CARDIOLOGY/ORIGINAL RESEARCH

Development of a Clinical Prediction Rule for 30-Day Cardiac Events in Emergency Department Patients With Chest Pain and Possible Acute Coronary Syndrome
Erik P. Hess, MD, MSc, Robert J. Brison, MD, Jeffrey J. Perry, MD, MSc, Lisa A. Calder, MD, MSc, Venkatesh Thiruganasambandamoorthy, MD, MSc, Dipti Agarwal, MBBS, Annie T. Sadosty, MD, Marco L. A. Silvilotti, MD, MSc, Allan S. Jaffe, MD, Victor M. Montori, MD, MSc, George A. Wells, PhD, Ian G. Stiell, MD, MSc
From the Department of Emergency Medicine, Division of Emergency Medicine Research (Hess, Agarwal, Sadosty) and the Department of Internal Medicine, Division of Cardiology (Jaffe) and Division of Endocrinology and Nutrition (Montori), Mayo Clinic College of Medicine, Rochester, MN; the Knowledge and Evaluation Research Unit, Rochester, MN (Hess, Montori); the Department of Emergency Medicine, Queens University, Kingston, Ontario, Canada (Brison, Silvilotti); and the Department of Emergency Medicine (Perry, Calder, Thiruganasambandamoorthy, Stiell) and Department of Epidemiology and Community Medicine (Perry, Calder, Thiruganasambandamoorthy, Wells, Stiell), University of Ottawa, Ottawa, Ontario, Canada.

Study objective: Evaluation of emergency department (ED) patients with chest pain who are at low risk for acute coronary syndrome is resource intensive and may lead to false-positive test results and unnecessary downstream procedures. We seek to identify patients at low short-term risk for a cardiac event for whom additional ED investigations might be unnecessary. Methods: We prospectively enrolled patients older than 24 years and with a primary complaint of chest pain from 3 academic EDs. Physicians completed standardized data collection forms before diagnostic testing. The primary adjudicated outcome was acute myocardial infarction, revascularization, or death of cardiac or unknown cause within 30 days. We used recursive partitioning to derive the rule and validated the model with 5,000 bootstrap replications. Results: Of 2,718 patients enrolled, 336 (12%) experienced a cardiac event within 30 days (6% acute myocardial infarction, 10% revascularization, 0.2% death). We developed a rule consisting of the absence of 5 predictors: ischemic ECG changes not known to be old, history of coronary artery disease, pain typical for acute coronary syndrome, initial or 6-hour troponin level greater than the 99th percentile, and age greater than 50 years. Patients aged 40 years or younger required only a single troponin evaluation. The rule was 100% sensitive (95% condence interval 97.2% to 100.0%) and 20.9% specic (95% condence interval 16.9% to 24.9%) for a cardiac event within 30 days. Conclusion: This clinical prediction rule identies ED chest pain patients at very low risk for a cardiac event who may be suitable for discharge. A prospective multicenter study is needed to validate the rule and determine its effect on practice. [Ann Emerg Med. 2012;59:115-125.] Please see page 116 for the Editors Capsule Summary of this article.
Provide feedback on this article at the journals Web site, www.annemergmed.com. A podcast for this article is available at www.annemergmed.com. 0196-0644/$-see front matter Copyright 2011 by the American College of Emergency Physicians. doi:10.1016/j.annemergmed.2011.07.026

INTRODUCTION
Background Chest pain is the second most common presenting complaint among adults presenting to emergency departments (EDs) across the United States, accounting for more than 6 million patient visits annually.1 Contemporary data suggest that 2.1% of patients with acute myocardial infarction receive misdiagnoses, with a reported 30-day mortality of 9.1%.2 As a
Volume , . : February

result, physicians initiate formal diagnostic testing for acute coronary syndrome at a very low risk threshold.3 Large numbers of low-risk patients undergo prolonged evaluations that result in false-positive test results, unnecessary downstream procedures, and increased health care costs. Importance A number of clinical prediction rules have been developed that identify patients at low risk for acute coronary syndrome.4-11 Of
Annals of Emergency Medicine 115

A Prediction Rule for 30-Day Cardiac Events

Hess et al Goals of This Investigation This studys goal was to derive and internally validate a clinical prediction rule to identify adults with chest pain and at very low risk for a cardiac event, for whom additional ED investigations might be unnecessary.

Editors Capsule Summary

What is already known on this topic Emergency physicians have a low threshold for admitting patients with potential acute coronary syndromes. What question this study addressed This study developed a clinical decision rule for identication of low-risk patients. What this study adds to our knowledge Patients younger than 50 years and with atypical chest pain, a nonischemic ECG result, and 2 negative troponin test results during 6 hours without known coronary disease are at low risk for death, myocardial infarction, or revascularization within 30 days (sensitivity 100% [95% condence interval 97.2% to 100%] and specicity 20.9% [95% condence interval 16.9% to 24.9%]). How this is relevant to clinical practice It is unlikely this rule, even if validated, would change clinical practice because it formally codies what has probably been common practice for a long time.

