Comparison of usual care and the HEART score for effectively and safely discharging patients with low‐risk chest pain in the emergency department: would the score always help?

Abstract Background Triage decisions for chest pain patients receiving usual care are based on a dynamic and comprehensive strategy performed in the physician's mind. It remains controversial whether simple, structured risk tools can surpass real, complex judgments. Hypothesis The potentially used History, Electrocardiogram, Age, Risk factors, Troponin (HEART) score would help identify low‐risk patients for discharge. Methods Patients with acute, non‐traumatic chest pain managed according to usual care were consecutively enrolled in a tertiary university hospital in China from August 24, 2015 to September 30, 2017. Major adverse cardiac events (MACE) included death, acute myocardial infarction, revascularization, and significant coronary stenosis (>50%) within 30 days. We compared the efficacy and safety of usual care and the potentially used HEART score in this population. Results Of 2185 patients analyzed, 926 (42.4%) patients were directly discharged by usual care, whereas HEART≤3 would have identified 524 (24.0%) patients as low‐risk (P < .001). The MACE rate in discharged patients was 2.2% (20/926) and would have been 5.2% (27/524) in those with HEART≤3 (P = .002). For discharged patients, the MACE rates in HEART≤3 vs HEART>3 groups were not significantly different (1.5% vs 2.7%, P = .225). Negative predictive value (NPV) was higher with usual care than with the HEART score (P = .003), but sensitivity was similar. For 340 patients with serial troponins, usual care was superior to the potentially used HEART score in regard to efficacy. Conclusions At this institution, usual care identified many more patients for discharge than the HEART score would have without apparently different outcomes in discharged patients with lower vs higher HEART scores. The HEART score would not appear to provide helpful risk stratification.

with the HEART score (P = .003), but sensitivity was similar. For 340 patients with serial troponins, usual care was superior to the potentially used HEART score in regard to efficacy.
Conclusions: At this institution, usual care identified many more patients for discharge than the HEART score would have without apparently different outcomes in discharged patients with lower vs higher HEART scores. The HEART score would not appear to provide helpful risk stratification.
chest pain, discharge, emergency department, HEART, usual care

| INTRODUCTION
Acute chest pain is one of the most common reasons for emergency department (ED) evaluation. Only a small proportion of patients receive a final diagnosis of acute coronary syndrome (ACS). 1,2 Reliably detecting patients with ACS remains a diagnostic dilemma. 3 Inappropriate admission of patients with benign disease is neither indicated nor cost-effective, whereas inadvertent discharge of patients with ACS from the ED is associated with increased mortality and liability. 4,5 Previous studies have shown that between 2.1% and 4.6% of patients with acute myocardial infarction (AMI) in the ED are mistakenly discharged. 4,6,7 Montassier et al found that 3.7% of patients were mistakenly discharged and presented with major adverse cardiac events (MACE) within 60 days. 8 The accurate and safe identification of patients who can be directly discharged from the ED is a challenge for physicians.
Current guidelines recommend the use of structured risk stratification tools to evaluate and triage patients with suspected ACS presenting to the ED. [9][10][11] Previous studies have indicated that the performance of the History, Electrocardiogram (ECG), Age, Risk factors, Troponin (HEART) score seems superior to other risk prediction scores. [12][13][14][15][16] It has been recommended that patients with a HEART score ≤3 should be discharged without further diagnostic testing, including no second cardiac troponin (cTn) measurement. 17,18 However, triage decisions for chest pain patients using usual care are based on a comprehensive strategy performed in the physician's mind. After all, dynamic variations can appear in every aspect of a patient's clinical condition, including symptoms, signs, ECG, markers, intentions to receive care, and others. It takes time to observe these changes, but currently used risk stratification scores merely focus on one cross-section of the timeline during an ED stay. [18][19][20] To solve this problem, a triage pathway, such as the HEART pathway, based on a clinical score and two serial cTn tests have been developed and evaluated. 21,22 Only the change in cTn over time is considered. Therefore, it remains controversial whether simple, structured risk tools can surpass real, complex judgments. 15,17,23 The performance of disposition decision protocols to discharge low-risk patients mainly depends on efficacy (defined as the proportion of patients discharged) and safety (quantified by the sensitivity and negative predictive value [NPV] for MACE). 24 In this study, we aimed to compare the efficacy and safety of usual care with the potentially used HEART score and pathway to identify patients for direct discharge in a tertiary hospital.

| Study population
This study was a retrospective analysis of a prospective observational cohort. Patients were consecutively recruited from the ED of Qilu Hospital of Shandong University, a tertiary university hospital that has 24 hours access to interventional angiography, between August 24, 2015 and September 30, 2017. All adult patients with acute, nontraumatic chest pain or other symptoms suggestive of ACS occurring in the previous 24 hours and for whom the attending physician requested cTn for suspected ACS were included. Other symptoms suggestive of ACS may include shortness of breath, nausea, vomiting, jaw pain, and others. 25 Patients were excluded if they were unwilling to provide informed consent, had an initial impression of ST-segment elevation myocardial infarction (STEMI), died in the ED, were transferred to other hospitals or left against medical advice.
This study was approved by the ethics committee of Qilu Hospital of Shandong University, and all patients provided written informed consent.

