Shared Decision-making for Syncope in the Emergency Department: A Randomized Controlled Feasibility Trial.

Abstract

Objective:

Significant practice variation is seen in the management of syncope in the emergency department (ED). We sought to evaluate the feasibility of performing a randomized controlled trial of a shared decision-making (SDM) tool for low-to-intermediate risk syncope patients presenting to the ED.

Methods:

We performed a randomized controlled trial of adults (≥30 years) with unexplained syncope who presented to an academic ED in the United States. Patients with a serious diagnosis identified in the ED were excluded. Patients were randomized, 1:1, to receive either usual care or a personalized syncope decision aid (SynDA) meant to facilitate SDM. Our primary outcome was feasibility, i.e. ability to enroll 50 patients in 24 months. Secondary outcomes included patient knowledge, involvement (measured with OPTION-5), and rating of care; and clinical outcomes at 30 days post-ED visit.

Results:

After screening 351 patients, we enrolled 50 participants with unexplained syncope from January 2017 to January 2019. The most common reason for exclusion was lack of clinical equipoise to justify SDM (n=124). Patients in the SynDA arm tended to have greater patient involvement, as shown by higher OPTION-5 scores: 52/100 vs 27/100 (between group difference −25.4; 95% CI −13.5 to −37.3). Both groups had similar levels of clinical knowledge, ratings of care, and serious clinical outcomes at 30 days.

Conclusions:

Among ED patients with unexplained syncope, a randomized controlled trial of a shared decision-making tool is feasible. Although this study was not powered to detect differences in clinical outcomes, it demonstrates feasibility, while providing key lessons and effect sizes that could inform the design of future SDM trials.

Introduction

Syncope is a chief complaint often encountered by clinicians working in the Emergency Department (ED), representing over 1.4 million visits per year and rising.^1,2 The healthcare costs and resource utilization associated with syncope care are substantial.^3,4 Even after extensive ED evaluation, involving laboratory, electrocardiogram, and radiologic testing,⁵ the actual cause of the syncope often remains unknown.^6,7 Most causes of syncope are benign.^8,9 Serious causes, including cardiac arrhythmias, acute coronary syndrome, and structural heart disease, are infrequent and difficult to diagnose in the ED.^10,11 As a result, there is substantial variation in ED management⁵ and a tendency to admit intermediate-to-high-risk patients for further testing and monitoring.³ Continued evaluation, either in an observation or inpatient unit are often low-value and may not reflect the values and preferences of the patient.^12,13

Shared decision-making (SDM) has been proposed as an approach to improve care for ED patients,¹⁴ including those with intermediate-risk syncope.^15,16 SDM has been defined as a collaborative process in which patients and providers make health care decisions together, taking into account scientific evidence, the clinician’s experience, as well as the patient’s values and preferences.¹⁷ Although the benefits of SDM have been demonstrated in randomized trials involving patients with low-risk chest pain in the ED, a clinical scenario with some parallels to syncope,^18,19 no studies have directly evaluated SDM for ED syncope care. We recently developed a personalized, paper-based decision aid to facilitate SDM for patients with unexplained syncope after ED evaluation.²⁰ In 2016, the Academic Emergency Medicine (AEM) Consensus Conference on SDM in the ED called for further research on the development and testing of SDM interventions and identified syncope as a relevant clinical scenario.²¹ Priority research questions identified during this conference included how to communicate risk and uncertainty with patients, what tools can facilitate SDM, and what outcomes are suitable for SDM research in the ED.

Given the recent emphasis on patient-centered care in the ED, the frequency with which unexplained syncope is encountered, and the multiple reasonable management options to consider for this condition, it is important to study the effects of SDM for intermediate-risk syncope patients. Potential benefits include increased patient knowledge and satisfaction and decreased low-value health services, while the risks include unanticipated serious clinical outcomes. A decrease in low-value health services could potentially decrease healthcare costs and ED crowding, two important challenges facing the American healthcare system today.^22,23

Our objective was to perform a pilot, randomized clinical trial to test the feasibility and preliminary efficacy of using a paper-based patient decision aid to facilitate SDM around the disposition decision for low- to intermediate-risk syncope patients in a busy ED. We sought to determine if execution of such a study protocol was feasible given the frequency of syncope visits in the ED and our inclusion/exclusion criteria, as well as to estimate the effect size of the intervention on various study outcomes to inform the design of future studies.