MATERIALS AND METHODS

Study Design The study had a prospective observational cohort design. Patients were enrolled from July 2007 to February 2010. Setting The study was conducted in 3 academic EDs located in Canada and the United States. The Civic Campus of the Ottawa hospital, located in Ottawa, Ontario, Canada, has 60,000 annual ED visits. The Kingston General Hospital is located in Kingston, Ontario, Canada, and has 53,000 annual ED visits. The Saint Marys hospital is located in Rochester, MN, and has 73,000 annual ED visits. Selection of Participants We enrolled consecutive adults older than 24 years and with a primary complaint of anterior chest pain, for whom cardiac troponin testing was ordered by the treating physician as a component of an assessment for acute coronary syndrome. We dened anterior chest pain as any discomfort located on the anterior thorax between the suprasternal notch and the xiphoid process, using the posterior axillary line as the border between the anterior and posterior thorax. Patients who had acute STsegment elevation on the initial ECG in at least 2 contiguous leads, hemodynamic instability (pulse rate persistently greater than 100 or less than 50 beats/min or systolic blood pressure persistently below 90 mm Hg), an unreliable clinical history as determined by the treating physician, a clear traumatic cause of chest pain, a clinical history of cocaine use or positive test result for cocaine, a terminal noncardiac illness, pregnancy, previous enrollment within the past 30 days, or inability to receive follow-up by telephone were excluded. A study nurse assessed eligibility in collaboration with the treating physician and obtained informed consent. The institutional review board at each site approved all study procedures. In the Ottawa and Kingston, Ontario, sites, cardiology consultation is obtained at the discretion of the attending emergency physician for patients considered to be at moderate or high risk for acute coronary syndrome. The consulting cardiologist determines whether admission to an inpatient observation unit, hospital ward, or cardiac ICU is indicated on a case-by-case basis. Patients considered to be at low risk for acute coronary syndrome are discharged for outpatient follow-up once the ED evaluation for acute myocardial infarction is complete. In the Rochester, MN, site, patients at high risk for acute coronary syndrome are admitted to the hospital ward or cardiac ICU, and patients at low to moderate risk for acute coronary syndrome are admitted to an ED-based observation unit, where
Volume , . : February

the published clinical prediction rules, none have adhered to the methodological standards for clinical prediction rules in their development and validation,12-14 have performed well when validated in other clinical settings,15 or have used contemporary cardiac troponin assays as advocated by the latest consensus denition.16 In the absence of a highly sensitive and reliable clinical prediction rule, a recent American Heart Association Scientic Statement recommends that hemodynamically stable patients with possible acute coronary syndrome and no objective evidence of ischemia (normal or near-normal ECG result and negative baseline cardiac injury markers) be admitted to a chest pain unit for further evaluation by an accelerated diagnostic protocol, including cardiac stress testing.17 In describing the role of risk scores such as the Thrombolysis in Myocardial Infarction (TIMI) or Global Registry of Acute Coronary Events (GRACE), it mentions that favorable (low-risk) scores facilitate identication of patients for whom chest pain unit management might be appropriate. The current investigation seeks to augment these recommendations by developing a clinical prediction rule that identies ED patients at very low risk for 30-day cardiac events who might be safe for discharge without chest pain unit admission, provided outpatient follow-up can be arranged within 30 days of the ED visit. 116 Annals of Emergency Medicine

Hess et al serial cardiac troponin levels, serial ECGs, and cardiac stress testing are obtained according to protocol.18 Patients who rule out for acute myocardial infarction and have no evidence of ischemia on serial ECGs and cardiac stress testing are discharged from the observation unit for outpatient follow-up. Data Collection and Processing The 64 variables selected for assessment in the study were chosen according to review of the literature, input from the investigators, and feedback solicited from physician assessors. A complete list of the variables selected can be found elsewhere.19 Variable denitions were determined according to standardized reporting guidelines.20 The study was designed before the publication of the reporting guidelines for acute coronary syndrome research in EDs in Australasia.21 On-duty attending or supervised emergency medicine resident physicians performed patient assessments and completed a standardized data collection form before obtaining results of diagnostic tests (other than the initial ECG). Interobserver assessments were conducted when feasible by an independent attending physician blinded to the rst physicians data collection form. After enrollment, a study nurse reviewed the hospital record to collect data on the cardiovascular history, cardiac risk factors, and all investigations related to the visit. Data were recorded on predesigned case record forms. Each site investigator monitored local compliance with patient enrollment procedures throughout the duration of the study. We dened the time of symptom onset as the time of onset of the episode of pain that prompted the patients visit to the ED. Clinicians categorized the period from the onset of symptoms to ED presentation into one of 3 categories: today ( 24 hours), yesterday (24 to 48 hours), and greater than or equal to 2 days ago. If the pain developed within the past 24 hours, the specic time of onset was recorded. Known coronary artery disease was dened as at least one 50% stenosis on coronary angiography; demonstrated ECG changes, perfusion defects, or wall motion abnormalities on exercise, pharmacologic, or rest imaging studies; or previous documentation of acute myocardial infarction. Patients who reported a history of coronary artery disease during their ED visit that was not conrmed by objective criteria in the medical record were classied as having a self-reported history of coronary artery disease.20 We instructed clinicians to classify the cause of the chest pain syndrome with the following method: if, after taking into account all the factors from the history and physical examination, the clinician considered the pain syndrome to be of cardiac cause, it was classied as typical. If the clinicians overall assessment was that the pain syndrome was of noncardiac cause, it was classied as atypical. Blinded to predictor variables and patient outcome, 2 boardcertied emergency physicians (E.P.H., R.J.B.) classied all ECGs with a structured standardized format20 previously demonstrated to correlate with 30-day outcomes in ED patients with chest pain.22 A subset of 80 randomly selected ECGs (40 interpreted as acute ischemia and 40 not interpreted as acute
Volume , . : February

