 2022). Thus, the influence of patient’s irritable behavior on nurses’ handoffs can have far reaching effects even when biases are communicated relatively subtly.

Lastly, we found that when a patient had a mental illness, the likelihood of a nurse recommending a critical test necessary for diagnosis was reduced compared to when a patient did not have a mental illness. This may reflect that patient mental illness distracted nurses from attending to other symptoms. Research demonstrates that patients with mental illness sometimes experience diagnostic overshadowing in which their physical symptoms are misattributed to their mental illness (Shefer et al., 2014; van Nieuwenhuizen et al., 2013). These patients may also have more complex clinical presentations, which may complicate clinical assessment. Given the notable health care disparities that exist for patients with mental illness (Sullivan et al., 2006; McGinty et al., 2015), future research is needed to explore these and other possibilities.

4.2. Strategies for reducing adverse effects of emotion on nursing care

The critical role emergency nurses play in patient care (Committee for Assessing Progress on Implementing the Recommendations of the Institute of Medicine Report The Future of Nursing, 2016; Manojlovich et al., 2022; Holly and Poletick, 2014) highlights the importance of not only identifying, but also remedying factors that compromise care quality. While our work does not test such strategies, it is important to examine existing evidence of interventions that may be particularly effective at reducing these effects. First, a greater focus on the value of reflexive practice in nursing along with greater support to allow nurses to engage in this practice may promote learning and improve patient care (Timmins, 2006). Reflexive practice enables clinicians to explore their emotional responses, beliefs, values, and assumptions to promote heightened awareness and reflection on how these factors impact care they provide to patients (English et al., 2022; Dawson et al., 2022; Josephsen, 2014). Engaging in reflexive practice requires metacognition (Josephsen, 2014), which requires self-knowledge, self-evaluation, and applicable knowledge.

Although nurses may learn to engage in reflexive practice during their education and training, it remains unclear how well it is implemented in nursing practice (English et al., 2022). Many barriers may prevent nurses from applying it, such as persistent, disruptive patient behaviors and elevated rates of workplace violence (Aljohani et al., 2021; Ayasreh and Hayajneh, 2021). The COVID-19 pandemic has also placed frontline healthcare workers and their patients under unique stressors (e.g., understaffing, longer wait times, infection risk) (Pariona-Cabrera et al., 2020). These conditions can understandably elicit strong emotions, impact patient care, and create poor conditions for engaging in reflexive practice. Interventions that reduce these work environment stressors (e.g., reduced nurse to patient ratios) would provide time and opportunities for this practice.

Facilitating teamwork, collaboration, and trust among emergency providers may also diminish the ill-effects of patient factors and improve handoffs specifically. Handoffs are complex social interactions (Holly and Poletick, 2014; Frankel et al., 2012), and attention to interpersonal dynamics may help overcome provider miscommunication (Kitzmiller et al., 2017). Teamwork quality has been shown to relate to the quality and safety of care delivery systems (Rosen et al., 2018) and improving teamwork via training interventions has been associated with improved patient safety. Additionally, interventions that standardize the handoff (Starmer et al., 2014; Haig et al., 2006; Riesenberg et al., 2009) process may counteract effects of emotion on handoff quality by ensuring patient information is communicated regardless of patient factors. In line with recent calls from The Joint Commission (2017) and others (Holly and Poletick, 2014; Arora et al., 2008; Welsh et al., 2010), handoff standardization interventions with nurses (Starmer et al., 2017; White-Trevino and Dearmon, 2018) and other healthcare professionals (Starmer et al., 2014; Chladek et al., 2021; Parent et al., 2018; Jorro-Barón et al., 2021) may improve care efficiency and patient safety for all patients, and especially those most vulnerable to handoff errors or biases (i.e., those exhibiting irritable behaviors).

Finally, emotional intelligence, or one’s ability to be aware of and understand emotional states in oneself and others (Salovey and Mayer, 1990; Bulmer Smith et al., 2009), is positively related to nurses’ physical and emotional caring for patients (Nightingale et al., 2018) and has been shown to increase following interventions (Hodzic et al., 2018). Thus, by raising nurses’ awareness of their own emotions and encouraging communication with peers about their emotions, emotional intelligence training may improve key skills needed to engage in reflexive practice, reduce negative effects of emotion on patient care, and benefit nurses, as prolonged experiences of negative emotion place nurses at risk for burnout and mental health problems (Vahey et al., 2004).