Methods

Study Design

We performed a single-center, parallel randomized controlled trial of syncope patients. Patients in the intervention arm were randomized to SDM facilitated by a paper-based, personalized, syncope patient decision aid (SynDA), while those in the control group received usual care. Given the nature of the intervention (decision aid to facilitate SDM), patients and providers were not able to be blinded. However, outcome assessors were blinded to study arm assignment. Full-time, paid research coordinators and a project manager were trained on all aspects of the research protocol, including screening, obtaining consent, randomization and follow-up prior to participating in study activities. This study was approved by the institutional review board at our medical center and registered at ClinicalTrials.gov (NCT02971163). This research was funded by a Career Development Grant from the National Institutes of Health, which had no role in the conduct or reporting of the study. As much as feasible, we adhered to the CONSORT criteria for randomized controlled feasibility trials.²⁴ Post hoc, we evaluated where on the pragmatic-explanatory continuum our trial lay using the PRECIS-2 guidelines.²⁵

Study Setting and Population

This trial was performed at an urban, academic ED in a tertiary care medical center in the U.S. with an annual volume of over 100,000 visits which serves a socioeconomically and ethnically diverse patient population. We enrolled a convenience sample of adult patients (≥30 years) who presented with syncope and were deemed appropriate for SDM by the treating attending physician. Patients were recruited for 12 hours/day Monday through Friday and 8 hours/day on Saturday and Sunday over a 24-month period. The minimum age of inclusion was lowered from 40 to 30 years mid-way through the trial to increase enrollment. There was no upper limit of age for participants. Exclusion criteria included inability to read in English, major communication barrier (e.g. significant visual or auditory impairment, altered mental status), lack of fixed address and telephone number, and a serious diagnosis identified in the ED, as determined by the treating clinician. Clinicians were also able to exclude patients from the trial if they deemed them to be inappropriate for observation care based on their risk profile (e.g. too low risk to be offered observation care, or too high risk to be offered direct discharge from ED), or not appropriate for SDM based clinical variables (e.g. dementia, altered mental status, intoxication, other disease process requiring admission). All clinicians (PAs, resident and attending physicians) were eligible to participate, signed informed consent, and were trained on the use of the decision aid to facilitate SDM during a 30-minute training course prior to the start of the trial.

Trained clinical research coordinators screened the electronic ED trackboard for eligible patients (looking for chief complaints of syncope, loss of consciousness, fall, dizziness, weakness, fainting, or passing out) and approached the treating clinician to verify eligibility before approaching the patients to explain the study and obtain written informed consent. After confirming the chief complaint (i.e. syncope), that the diagnostic evaluation had not revealed a serious diagnosis, and that the treating attending physician deemed the patient appropriate for SDM, (i.e. has medical decision-making capacity and was neither too high nor too low risk), research staff randomized patients to the SynDA arm versus usual care. The was done after all of the results of ED testing had returned, with the exception of a second troponin test, if ordered. In that scenario, patients were informed that their continued eligibility was contingent on the second troponin results being negative. Randomization was achieved using sealed, opaque envelopes assigning patients on a 1:1 basis to SynDA or to usual care, using a variable block sequences of 4 to 6, and was stratified by age (less than or greater than 60 years) to ensure a balance of older patients in both groups. The random allocation sequence was created by a statistician at our institution.

Interventions

Patients were randomized to receive either usual care or a personalized syncope decision aid, named SynDA, designed to facilitate SDM (See Figure 2) around the disposition decision, i.e. direct ED discharge with primary care or cardiology follow-up, versus observation care. This paper-based decision aid was iteratively developed over a 1-year period with broad stakeholder input from patients, emergency physicians, a physician assistant (PA), a healthcare designer, and experts in the field of SDM research. This development process has been published previously.²⁰ Briefly, the tool uses simple language to explain what syncope is, what potential underlying conditions are still possible, what the patient’s risk of 30-day a serious medical event is, and what the possible disposition options are (i.e. direct discharge with primary care provider follow-up, direct discharge with cardiology follow-up, or admit to the observation unit).

Figure 2.