A Prediction Rule for 30-Day Cardiac Events ischemia) was independently interpreted by a third investigator (A.S.J.), and interrater reliability was assessed. Cardiac troponin T (Elecsys; Roche Diagnostics, Indianapolis, IN) (99th percentile reference limit 0.01 g/L; lower limit of detection 0.01 ng/mL; 10% coefcient of variation 0.035 ng/mL) was the cardiac biomarker used at the Rochester and Ottawa sites. At the Kingston site, cardiac troponin I (Beckman Coulter, Brea, CA) (99th percentile reference limit 0.04 g/L; lower limit of detection 0.01 g/L; 10% coefcient of variation 0.06 g/L23) was the cardiac biomarker in routine clinical use. Serum for cardiac troponin T or I was to be obtained at patient arrival and greater than or equal to 6 hours from pain onset, with at least 3 hours between samples.16,24 Outcome Measures We dened the primary composite outcome to be acute myocardial infarction, coronary revascularization, or death of cardiac or unknown cause within 30 days, whether the outcome occurred in the ED, in the hospital, or after discharge. Acute myocardial infarction was dened as a cardiac troponin measurement above the 99th percentile, with an increasing or decreasing pattern on serial measurements (for troponin I, dened as a 20% change between samples24; for troponin T, signicant change dened according to an algorithm developed in the Mayo Clinic cardiovascular laboratory), development of pathologic Q waves on the ECG, or ECG evolution consistent with acute myocardial infarction.16 Revascularization was dened as reestablishment of coronary patency by percutaneous coronary intervention or coronary artery bypass graft surgery. The primary outcome was ascertained by investigators blinded to the predictor variables. If a diagnosis could not be assigned, a cardiologist (A.S.J.) reviewed all the clinical data and assigned an adjudicated outcome diagnosis. All positive and 10% of randomly selected negative outcomes were conrmed by a second coinvestigator blinded to the standardized data collection forms. Disagreements were resolved by consensus. All management decisions were left to the discretion of the treating physician. Because not all patients with chest pain routinely undergo coronary angiography, structured telephone follow-up was conducted for all subjects by a study nurse, starting at 30 days. Patients were asked whether they had consulted another physician, had experienced a heart attack, or had undergone any cardiac testing, including coronary angiography with or without invasive intervention. If the patient answered in the afrmative or was unsure of the answer to any of these questions, additional documentation was sought from the facility in which any procedures were performed. If the study nurse was unable to contact the patient by telephone, outcomes were ascertained by review of the hospital medical record. For those subjects unable to be contacted by telephone and with no subsequent hospital visits, we reviewed the provincial coroners database (Ottawa and Kingston sites) or
Annals of Emergency Medicine 117

A Prediction Rule for 30-Day Cardiac Events the Social Security Death Index25 (Rochester site) to ascertain vital status. Primary Data Analysis We measured interobserver agreement for each variable by calculating the coefcient for dichotomous or categorical variables and the intraclass correlation coefcient26 for continuous variables. Variables with values greater than or equal to 0.6 were considered to represent substantial agreement and were considered for inclusion in the rule. Values were not calculated for variables collected from the medical record (eg, age, increased troponin level, history of coronary artery disease). We further explored variables of potential interest (either on a statistical or clinical basis) by using cut points to ensure that the nal model did not contain any continuous variables so that clinicians could categorize patients as high or low risk without performing any calculations. To aid in selection of the best variables for multivariable analyses, we used univariable logistic regression to determine the strength of association between each variable and the composite primary outcome. Variables found to be both reliable ( 0.6) and associated with the primary outcome ( .20) were evaluated with recursive partitioning. The objective was to identify the best combination of predictor variables that was highly sensitive for the primary outcome, with the maximum possible specicity and clinical sensibility. Recursive partitioning creates a branching decision tree by dividing the patient population into subgroups with and without the outcome of interest, according to contents of predictor variables in the subgroup. Because our previous experience suggested that recursive partitioning may be more suitable than logistic regression when the intent is to classify one outcome at the expense of the other (when high sensitivity is preferred to overall accuracy), we performed recursive partitioning with Answer Tree software (version 3.1; IBM SPSS, Chicago, IL). The variables chosen by the best model constituted the decision rule. We cross-validated the derived decision rule by comparing the classication of each patient to the actual status for the primary outcome, allowing us to estimate the sensitivity and specicity of the rule. Bootstrapping was used to calculate 95% condence intervals (CIs) for both the sensitivity and specicity estimates.27 For bootstrapping, 5,000 samples of 500 observations were generated, and the mean CIs for both estimates were reported. Because the sensitivity estimate was close to 1, a continuity correction was calculated for the CI, and a 1-sided test was used. To determine the 95% CI for proportions with correction for continuity, we used the method described by Newcombe.28 We estimated the sample size according to the precision of the sensitivity of the derived decision rule. As with previous decision rule studies, we prespecied 120 outcome events to derive a rule that is 100% sensitive, with a lower 95% condence limit of 97.0%.29,30 To have the greatest utility for practicing emergency physicians, we aimed to include at least 120 outcome events occurring outside the ED (inhospital or 118 Annals of Emergency Medicine

Hess et al

Figure 1. Flow diagram of multicenter prospective cohort study of ED patients with chest pain.

after ED discharge). Review of quality data from the Ottawa hospital indicated that 10% of patients who presented to the ED with chest pain would meet outcome criteria within 30 days. We estimated that half of these events would occur after hospital admission or ED discharge. The a priori sample size was estimated to be 2,400 patients.