| Data collection
Clinical data were collected on a standardized case report form (CRF) in accordance with clinical data standards by trained research assistants. 25 All the elements of the HEART score were included. The disposition after ED evaluation was categorized into discharged and undischarged. Undischarged included hospitalization and referral to a cardiologist. Hospitalization was defined as admission to an inpatient unit or to an observation room in the ED for at least 24 hours. 25 Follow-up after 30 days was conducted by trained research assistants to acquire information about adverse events, hospital revisits, and readmission over the telephone. The relevant medical records were obtained if a hospital admission was reported during the follow-up period. Local death registry data were checked to ensure whether patients lost to contact were deceased.

| Decision strategies
Every subject in this study was managed by usual care (discharged or undischarged) and retrospectively evaluated to be low or high risk according to the potentially used HEART score (≤3 or > 3). The methods for calculating the HEART score and HEART pathway have been previously described. 21 The HEART score consisted of five elements: history, ECG, age, risk factors, and troponin. The history component was scored using a list of predefined chest pain characteristics that were categorized as typical or atypical. The ECG component was scored based on the impression of the first ECG by treating physicians. The first cTn results and the 99th percentile of the upper reference limit (URL) were used to calculate the scores. The overall HEART scores were retrospectively determined by the SAS program to guarantee their veracity and consistency. Patients with a HEART score of 0 to 3 were categorized as low-risk of developing MACE and considered eligible for direct discharge from the ED without further diagnostic testing. The HEART pathway (−) indicated patients were low risk if their HEART score was ≤3 and two cTn tests were both negative (the first and second cTn values after presentation). Since the use of the HEART score would have indicated discharge immediately after the low-risk score was assigned, the time of the initial cTn report (completing the HEART score) was taken as the discharge time in the HEART score group.

| Clinical outcome
The primary outcome was MACE within 30 days after initial presentation, including death from all causes, index (being the cause for the initial presentation) or subsequent (occurring during the follow-up) AMI, revascularization (emergency/urgent/elective percutaneous coronary intervention [PCI] or coronary artery bypass grafting [CABG]), and coronary angiography revealing significant stenosis (>50%) with conservative treatment. The secondary outcome was the composite of death from all causes, index or subsequent AMI and emergency revascularization within 30 days. Each event in MACE was adjudicated by two independent cardiologists for all patients in accordance with the definitions following a review of all available medical records. [25][26][27][28] In case of disagreement, a third cardiologist reviewed the record. AMI referred to a type 1 myocardial infarction and was defined as myocardial necrosis in the context of myocardial ischemia due to a definite or highly suspected plaque rupture and coronary thrombosis. Myocardial necrosis was diagnosed based on the rise and/or fall (a delta of ≥20% was used) of cTn with at least one value above the diagnostic threshold. 10,28

| Statistical analysis
Continuous variables are presented as the mean ± SDs or median (interquartile range), and categorical variables are summarized as numbers and percentages. The outcome rates were compared with a χ 2 test between groups. Diagnostic accuracy with a 95% confidence interval (CI) for usual care and the potentially used HEART score or pathway for MACE were determined, including sensitivity, specificity, NPV, and positive predictive value (PPV). The efficacy (proportion of patients identified as low risk), sensitivity, and specificity of the different strategies were compared using the McNemar test based on paired 4-fold tables. 29 The NPV and PPV of the different strategies were compared using χ 2 tests for the respective proportions. A Pvalue less than .05 (two-sided significance testing) was considered statistically significant. All statistical analyses were performed using SAS V.9.4 (SAS Institute, Inc., Cary, North Carolina).  Table 1.

| Efficacy of usual care vs the HEART score
Usual care triaged 926 (42.4%) patients to be directly discharged without further testing. If the HEART score was used in the entire cohort, 524 (24.0%) patients would have been identified as low risk (Figure 1).
Based on the paired 4-fold table stratifying chest pain by usual care vs the potentially used HEART score (Table S1), the difference between these two percentages was significant (P < .001). The specificity of usual care to rule in events was 0.577 (0.553, 0.602), which was superior to the HEART score with 0.317 (0.294, 0.340) (P < .001) ( Table 2).
The baseline characteristics of the low-vs high-risk groups categorized by usual care or the potentially used HEART score are shown in Table S2, and the median LOS of patients discharged by usual care was 5.5 (1.7, 8.7) hours, which was longer than the assumed time using a HEART score ≤ 3, 1.5 hours (1.4,1.7). For the composite of death, AMI and emergency revascularization, the specificity of usual care outperformed the HEART score (Table S3).