SynDA: Options for care after passing out

Differing versions of the SynDA tool were used based on the patient’s individualized Canadian Syncope Risk Score and the corresponding estimated probability of a serious medical event within 30 days.¹¹ “We chose the Canadian Syncope Risk Score based on the following factors: 1) rigorous developmental methodology, 2) largest sample size of any ED syncope risk score, and 3) stratified outcome with graduated levels of risk.” The components of the Canadian Syncope Risk Score were electronically entered into a database by the research assistant with the guidance of the treating clinician, generating the personalized risk estimate. Versions of the tool differed only in their displayed estimated risk, which ranged from 0.5% to 9%, communicated using percent, natural frequency, and a 100-person, colored pictogram to maximize patient comprehension. If the calculated 30-day risk exceeded the comfort level of the attending physician, he or she had the option to exclude the patient from the study.

Participants in the usual care arm received management as per standard care for ED syncope patients, without any study intervention or decision aid. This may or may not have included some form of SDM with the patient, depending on the usual practice of the clinician involved. ED disposition for all patients was determined by the treating clinician and not by study protocol. This disposition discussion, with or without the SynDA tool, was performed either by the attending physician or the emergency medicine resident caring for the patient after all ED testing had been completed.

Measurements

Research staff recorded: 1) the proportion of eligible syncope patients enrolled in the study, 2) proportion of enrolled patients successfully contacted at 30 days; and 3) proportion of patients deemed inappropriate for SDM by clinicians. Patient demographics, numeracy, health literacy, Control Preferences Scale^26,27, education and income were collected at baseline. The Control Preferences Scale (see Appendix 1) is a validated single question measure of patient’s desire for involvement in medical decisions that ranges from 1 (active) to 3 (collaborative) to 5 (passive).²⁸ Numeracy and health literacy were measured using two validated scales: the Subjective Numeracy Scale (SNS) and the Short Literacy Survey (SLS), respectively. The SNS consists of three questions summed to give a score of 3-18 with higher scores indicating higher subjective numeracy skills.²⁹ The SLS consists of three questions summed to give a score of 3-15 with higher scores indicating greater subjective health literacy.³⁰ The choice of health literacy/numeracy assessments was based on brevity, ease of use, and prior validation in the ED setting. All enrolled participants completed a post-encounter survey after the disposition decision to assess their perceptions about the amount, clarity, and helpfulness of the information shared, their knowledge about syncope, their level of decisional conflict, and their overall rating (0-10) of the ED visit. ED providers (the resident, attending, or PA) also completed a survey assessing their perception of the amount, clarity, and helpfulness of the information shared and, for encounters randomized to the intervention, the acceptability of the tool. Patient and clinician surveys were adapted from prior studies and have demonstrated construct validity (See Appendix 2A and 2B).¹⁹

All participants were offered the option of having the final disposition discussion audio-recorded or directly observed by research staff to evaluate patient engagement via the OPTION-5 scale.³¹ The duration of each discussion was also recorded. The OPTION-5 scale is a brief, 5-item instrument used by a trained observer to measure the degree of patient involvement in clinical decision-making and has demonstrated validity and reliabity.³² The principal investigator (PI) trained two research coordinators on OPTION-5 scoring using an online course from the Preference Laboratory at The Dartmouth Institute and demonstration audio-recordings of patient and clinician actors. The demonstration audio-recordings were created by the PI using two volunteer patient actors not aware of the study hypothesis. Six separate audio-recordings were created to illustrate minimal, moderate, and maximal clinician effort towards patient involvement, with and without the decision aid. The two research coordinators listened to the demonstration recordings, while blinded to the intended effort, with the PI and rated and discussed each one until consensus was achieved. Disagreements were resolved by discussion with the PI. Both coordinators were blinded to study assignment when evaluating the study recordings. Intraclass correlation (ICC 3,1) for OPTION-5 scores on a subsample of eight recordings was calculated to measure inter-rater agreement. ICC is a means to measure the reliability of quantitative measurements made by different observers³³ recommended by the creators of the OPTION-5 score.³² We used a two-way mixed-effects ICC model, where each subject is assessed once by each rater, and the selected raters are the only raters of interest.