RESULTS
We enrolled 2,718 patients from 3 academic EDs (Figure 1), representing 82% of eligible patients. There were 1,017 (37.4%) patients enrolled in Ottawa, Ontario, 496 (18.2%) in Kingston, Ontario, and 1,205 (44.3%) in Rochester, MN. Outcome status was ascertained at 30 days for all patients; 2,629 (97%) were able to be contacted by telephone for the 30-day structured follow-up, 58 (2%) had their vital status conrmed by review of subsequent visits in the hospital medical record, and the vital status of 31 (1%) was assessed by searching the provincial coroners database (Ottawa and Kingston sites) or the Social Security Death Index (Rochester site). There were 12 disagreements in outcome ascertainment between investigators that required adjudication by a third investigator. Of all patients, 336 (12%) met at least 1 component of the composite outcome within 30 days of the ED visit (Table 1). There were 167 (6%) patients who met criteria for acute myocardial infarction, 271 (10%) who underwent coronary revascularization, and 6 (0.2%) who died of a cardiac or unknown cause. The frequency of acute myocardial infarction and death was similar between the 3 study sites (Table E1, available online at http://www.annemergmed.com). Demographic and cardiovascular history characteristics of the enrolled patients were similar to those of the eligible missed patients (Table E2, available online at http://www. annemergmed.com). Table E3 (available online at http://www. annemergmed.com) details the proportion of patients who underwent cardiac diagnostic investigations to evaluate for acute coronary syndrome.
Volume , . : February

Hess et al
Table 1. Baseline patient characteristics, ECG ndings, and 30-day outcomes for 2,718 ED patients with chest pain.
Characteristic Demographics Mean age, y (SD) Age range, y Male patient Arrival by ambulance Median length of stay in department, h (IQR) Range Cardiac risk factors Hypertension Diabetes mellitus Hypercholesterolemia Family history of cardiac disease Smoking history Current Former ( 1 y ago) Former ( 1 y ago) No cardiac risk factors Cardiovascular history Previous myocardial infarction Angina (chest pain on exertion) Known coronary artery disease Congestive heart failure Cardiac arrest Ventricular tachycardia Atrial brillation Stroke or transient ischemic attack Peripheral vascular disease No cardiovascular history Initial ECG interpretation Normal (%) Nonspecic ST-segment changes Abnormal, not diagnostic Ischemia known to be old Ischemia not known to be old Other ECG ndings ST-segment depression 0.5 mm T-wave inversion Left bundle branch block Right bundle branch block Q waves Management Admitted to the hospital Admitted to the observation unit (n 1,205) Cardiac stress testing performed Coronary angiography performed Coronary stenosis 70% Outcomes Cardiac event Acute myocardial infarction Revascularization Death of cardiac/unknown cause Cardiac event occurring after ED disposition (inhospital or outof-hospital) No. (%) of Patients (n 2,718)* 60.0 (14.9) 2599 1,439 (52.9) 661 (24.3) 3.0 (3.1) 0.416.3 1,592 (58.6) 559 (20.6) 1,458 (53.6) 784 (28.8) 1,393 (51.3) 483 (17.8) 155 (5.7) 755 (27.8) 274 (10.1) 630 (23.2) 626 (23.0) 892 (32.8) 185 (6.8) 41 (1.5) 31 (1.1) 232 (8.5) 190 (7.0) 133 (4.9) 1,395 (51.3) 1,169 (43.0) 599 (22.0) 578 (21.3) 241 (8.9) 131 (4.8) 96 (3.5) 195 (7.2) 91 (3.4) 118 (4.3) 247 (9.1) 865 (31.8) 577 (47.9) 896 (33.0) 510 (18.8) 329 (12.1) 336 (12.4) 167 (6.1) 271 (10.0) 6 (0.2) 175 (6.4)

A Prediction Rule for 30-Day Cardiac Events

Table 1. Continued.
Characteristic Acute myocardial infarction Revascularization Death of cardiac/unknown cause No. (%) of Patients (n 2,718)* 8 (0.3) 171 (6.3) 4 (0.2)

IQR, Interquartile range. *Unless otherwise indicated. Among the 892 patients with known coronary artery disease, 666 (74.7%) were conrmed with objective criteria in the medical record and 226 (25.3%) were ascertained by patient self-report. Only 1 site (Rochester, MN) had an observation unit. In at least 1 major epicardial coronary artery. Acute myocardial infarction, revascularization, or death of cardiac/unknown cause within 30 days of the ED visit. Some patients experienced more than one type of cardiac event.

Table 2 shows the association between the predictor variables and the composite outcome, as determined by univariable analyses and the interobserver agreement for predictor variables obtained from the chest pain history and ECG. Of the 250 interobserver cases, the pain began in less than 24 hours in 124 cases. The intraclass correlation coefcient for time of onset was 0.916, indicating excellent interrater agreement. In general, patients with subsequent events had more previous cardiac events, had more risk factors by history, and were more apt to have abnormal ECG ndings. The model developed with recursive partitioning analysis is displayed in Figure 2. The variables in this nal statistical model were then combined into a simple, clinically sensible list to present the decision rule, titled the North American Chest Pain Rule (Box). The classication performance of the rule was accurate across 5,000 bootstrapped test sets and was consistent with the estimation from the original data set. Table 3 shows the classication performance of the model for predicting acute myocardial infarction, revascularization, or death within 30 days, using the bootstrapped 95% CI estimates. Figure 3A shows the performance of the prediction rule at an age cutoff of aged 50 years or younger. In this model, patients aged 40 years or younger and who otherwise have none of the risk factors can be safely discharged from the ED if the initial cardiac troponin result is negative. This group was represented in 7% of the cohort. Patients aged 41 to 50 years require a second troponin test, obtained at least 6 hours from pain onset, and can be discharged if the troponin measurement is negative and no other criteria are met. This group was represented in 11% of the cohort. If the age cutoff is increased to 60 years or younger, the specicity of the rule increases to 34% and 4 positive outcomes were missed. Each of these cases was considered to be an atypical presentation by the attending physician, had no new ischemic changes on ECG, and had negative serial cardiac troponin results. One patient developed ST-segment elevation on the ECG component of a dobutamine stress echo, had a 90% lesion in the right posterior descending artery identied on
Annals of Emergency Medicine 119