Taken together, efforts to promote reflexive practice among nurses and facilitate workplace conditions that support opportunities for reflexive practice are likely to be most effective at reducing effects of patient factors. However, the emergency department is buffeted by system- and societal-level challenges, often serving highly vulnerable patient populations and patient needs while operating with limited resources (Croskerry and Sinclair, 2001; Isbell et al., 2020a). Public policy changes that increase resources and funding are urgently needed to allow emergency departments to manage the complex physical and emotional needs of their patients, and reduce stressful, emotion-eliciting conditions. Addressing large-scale challenges will likely reduce the intensity and frequency of negative emergency department care experiences (Moskop et al., 2009).

5. Limitations and strengths

Our participants are restricted to a convenience sample of nurses recruited from seven hospitals in the Northeastern US and one in the mid-Atlantic region of the US, and our sample is disproportionately White and female, all of which may limit generalizability. Further, our sampling method may have led to self-selection bias in which those who chose to participate in the study are different from those who did not. To reduce this concern, we did not provide nurses with specific information about the study focus in our recruitment materials. Importantly, given that all nurses were randomly assigned to experimental conditions, these sampling concerns may limit generalizability but do not threaten the interpretation of our findings.

Our results reveal effects of emotionally evocative patient behavior and mental illness across four clinical cases, which is a strength. However, all patients in these cases are White adults, highlighting a limitation. This decision was made to avoid confounding of known effects of patient demographic characteristics (Zhang et al., 2020; Institute of Medicine (US) Committee on Understanding and Eliminating Racial and Ethnic Disparities in Health Care, 2003; Blanchard et al., 2003) outside of our research scope. Work is needed to examine how effects of patient behavior may vary across patient and nurse demographics. Also, our mental illness manipulation may have not been very salient, as it appeared once in the electronic health record and patient video. This may explain the small number of mental illness effects but may also be due to nurses holding mixed attitudes towards mental illness (de Jacq et al., 2016). We cannot parse these possibilities without additional research.

Lastly, like all vignette studies, this experiment is limited in external validity. Although we employed detailed patient scenarios presented in multimedia vignettes with interactive elements, our experimental procedures could not capture all relevant real-world factors (e.g., cognitive load, distraction, fatigue). Even so, simulated patient encounters produce excellent experimental control, and research suggests that case vignettes are good proxies for investigating clinical behaviors (Mohan et al., 2014; Peabody et al., 2000). Further, given that emergency department dynamics likely exacerbate the impacts of patient factors, our findings may underestimate real-world effects. Ultimately, experimental work helps to lay the foundation for work in real-world settings. Multi-method research conducted both inside and outside of the emergency department and other clinical contexts is needed to develop a comprehensive understanding of how emotions influence patient care.

6. Conclusions

Although emotional experiences in the emergency department can lead to lower quality patient care, errors, and patient safety risks, research has been limited, and studying such phenomena in real-world contexts is difficult. Utilizing an experimental design, we provide the first evidence that emergency nurses’ emotions, clinical assessments, and handoff quality are impacted by emotionally evocative patient behaviors. As a central part of the clinical care team and the largest component of the healthcare workforce (Committee for Assessing Progress on Implementing the Recommendations of the Institute of Medicine Report The Future of Nursing, 2016), nurses represent a crucial link in care. Thus, these effects may constitute a substantial risk to patient safety. This work is part of a large, complicated puzzle, highlighting an urgent need to study the impact of emotional experiences on patient care in real emergency department contexts.

What is already known

• Engaging with human emotions is an integral but poorly understood part of the work of emergency healthcare providers.

• As members of the diagnostic team and key communicators of patient information, emergency nurses embody a critical point of care.

• Patient safety and quality of care require that patients receive timely and appropriate assessment and management of health conditions, but little is known about the extent to which nurses’ emotions may influence these outcomes.

What this paper adds

• Irritable (vs. calm) patient behaviors elicited anger in emergency nurses and adversely influenced their patient assessments, beliefs about a patient’s prognosis, and handoff quality.

• Patients with mental illness (vs. without) were less likely to have necessary tests recommended for them by emergency nurses, which may indicate diagnostic overshadowing.

Data availability

The data for this manuscript are available from the authors on request.