To assess 30-day outcomes, research staff completed electronic chart review and contacted patients by telephone to confirm major adverse cardiac events, hospital admissions, ED visits, physician office visits, and further testing since discharge within 30 days of the index visit. Staff collecting this outcome were blinded to study assignment. Patients received one $25 gift card at the time of enrollment, and a second $25 gift card upon completion of the 30-day follow-up telephone survey. All screened patients were reminded that participation in the study was entirely optional and that refusal would not affect the medical care they received.

Key Outcomes Measures

Our primary outcome was feasibility, defined as ability to reach target enrollment: 50 patients in 24 months. This sample size was based on the time and resources available for study completion and is consistent with prior ED feasibility studies.^34–36 Secondary outcomes included patient knowledge, satisfaction, engagement (as measured using the validated OPTION-5 scale), decisional conflict (as measured using the validated Decision Conflict Scale)³⁷, admission rates, and clinical outcomes at 30 days post-index visit. We also measured provider satisfaction with the SynDA tool.

Data Analysis

Descriptive statistics for baseline demographic and clinical variables were conducted. Secondary outcomes with continuous distributions were evaluated using independent sample t-tests and a two-sample test of proportions for dichotomous variables. For continuous variables, the Shapiro-Wilk normality and Anderson-Darling tests were used to determine if the continuous variables were normally distributed. For the continuous data that were not normally distributed, we used the Wilcoxon rank-sum test by groups. Point estimates and 95% confidence intervals (CI) were calculated. Numerical and categorical data were collected with the Research Electronic Data Capture (REDCap) web-based program (Vanderbilt University, Nashville, TN). Audio-recordings of the disposition discussion were collected using a tablet device with an embedded digital microphone. Analyses were conducted using STATA version 15.2 (StataCorp LP, College Station, Texas) with an intention-to-treat approach.

Results

Characteristics of study subjects

After screening 350 patients with syncope, we enrolled 50 (14.3%) from January 2017 to January 2019 (see Figure 1). One patient was excluded after randomization since the etiology of their loss of consciousness was actually hypoglycemia, not syncope. The most common reason for exclusion (n=124, 35.4%) was lack of clinical equipoise to justify SDM (i.e. patient too high/low risk). Among enrolled patients, 58% were female, and the median age was 58.5 years (range 33-89). Thirty-four percent identified as white (Table 1). The two study arms were well-balanced with regard to numeracy, health literacy, and education, but not with regard to race and gender.

Figure 1.

SynDA Trial Patient Flow Diagram

Table 1.

Baseline characteristics of study participants. No. (%)

Characteristic	All, N=50	SynDA group, n=24	Control group, n=26
Sex:
Female	29 (58%)	12 (50%)	17 (65%)
Age in years, median (IQR)	58.5 (46.3, 71.5)	54.5 (45.3, 70)	60 (47.8, 73)
Race/Ethnicity:
Black/African-American	18 (36%)	7 (29%)	11 (42%)
Hispanic/Latino	10 (20%)	5 (21%)	5 (19%)
White	17 (34%)	10 (42%)	7 (27%)
Asian	1 (2%)	0 (0%)	1 (4%)
Other	4 (8%)	2 (8%)	2 (8%)
Control Preferences Scale:
1	3 (6%)	3 (13%)	0 (0%)
2	15 (30%)	8 (33%)	7 (27%)
3	23 (46%)	10 (42%)	13 (50%)
4	4 (8%)	2 (8%)	2 (8%)
5	2 (4%)	0 (0%)	2 (8%)
Numeracy score, mean (SD)	12.9 (2.6)	13.2 (2.6)	12.6 (2.6)
Health Literacy score, mean (SD)	5.3 (2.7)	5.0 (2.6)	5.5 (2.8)
Education:
Some High School	2 (4%)	1 (4%)	1 (4%)
High School Graduate/GED	7 (14%)	3 (13%)	4 (15%)
Some College	15 (30%)	8 (33%)	7 (27%)
College Graduate	10 (20%)	4 (17%)	6 (23%)
Graduate School	16 (32%)	8 (33%)	8 (31%)
Income:
Less than $20,000	9 (18%)	2 (8%)	7 (27%)
$20,000 to $29,999	5 (10%)	1 (4%)	4 (15%)
$30,000 to $39,999	4 (8%)	2 (8%)	2 (8%)
$40,000 to $59,999	5 (10%)	3 (13%)	2 (8%)
$60,000 to $79,999	3 (6%)	3 (13%)	0 (0%)
$80,000 to $99,999	3 (6%)	2 (8%)	1 (4%)
$100,000 or more	9 (18%)	4 (17%	5 (19%)
Rather Not Say	12 (24%)	7 (29%)	5 (19%)
Canadian Syncope Risk Score:
Low: 0 (≈2% risk)	9 (18%)	3 (13%)	6 (23%)
Medium: 1 (≈3% risk)	28 (56%)	13 (54%)	15 (58%)
Medium: 2 (≈5% risk)	13 (26%)	8 (33%)	5 (19%)