Volume , . : February

A Prediction Rule for 30-Day Cardiac Events

Hess et al

Table 2. Univariate association between demographic, clinical, and ECG predictor variables and the primary outcome in 2,718 ED patients with chest pain.
Group, No. (%) of Patients* Variable Demographics Mean age, y (SD) Male patient Arrival by ambulance Cardiac risk factors Hypertension Diabetes mellitus Hypercholesterolemia Family history of cardiac disease Any history of smoking Cardiovascular history Acute myocardial infarction Angina (chest pain on exertion) Known coronary artery disease Congestive heart failure Cardiac arrest Ventricular tachycardia Atrial brillation Stroke or transient ischemic attack Peripheral vascular disease Chest pain history Present on ED arrival Present at rest Resolution of pain before evaluation Worse with exertion Pleuritic Similar to previous ischemia Change in usual pattern of angina in past 24 h 2 episodes of pain in past 24 h Onset Abrupt ( 1 h) Gradual ( 1 h) Location on chest Center Left anterior Left lateral Right anterior Right lateral Pain description Pressure/squeezing Heavy Sharp Indigestion/burning quality Radiation Left arm/shoulder Right arm/shoulder Both arms/shoulders Neck/jaw Back Abdomen Associated symptoms Nausea or vomiting Shortness of breath Diaphoresis Worse with movement Typical for acute coronary syndrome Atypical for acute coronary syndrome Initial interpretation of the ECG (n for 80) Normal (%) Positive for Primary Outcome (n 336) 66.4 (12.8) 231 (68.8) 104 (31.0) 243 (72.3) 99 (29.5) 239 (71.1) 95 (28.3) 193 (57.4) 123 (36.6) 142 (42.3) 193 (57.4) 33 (9.8) 3 (0.9) 5 (1.5) 34 (10.1) 35 (10.4) 28 (8.3) 191 (56.8) 275 (81.8) 176 (52.4) 179 (53.3) 22 (6.5) 159 (47.3) 92 (27.4) 194 (57.7) 255 (75.9) 71 (21.1) 236 (70.2) 100 (29.8) 19 (5.7) 17 (5.1) 4 (1.2) 188 (60.0) 60 (17.9) 40 (11.9) 39 (11.6) 95 (28.3) 18 (5.4) 38 (11.3) 79 (23.5) 39 (11.6) 2 (0.6) 71 (21.1) 158 (47.0) 95 (28.3) 50 (14.9) 250 (74.4) 79 (23.5) 84 (25.0) Negative for Primary Outcome (n 2,382) 59.1 (15.0) 1,208 (44.4) 557 (20.5) 1,349 (49.6) 460 (16.9) 1,219 (44.8) 689 (28.9) 1,200 (50.4) 507 (21.3) 484 (20.3) 699 (29.3) 152 (6.4) 38 (1.6) 26 (1.1) 198 (8.3) 155 (6.5) 105 (4.4) 1,583 (66.5) 2,154 (90.4) 998 (41.9) 688 (28.9) 441 (18.5) 504 (21.2) 325 (13.6) 987 (41.4) 1,612 (67.7) 686 (28.8) 1,325 (55.6) 847 (35.6) 183 (7.7) 154 (6.5) 45 (1.9) 1,084 (45.5) 408 (17.1) 585 (24.6) 202 (8.5) 582 (24.4) 101 (4.2) 103 (4.3) 361 (15.2) 316 (13.3) 41 (1.7) 550 (23.0) 923 (38.8) 496 (20.8) 481 (20.2) 778 (32.6) 1,527 (64.1) 1,076 (45.2) Odds Ratio (95% CI) 1.03 (1.021.04) 2.14 (1.682.73) 1.48 (1.151.89) 2.00 (1.562.58) 1.75 (1.352.25) 2.35 (1.833.02) 0.97 (0.751.25) 1.33 (1.061.67) 2.14 (1.682.72) 2.87 (2.263.64) 3.25 (2.574.11) 1.60 (1.082.37) 0.56 (0.171.81) 1.37 (0.523.59) 1.24 (0.851.82) 1.67 (1.142.46) 1.99 (1.143.49) 0.67 (0.530.84) 0.48 (0.350.65) 1.53 (1.211.92) 2.81 (2.233.54) 0.31 (0.200.48) 3.35 (2.654.24) 2.39 (1.833.12) 1.93 (1.532.43) 1.50 (1.151.96) 0.66 (0.500.87) 1.88 (1.472.41) 0.77 (0.600.99) 0.72 (0.441.17) 0.77 (0.461.29) 0.63 (0.221.75) 1.52 (1.211.91) 1.05 (0.781.42) 0.42 (0.300.59) 1.42 (0.992.04) 1.22 (0.951.57) 1.28 (0.762.14) 2.82 (1.914.17) 1.72 (1.312.27) 0.86 (0.601.22) 0.34 (0.081.42) 0.89 (0.681.18) 1.40 (1.121.77 1.50 (1.161.94) 0.69 (0.500.95) 5.99 (4.627.77) 0.17 (0.130.22) 0.40 (0.310.51) Interobserver Agreement, Value (n 250) NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA 0.63 0.63 0.67 0.53 0.74 0.74 0.62 0.61 0.45 0.45 0.52 0.51 0.63 0.48 0.66 0.47 0.27 0.55 0.68 0.55 0.33 0.34 0.45 0.63 0.66 0.54 0.42 0.61 0.43 0.75 0.75 0.85 NA

120 Annals of Emergency Medicine

Volume , . : February

Hess et al
Table 2. Continued.