| Outcomes
A total of 615 (28.2%) patients had 30-day MACE in this chest pain cohort ( Figure 1). As shown in Table 3, the MACE rate in patients discharged by usual care was 2.2% (20/926), and the rate would have been 5.2% (27/524) in those with a low HEART score (P = .002). For patients deemed to be low risk by usual care (discharged), the MACE rates in the HEART score ≤ 3 vs HEART score > 3 groups were not significantly different (1.5% vs 2.7%, P = .225). For patients with a low HEART score, the MACE rate in the discharged group was much lower than that in the undischarged group (1.5% vs 17.1%, P < .001).
The incidence of the composite of death, AMI, and emergency  revascularization was similar in patients discharged by usual care and in patients with the potentially used HEART score ≤ 3 (0.5% vs 1.5%, P = .079) ( Table 3).

| Safety of usual care vs the HEART score
The sensitivity of usual care was 0.967 (0.953, 0.981) and would have been 0.956 (0.940, 0.972) for the potentially used HEART score (P = .311). The difference in NPV was significant (P = .003) between usual care at 0.978 (0.969, 0.988) and the HEART score at 0.948 (0.930, 0.967) ( Table 2). For the composite of death, AMI, and emergency revascularization, the sensitivity and NPV showed no significant differences between the two strategies (Table S3).

| Performance of usual care vs the HEART pathway
As shown in Table 4, among 340 patients who received serial cTn tests, the proportion of low-risk chest pain identified by usual care was 25.9% (88/340), whereas use of the HEART pathway would have noted only 11.5% (39/340) of patients as low risk (P < .001). The MACE rate in discharged patients was similar to that would have been in the HEART pathway low-risk group (3.4% vs 7.7%, P = .370) ( Table 4). For discharged patients, the difference in the MACE rates between the HEART pathway (−) and HEART pathway (+) groups was not significant (0% vs 4.7%, P = .559). For patients with a negative HEART pathway, the MACE rate in the discharged group was lower than that in the undischarged group (0% vs 20.0%, P < .050) (   safety of usual care and the potentially used HEART strategies to discharge low-risk patients were compared. In this Chinese tertiary hospital, usual care identified a much larger proportion of patients to directly discharge needing no further investigations than the HEART score would have. The rate of MACE in discharged patients was lower than that would have been in those with a low HEART score (≤3). For discharged patients, the high HEART score (>3) group showed no apparent difference in outcomes compared with the low HEART score (≤3) group. On the contrary, for patients with a low HEART score, those deemed low risk by usual care (discharged) did have lower MACE rate than high-risk (undischarged) patients.
Previous studies have demonstrated that clinician gestalt alone is insufficient or not superior for identifying patients who are safe to discharge from the ED. 15,23,30,31 However, the unstructured assessment used in these studies is generally a scale or an impression code of ACS probability completed based on history, symptoms, and physical examination performed by physicians. It has been shown that medical history, risk factors, symptoms, and physical examination do not have a sufficient discriminatory ability to "rule-in" or "rule-out" ACS in the ED. 32  and NPV of 99.5%. This level of safety would be acceptable (>99%) to most physicians according to an international survey. 35 The specificity of usual care still outperformed the HEART score with a similar level of safety. However, safe discharge is not only the avoidance of acute critical events (like AMI or death) but also potential events to which coronary stenosis may lead. Therefore, it is more appropriate to cover all these events to evaluate the performance of strategies to directly discharge chest pain patients without further diagnostic testing. 17 A metaanalysis demonstrated that a HEART score of 0 to 3 could miss 1.6% MACE (range 0.9%-5.9%) with a pooled sensitivity and specificity of 96.7% (range 75.5%-100%) and 47.0% (range 31.8%-67.5%), respectively. 36 In our analysis, the incidence of MACE in patients with a low HEART score would have been much higher (5.2%) with lower sensitivity and specificity than the pooled values mentioned above. The possible reasons for these disparities may be the different definition of MACE and ACS prevalence. 32,37 After all, the percent of patients suffering 30-day MACE in our study was 28.2%, which was higher than that in most prior research. 36 There were several limitations in the study. First, this study was a single-center observational study, which may limit the generalizability of the findings. The performance of usual care and clinical scores in broader patient populations should be determined by further studies of heterogeneous groups. In particular, the findings need to be confirmed in a randomized trial. Our report may provide implications for improving clinical risk stratification tools, of which objectivity and repeatability may overcome the variability between physicians and between hospitals. Second, we did not invite the discharged patients to take cTn tests and ECGs in a follow-up assessment. Therefore, it was possible that the incidence of MACE was underestimated. However, to reduce this underestimation, we reviewed all relevant medical records to determine the occurrence of any outcome if patients revisited a hospital during the follow-up period.