IQR: Interquartile range; SD: Standard Deviation; GED: General Educational Development;

Main results

We were able to randomize 50 syncope patients from the ED, demonstrating feasibility. We stopped the trial after reaching our minimum target enrollment, n=50. Our telephone follow-up rate at 30 days was 88% (44/50). Patients and providers both expressed satisfaction with respect to amount, clarity, and helpfulness of the information shared (see Table 2). Overall rating of care by patients was similar in the two groups, 7.6/10 in the control group vs 8.3/10 in the intervention group (between group difference −0.7; 95% CI −1.9 to 0.5). Patient knowledge was also similar in the two groups: 5.3/10 vs 5.2/10 in the control vs intervention group (between group difference 0.1; 95% CI −0.9, 1.1). Patient engagement tended to be higher in the SynDA arm as reflected by a higher mean OPTION-5 scores: 52/100 vs 27/100 (between group difference −25.4; 95% CI −13.5 to −37.3), as shown in Table 3. OPTION-5 scores were found to be normally distributed using the Anderson-Darling test. Interrater agreement for OPTION-5 scores was found to be moderately high (ICC=0.73) on the subsample of audio recordings.³⁸ The median duration (not normally distributed) for each the disposition discussion was 3:29 minutes in the control arm and 5:26 minutes in the SynDA arm (Wilcoxon two-sample test, p=0.003).

Table 2.

Outcome comparisons between participants randomized to SynDA tool vs. Control.