A Prediction Rule for 30-Day Cardiac Events

Group, No. (%) of Patients* Variable Nonspecic ST-segment changes (%) Abnormal but not diagnostic of ischemia (%) Infarction or ischemia known to be old (%) Infarction or ischemia not known to be old (%) Other ECG ndings ST-segment depression 0.5 mm T-wave inversion Left bundle-branch block Right bundle-branch block Q waves Positive for Primary Outcome (n 336)* 67 (19.9) 82 (24.4) 36 (10.7) 67 (19.9) 58 (17.3) 50 (14.9) 24 (7.1) 18 (5.4) 39 (11.6) Negative for Primary Outcome (n 2,382) 532 (22.3) 496 (20.8) 205 (8.6) 64 (2.7) 67 (2.8) 145 (6.1) 67 (2.8) 100 (4.2) 208 (8.7) Odds Ratio (95% CI) 0.86 (0.651.15) 1.22 (0.941.60) 1.24 (0.851.82) 9.03 (6.3012.95) 7.21 (4.9710.47) 2.70 (1.913.81) 2.66 (1.644.30) 1.29 (0.772.16) 1.37 (0.961.97)

Interobserver Agreement, Value (n 250) NA NA NA NA NA NA NA NA NA

*Unless otherwise indicated. The primary outcome was acute myocardial infarction, coronary revascularization, or death within 30 days of the ED visit. Attending physicians were instructed to record whether the chest pain syndrome was cardiac (typical for acute coronary syndrome) or noncardiac (atypical for acute coronary syndrome) in cause. Was dichotomously assessed according to whether ECGs were classied as infarction or ischemia not known to be old (acute ischemia) or other. Infarction or ischemia on ECG was dened as T-wave inversion greater than or equal to 0.2 mm or ST-segment depression greater than or equal to 1 mm in at least 2 contiguous leads.

coronary angiography that received a stent, and subsequently experienced a postrevascularization nonST-segment acute myocardial infarction. The other 3 patients underwent percutaneous coronary intervention and did not experience myocardial infarction or death.

LIMITATIONS
Our study has several potential limitations. We enrolled only patients who presented with a chest pain syndrome. Patients at risk for acute coronary syndrome who presented with non chest pain syndromes such as shortness of breath, nausea, or generalized fatigue were not included, which limits the application of our ndings to those who have chest pain as a presenting symptom. Risk stratication of patients with possible acute coronary syndrome who present with non chest pain syndromes should proceed as otherwise indicated in standard practice. Not all (82%) eligible patients were enrolled, likely because they were missed when the ED was busy. We did, however, collect demographic and cardiovascular history characteristics of eligible missed patients, and they were similar to those of eligible enrolled patients. Investigators, blinded to predictor variables and outcomes, interpreted ECGs. If the prediction rule were to be applied in clinical practice, the treating physician would interpret ECGs in the course of clinical care. However, the study investigators who interpreted the ECGs were blinded to both the predictor variables and outcome at the ECG interpretation and are boardcertied, practicing emergency physicians with training similar to that of those who would apply the rule in clinical practice. We did not include patients with potential acute coronary syndrome who were medically managed as a component of the primary outcome. As such, the prediction rule may not identify some patients who may benet from pharmacologic management. All enrolled patients did not undergo denitive
Volume , . : February

cardiac testing such as coronary angiography or cardiac stress testing introducing the potential for evaluation bias. We did, however, conduct structured telephone follow-up for all enrolled patients as a proxy outcome measure to attenuate the risk of evaluation bias. In deriving the rule, we assessed univariate associations between all potentially important predictors and included variables that were both associated with the outcome and of sufcient interrater reliability in multivariate analyses. Although this approach to prediction rule development involves multiple testing and introduces the possibility of overtting, we thought it important that all potentially important predictors that clinicians might consider important be evaluated in the process of deriving the rule. In addition, clinical prediction rule development involves several phases, including derivation, validation, and implementation, and the robustness of the model is tested in each phase. Some may question including coronary revascularization as an outcome because the decision to treat a coronary stenosis may vary between interventional cardiologists and between regions of the country. However, the absence of acute ischemic ECG changes and troponin increases does not necessarily indicate the absence of ischemic heart disease. Cardiac ischemia may not be severe enough to cause sufcient myocardial damage to release detectable quantities of troponin,31 and the initial ECG result may be normal or nonspecic, even in cases that ultimately meet criteria for acute myocardial infarction.32 Despite the user-dependent nature of coronary revascularization, excluding coronary revascularization as an outcome criterion may miss cases of acute coronary syndrome.

DISCUSSION
We successfully developed a highly sensitive clinical decision rule to identify patients at very low risk for acute myocardial
Annals of Emergency Medicine 121

A Prediction Rule for 30-Day Cardiac Events

Hess et al

Figure 2. Derivation of clinical decision rule by recursive partitioning.

infarction, revascularization, or death within 30 days of the ED visit. This rule would allow emergency physicians to determine which patients require additional ED investigations and which can be safely discharged for further outpatient evaluation. Patients who have none of the risk factors in the decision rule and are aged 40 years or younger can be safely discharged if the initial cardiac troponin level is less than the 99th percentile reference limit; patients aged 41 to 50 years can be discharged if a second cardiac troponin value less than the 99th percentile is obtained at least 6 hours from pain onset. Patients aged 60 years or younger have a very low event rate and can be considered for discharge in practice settings in which a less than 2% miss rate is acceptable, depending on local resources. To the best of the authors knowledge, this is the largest clinical decision rule study in ED patients with chest pain that has been conducted strictly according to the methodological standards for clinical 122 Annals of Emergency Medicine

prediction rules. Nevertheless, the prediction rule requires prospective multicenter validation before use in clinical practice. Several prediction models have been developed to identify ED chest pain patients who are at low risk for a cardiac event. The TIMI risk score, which was initially developed in patients with unstable angina or nonST-segment myocardial infarction, has been broadly validated in several ED populations with chest pain.33 The pooled sensitivity of the TIMI risk score in a recent systematic review was 97%, the pooled specicity was 25%, and the miss rate of patients with a score of zero was 2%. This miss rate is likely too high for use in the US setting. Than et al34 recently reported the performance of the TIMI risk score combined with a point-of-care multibiomarker panel using troponin, creatine kinase MB, and myoglobin. Their protocol efciently identied 418 of 421 major adverse cardiac events within 2 hours of presentation (sensitivity 99.3%; specicity
Volume , . : February