Outcome Variable	SynDA Group n=24		Control Group n=26		Difference between groups (95%CI)
Patient Reported Outcomes (Index visit)
Amount of Information: n (%)
1= Too little information	3	(12.5%)	1	(3.8%)	−8.7% (−23.8, 6.5)
2=	0	(0.0%)	2	(7.7%)	7.7% (−2.6, 17.9)
3= Just Right*	15	(62.5%)	21	(80.8%)	18.3% (−6.3, 43)
4=	5	(20.8%)	1	(3.8%)	−17.0% (−34.8, 0.9)
5= Too much information	1	(4.2%)	0	(0.0%)	−4.2% (−12.2, 3.8)
Clarity of Information: n (%)
1= Not clear at all	1	(4.2%)	1	(3.8%)	−0.3% (−11.2, 10.6)
2	1	(4.2%)	1	(3.8%)	−0.3% (−11.2, 10.6)
3= Somewhat Clear	4	(16.7%)	4	(15.4%)	−1.3% (−21.6, 19.1)
4	8	(33.3%)	11	(42.3%)	9.0% (−17.8, 35.7)
5= Extremely Clear*	10	(41.7%)	9	(34.6%)	−7.1% (−33.9, 19.8)
Helpfulness of Information: n (%)
1= Not helpful at all	0	(0%)	2	(7.7%)	7.7% (−2.6, 17.9)
2=	0	(0%)	1	(3.8%)	3.8% (−3.5, 11.2)
3= Somewhat helpful	6	(25.0%)	5	(19.2%)	−5.8% (−28.8, 17.2)
4=	5	(20.8%)	8	(30.8%)	9.9% (−14.1, 34.0)
5= Extremely helpful*	13	(54.2%)	10	(38.5%)	−15.7% (−43.0, 11.6)
Decisional Conflict Score (range = 0-10)	8.9		9		0.1 (−0.8, 1.0)
Patient Knowledge, mean (range = 0–9)	5.2		5.3		0.1 (−0.9, 1.1)
Recommend the way of sharing information: n (%)
1= Strongly recommend*	19	(79.2)	13	(50.0%)	−29.2% (−54.3, −4.0)
2=	4	(16.7)	5	(19.2%)	2.6% (−18.7, 23.8)
3= Not sure	0	(0%)	2	(7.7%)	7.7% (−2.6, 17.9)
4=	1	(4.2%)	2	(7.7%)	3.5% (−9.5, 16.5)
5= Strongly recommend against	0	(0%)	3	(11.5%)	11.5% (−0.7, 23.8)
Refused	0	(0%)	1	(3.8%)	3.8% (−3.5, 11.2)
Overall Rating of Care, mean (range = 0-10)	8.3		7.6		−0.7 (−1.9, 0.5)
Likelihood to recommend this ER: n, (%)
-Definitely No	1 (4.2%)		1 (3.8%)		−0.3% (−11.2, 10.6)
-Probably No	1 (4.2%)		1 (3.8%)		−0.3% (−11.2, 10.6)
-Probably Yes	8 (33.3%)		8 (30.8%)		−2.6% (−28.5, 23.3)
-Definitely Yes*	14 (58.3%)		15 (57.7%)		−0.6% (−28.0, 26.7)
Provider Reported Outcomes (Index visit)
Amount of Information: n (%)
1= Too little information	1	(3.8%)	0	(0.0%)	3.8% (−3.5, 11.2)
2=	3	(11.5%)	1	(4.2%)	7.4% (−7.3, 22.0)
3= Just Right*	19	(73.1%)	18	(75.0%)	−1.9% (−26.2, 22.4)
4=	3	(11.5%)	5	(20.8%)	−9.3% (−29.7, 11.1)
5= Too much information	0	(0.0%)	0	(0.0%)	0.0%
Clarity of Information: n (%)
1= Not clear at all	0	(0.0%)	0	(0.0%)	0.0%
2	1	(4.2%)	1	(3.8%)	−0.3% (−11.2, 10.6)
3= Somewhat Clear	0	(0.0%))	11	(42.3%)	42.3% (23.3, 61.3)
4	4	(16.7%)	9	(34.6%)	17.9% (−5.6, 41.5)
5= Extremely Clear*	1	(4.2%)	5	(19.2%)	15.1% (−2.1, 32.2)
Helpfulness of Information: n (%)
1= Not helpful at all	0	(0.0%)	0	(0.0%)	0%
2=	2	(8.3%)	0	(0.0%)	−8.3% (−19.4, 2.7)
3= Somewhat helpful	5	(20.8%)	16	(61.5%)	40.7% (15.9, 65.5)
4=	10	(41.7%)	7	(26.9%)	−14.7% (−40.8, 11.3)
5= Extremely helpful*	7	(29.2%)	3	(11.5%)	−17.6% (−39.6, 4.3)
Would you want to present information about other chief complaints/clinical decisions in the same way? ¶ n (%)			N/A		N/A
1= Yes, for sure*	8	(33.3%)
2=	8	(33.3%)
3= Not sure	5	(20.8%)
4=	3	(12.5%)
5= No, not at all	0	(0.0%)
Would you recommend to other providers the way that you presented information today? ¶ n (%)			N/A		N/A
1= Yes, strongly recommend*	13	(54.2%)
2=	8	(33.3%)
3= Not sure	2	(8.3%)
4=	1	(4.1%)
5= No, strongly recommend against	0	(0.0%)

^*Desired end of the scale.

^¶SynDA group only.

N/A: Not applicable. ER: Emergency Room;

Table 3.

Utilization and clinical outcomes of participants randomized to SynDA tool vs. Control.

Outcome Variable	SynDA Group n=24	Control Group n=26	Difference between groups (95%CI)
Utilization Outcomes
Index Visit
Admitted to hospital	0 (0%)	0 (0%)	0%
Sent to observation unit	4 (17%)	7 (27%)	10.3% (−13, 32)
Discharged from ED	20 (83%)	19 (73%)	−10.3% (−32, 13)
CT scan	1 (4.2%)	1 (3.8%)	−0.3% (−1.6, 1.5)
30-day follow-up period	n=20	n=24
Repeat ED visit	1 (5.0%)	2 (8.3%)	3.3% (−16, 21)
Office visit	14 (70%)	17 (71%)	0.8% (−24, 27)
Hospital admission	1 (5.0%)	1 (4.2%)	−0.8% (−20, 16)
Clinical Outcomes at 30 days
New significant diagnosis	2 (10%)	2 (8.4%)	−1.7% (−23, 17)
-Acute stroke	1 (5.0%)	0 (0%)	−5% (−24, 9.4)
-Heart failure	1 (5.0%)	0 (0%)	−5% (−24, 9.4)
-Cardiac arrhythmia	0 (0%)	1 (4.2%)	4.2% (−12, 20)
-Large pericardial effusion	0 (0%)	1 (4.2%)	4.2% (−12, 20)
Mortality	0 (0%)	0 (0%)	0%
Patient Involvement (Index Visit)	n=22	n=22
OPTION-5 Score, mean (SD) Range: 0-100 (Audio recording)	52.0 (SD=18)	26.6 (SD=21)	−25.4 (−13.5, −37.3)
Duration of discussion, median (IQR) minutes:seconds	5:26	3:29	Wilcoxon two-sample test, p=0.003

ED: Emergency Department; CT: Computed tomography; OPTION-5: Observing Patient Involvement in Decision-Making; SD: Standard Deviation; IQR: Interquartile Range.

Clinical Outcomes

Based on data from the telephone follow-up and the electronic chart review, the rate of new clinically important diagnoses at 30 days 2/20 (10%) in the SynDA group and 2/24 (8.4%) in the control group (See Table 3). The two clinically important diagnoses in the SynDA group were an acute stroke and heart failure, while the two in the control group were a cardiac arrhythmia and a large pericardial effusion. There were no deaths in either group.

Resource Utilization

Resource utilization was similar in the two groups with no patients in either group being admitted to an in-patient service during the index visit. The rate of observation unit care was 26.9% in the control group and 16.7% in the SynDA group (Risk Difference: 10.2%, 95% CI: −12.4%, 32.9%). One patient in each group was admitted to the hospital during the 30-day follow-up period - one for acute stroke and one for a large pericardial effusion.

Discussion

Our single-center, randomized controlled pilot trial evaluating a decision aid for patients with unexplained syncope demonstrated that such a study is feasible in the ED. Our study directly address research gaps identified during the 2016 AEM Consensus Conference on “Shared Decision Making”, including how best to communicate risk and uncertainty with patients, what tools can facilitate SDM, and what outcomes can and should be measured in emergency care SDM research. Although our trial was not powered to detect differences in clinical, utilization, or patient-oriented outcomes, our results suggest a tendency towards increased patient engagement, lower resource utilization, and no difference in clinical outcomes with use of the SynDA tool. The relatively low enrollment rate of 2-3 patients per month, even after age range expansion, does speak to the substantial effort that is needed to complete SDM trials in a busy, academic ED with medically complex patients. Since the clinicians were the ones asked to perform SDM with the patients, we needed their approval and were not able to mandate inclusion in the trial using standardized criteria. Our results provide an estimate of the effect size of the intervention on various study outcomes, and can be used to inform power calculations for future studies. The primary outcome of interest in a future trial of this nature would be difference in observation unit rates across the two study arms. A minimally clinically important difference would be approximately 5%. Key secondary outcomes would include differences in OPTION-5, patient knowledge, and patient satisfaction, along with clinical safety outcomes. We believe a MCID for OPTION-5, patient knowledge, and patient satisfaction scores to be 15/100, 1.5/10, and 1/10, respectively. To recruit an adequate sample size for a definitive trial, a significant investment of time and resources with multiple recruitment centers would be needed, preferably with some of these being high-volume community EDs caring for less medically complex patients.