Hess et al
Table 3. Prognostic accuracy of clinical decision rule for 30day cardiac events in 2,718 patients with chest pain and possible acute coronary syndrome.
A, Age cutoff 50 years or younger Decision Rule Yes No Cardiac Event Within 30 Days 336 0 No Cardiac Event Within 30 Days 1,885 497

A Prediction Rule for 30-Day Cardiac Events

Sensitivity 100.0% (95% CI 97.2% to 100.0%); specicity 20.9% (95% CI 16.9% to 24.9%); positive predictive value 15.1% (95% CI 13.7% to 16.7%); negative predictive value 100.0% (95% CI 99.0% to 100.0%); stress testing proportion 81.7% (95% CI 80.2% to 83.1%).

B, Age cutoff 60 years or younger Decision Rule Yes No Cardiac Event Within 30 Days 332 4 No Cardiac Event Within 30 Days 1,581 801

Sensitivity 98.8% (95% CI 95.4% to 100.0%); specicity 33.6% (95% CI 28.9% to 38.1%); positive predictive value 17.4% (95% CI 15.7% to 19.1%); negative predictive value 99.5% (95% CI 98.6% to 99.8%); stress testing proportion 70.4% (95% CI 68.6% to 72.1%).

11.0%) and identied 10% of the cohort with a TIMI risk score of zero who could be safely discharged from the ED. Our clinical decision rule identied 18% of a cohort who could be safely discharged from the ED, using only contemporary cardiac troponin assays, and as more sensitive cardiac troponin assays are developed this proportion is likely to increase.35 The History, ECG, Age, Risk factors and Troponin (HEART) score has been retrospectively derived36 and validated37 in ED patients with chest pain. The score uses data from the history, ECG, age, cardiac risk factors, and troponin level to predict the short-term risk of acute myocardial infarction, revascularization, and death. In the derivation phase, 2.5% of patients classied as low risk (HEART score 0 to 3) had a cardiac event within 3 months. In the validation phase, the sensitivity of a HEART score of 0 to 3, however, was 98.1% and the specicity was 15.5%. These are less optimal prognostic accuracy estimates compared with those observed in the current investigation. The Vancouver Chest Pain Rule, which was developed according to the methodological standards for clinical prediction rules, also identied patients younger than 40 years as a low-risk group appropriate for discharge from the ED.8 For patients older than 40 years, these investigators found that patients without certain chest pain characteristics (pain radiating to neck, arm, or jaw and pain does not increase with deep breath or palpation) could also be classied as low risk and considered for ED discharge. In our study, however, which included 250 cases in which the interobserver reliability of predictor variables was assessed, only radiation of pain to the back or abdomen had a of at least 0.6; radiation of pain to other areas was insufciently reliable to be considered for inclusion in the rule. We also used a more sensitive marker of
Volume , . : February

Figure 3. North American Chest Pain Rule.* *Does not apply to patients with acute ST-segment elevation, hemodynamic instability, a clear traumatic cause of chest pain, a history of cocaine use or tested positive for cocaine, a terminal noncardiac illness, or who were considered to have an unreliable history by the treating physician. Dened as ST-segment deviation greater than or equal to 1 mm or T-wave inversion greater than or equal to 0.2 mm in at least 2 contiguous leads. As determined by the attending emergency physician. Samples for contemporary cardiac troponin assay analysis were obtained at presentation and at least 6 hours from symptom onset in this cohort. Positive was dened as greater than 99th percentile reference limit. In addition, an age cutoff of 60 years or younger can be used with a less than 2% miss rate and lower proportion of patients who require urgent stress testing in practice settings in which this miss rate is acceptable.

cardiac injury, cardiac troponin, rather than creatine kinase MB, as a predictor variable in our study. Marsan et al10 reported ndings similar to those observed in the current investigation. These investigators observed that for ED chest pain patients younger than 40 years and with a normal ECG result, an initial normal cardiac biomarker result, no cardiac risk factors, and no abnormal cardiac history, there were no cardiac events at 30 days. Collin et al38 reported no cardiac events at 1 year for a cohort of 609 patients with similar characteristics. Overall, the observations in the current investigation add to the growing literature that urgent cardiac stress testing in ED chest pain patients younger than 40 years has limited utility and is likely an inefcient use of resources. However, in our population we also observed that a physician impression of atypical or noncardiac chest pain had a high degree of interobserver reliability ( 0.75), had a strong negative association with 30-day cardiac events (odds ratio 0.14;
Annals of Emergency Medicine 123