The proportion of patients sent to the observation unit was lower in the SynDA arm by roughly 10%, corresponding to a 10% higher discharge rate, with no difference in hospitalizations. This suggests that, when offered the option, ED patients may be more likely to choose to be discharged home. This is consistent with prior studies showing that ED patients often have greater risk tolerance than expected after being fully informed of the risks,^39,40 and consistent with systematic review suggesting that SDM tools can reduce resource utilization in non-ED settings.⁴¹ This lower observation rate is also consistent with two prior randomized trials of shared decision-making tools for use with patients presenting with low-risk chest pain.^18,19 Hess et al. reported that the Chest Pain Choice tool was safe and effective for use in this clinical scenario. Similar to our study, the OPTION-5 scores were higher in the intervention arm in these two trials. Our study adds to the growing body of literature demonstrating that shared decision-making is feasible in emergency care, and may decrease resource utilization.^14,42–45

Patients and providers appeared to be generally satisfied with use of the decision aid to facilitate shared decision-making. Patients in both groups tended to offer high ratings with respect to the amount, clarity, and helpfulness of information shared. On average, use of the decision aid was associated with an additional 2 minutes of discussion time; all potential benefits must be weighed against this increased time. Surprisingly, we did not observe an increase in patient knowledge, as was seen in the two chest pain trials. There also did not appear to be a tendency toward greater patient satisfaction in the SynDA group, a result we had not anticipated. Three possible explanations for this are, 1) selection bias; patients who were generally pleasant and agreeable may have been more likely to participate in the trial, 2) the Hawthorne effect may have caused providers to be particularly kind and attentive when discussing disposition options at the end of the visit, and 3) subjects enrolled in randomized controlled trials often get increased attention from both study staff and clinicians, which may increase satisfaction.

Our enrollment rate of 14%, while relatively low, was roughly as expected given the multiple reasons for exclusion including the patient being too low risk for consideration for the observation unit, and conversely, being so high risk as to merit admission. SDM in this scenario is only appropriate when two disposition options are medically reasonable and the patient has decision-making capactity.¹⁴ Nonetheless, there is still a substantial proportion of patients for whom direct discharge or observation are both reasonable, and for these, shared decision-making is most likely to have the greatest benefit. Of note, the adverse event rates of 10% and 8% were similar in the intervention and control groups, which confirms that the enrolled subjects were of intermediate risk and suggests that the decision aid is safe for use in the ED. These rates were as expected given our exclusion of low-risk patients. The two clinically important diagnoses in each arm were felt to be potentially related to the initial presentation of syncope by the study team.

Our study may be helpful to other researchers interested in studying SDM in other ED clinical scenarios, such as transient ischemic attack or minor stroke, imaging for possible appendicitis or acute flank pain, and testing for possible subarachnoid hemorrhage, among others. Researchers would be wise to invest early in training clinicians in SDM, and involve a broad coalition of stakeholders including residents, faculty, physician assistants, and nurses. Financial incentives for providers (as well as patients) may be helpful in increasing enrolment rate, if approved by the IRB. To illustrate where our trial lies on the pragmatic-explanatory continuum, we have included a PRECIS-2 Table and Figure in Appendix 3 and 4, respectively.²⁵

Limitations

There are a number of important limitations to our study. Our small sample size precluded any definitive statistical comparisons between patients in each study arm. We included only English-speaking patients from a single, busy, urban, academic ED; our results may not apply to other patient populations. Patients and providers were not blinded to the intervention, which may have biased the results. As mentioned above, blinding for a SDM intervention is not feasible. All patient encounters were either directly observed or audio-recorded by research staff, which created the potential for a Hawthorne effect. However, since this direct observation occurred in both study arms, it should not have led to differences between the two. Our enrollment rate of roughly 14% has the potential for selection bias and limits generalizability, which was anticipated given that SDM is only appropriate for select ED patients under certain clinical circumstances. Due to random chance and our small sample, there were more women and more African-Americans in the control arm; it’s unlikely the imbalance had an effect on our results. Since clinicians in the control arm were not prohibited from using SDM, it is possible that some contamination occurred. However, such contamination would likely bias results towards the null hypothesis. In addition, 12% of patients were lost to follow-up at 30 days (four in the SynDA arm and two in the control arm), which is lower than expected compared to studies of this duration; it is unlikely that higher follow-up rates would change the statistical significance of our findings.

Conclusions

In summary, our study demonstrates that a randomized controlled trial of a SDM intervention for unexplained syncope patients is feasible in the ED. Although our results are promising with regard to improved patient engagement and resource utilization, a larger randomized trial would be necessary to determine the safety and efficacy of such an intervention. Our study provides key lessons with respect to feasibility, enrollment strategies, and effect sizes that could inform the design of future SDM trials conducted in the ED.