A Prediction Rule for 30-Day Cardiac Events 95% CI 0.13 to 0.21), and, when combined with ndings routinely available from the history and physical examination, ECG, and troponin level, assists in identifying patients aged 50 years or younger who are at very low risk for a cardiac event within 30 days. Our prediction rule may differentially affect resource use, depending on local practice patterns. In the Ottawa and Kingston sites, the hospital admission rates were 18% and 21%, respectively, and neither of these sites have an ED observation unit. Use of the prediction rule in these sites would likely increase resource use. Using the age cutoff of 60 years or younger may attenuate the effect on resource use and could be anticipated to have a miss rate of less than 2%. In the Rochester site, however, the hospital admission rate was 48% and the ED observation unit admission rate was 48% (combined admission rate of 96%). At the age cutoff of 50 years or younger, the estimated testing rate was 82%. Use of the rule at the Rochester site would likely decrease resource use. The safety and potential utility of our prediction rule depends on the availability of outpatient follow-up within 30 days of the ED visit. Some patients may not have ready access to a primary care provider. Additional ED investigations may be appropriate for patients with unreliable access to follow-up or in health care settings in which outpatient follow-up is not readily available. In summary, we derived and internally validated a clinical prediction rule that identied adults with chest pain at very low risk for a cardiac event for whom additional ED investigations may be unnecessary. Application of the rule has potential to increase the efciency of diagnostic investigation; patients who do not meet any of the high-risk criteria could be safely discharged from the ED, and patients with any of the risk factors in the rule could be considered for additional investigation or observation unit admission. Prospective multicenter validation is needed before use in clinical practice.
Supervising editor: Judd E. Hollander, MD Author contributions: All authors contributed signicantly to the study and the preparation of the article. EPH, JJP, ASJ, GAW, and IGS were responsible for the conception and design of the study. LAC, VT, DA, and ATS assisted with data acquisition. EPH and RJB were responsible for data acquisition, data entry, and preliminary analysis. EPH and GAW performed all statistical analyses. All authors participated in the drafting and multiple revisions of the article and gave nal approval of the submitted work. EPH takes responsibility for the paper as a whole. Funding and support: By Annals policy, all authors are required to disclose any and all commercial, nancial, and other relationships in any way related to the subject of this article as per ICMJE conict of interest guidelines (see www.icmje.org). The study was funded by a Fellow-to-Faculty Transition award from the American Heart Association, the Society for Academic Emergency Medicine, and the Emergency Medicine Foundation (award 0775030N). The study was also

Hess et al
supported by the Ontario Innovation Fund and the University of Ottawa Department of Emergency Medicine. Publication dates: Received for publication April 29, 2011. Revisions received June 27, 2011, and July 5, 2011. Accepted for publication July 13, 2011. Available online September 1, 2011. Earn CME Credit: Continuing Medical Education is available for this article at http://www.ACEP-EMedHome.com. Address for correspondence: Erik P. Hess, MD, MSc, E-mail hess.erik@mayo.edu.

REFERENCES
1. Niska RW, Bhuiya F, Xu J. National Hospital Ambulatory Medical Care Survey: 2007 emergency department summary. Natl Health Stat Rep. 2010;26:1-32. 2. Schull MJ, Vermeulen MJ, Stukel TA. The risk of missed diagnosis of acute myocardial infarction associated with emergency department volume. Ann Emerg Med. 2006;48:647-655. 3. Pines JM, Szyld D. Risk tolerance for the exclusion of potentially life-threatening diseases in the ED. Am J Emerg Med. 2007;25: 540-544. 4. Baxt WG, Skora J. Prospective validation of articial neural network trained to identify acute myocardial infarction. Lancet. 1996;347:12-15. 5. Goldman L, Cook EF, Brand DA, et al. A computer protocol to predict myocardial infarction in emergency department patients with chest pain. N Engl J Med. 1988;318:797-803. 6. Goldman L, Weinberg M, Weisberg M, et al. A computer-derived protocol to aid in the diagnosis of emergency room patients with acute chest pain. N Engl J Med. 1982;307:588-596. 7. Selker HP, Beshansky JR, Grifth JL, et al. Use of the acute cardiac ischemia time-insensitive predictive instrument (ACI-TIPI) to assist with triage of patients with chest pain or other symptoms suggestive of acute cardiac ischemia. A multicenter, controlled clinical trial. Ann Intern Med. 1998;129:845-855. 8. Christenson J, Innes G, McKnight D, et al. A clinical prediction rule for early discharge of patients with chest pain. Ann Emerg Med. 2006;47:1-10. 9. Boersma E, Pieper KS, Steyerberg EW, et al. Predictors of outcome in patients with acute coronary syndromes without persistent ST-segment elevation. Results from an international trial of 9461 patients. The PURSUIT Investigators. Circulation. 2000;101:2557-2567. 10. Marsan RJ Jr, Shaver KJ, Sease KL, et al. Evaluation of a clinical decision rule for young adult patients with chest pain. Acad Emerg Med. 2005;12:26-31. 11. Lyon R, Morris AC, Caesar D, et al. Chest pain presenting to the emergency departmentto stratify risk with GRACE or TIMI? Resuscitation. 2007;74:90-93. 12. McGinn TG, Guyatt GH, Wyer PC, et al. Users guides to the medical literature, XXII: how to use articles about clinical decision rules. Evidence-Based Medicine Working Group. JAMA. 2000;284: 79-84. 13. Laupacis A, Sekar N, Stiell IG. Clinical prediction rules. A review and suggested modications of methodological standards. JAMA. 1997;277:488-494. 14. Stiell IG, Wells GA. Methodologic standards for the development of clinical decision rules in emergency medicine. Ann Emerg Med. 1999;33:437-447. 15. Hess EP, Thiruganasambandamoorthy V, Wells GA, et al. Diagnostic accuracy of clinical prediction rules to exclude acute

124 Annals of Emergency Medicine

Volume , . : February

Hess et al
coronary syndrome in the emergency department setting: a systematic review. CJEM. 2008;10:373-382. Thygesen K, Alpert JS, White HD, et al. Universal denition of myocardial infarction. Circulation. 2007;116:2634-2653. Amsterdam EA, Kirk JD, Bluemke DA, et al. Testing of low-risk patients presenting to the emergency department with chest pain: a scientic statement from the American Heart Association. Circulation. 2010;122:1756-1776. Cullen MW, Reeder GS, Farkouh ME, et al. Outcomes in patients with chest pain evaluated in a chest pain unit: the Chest Pain Evaluation in the Emergency Room study cohort. Am Heart J. 2011;161:871-877. Hess EP, Wells GA, Jaffe A, et al. A study to derive a clinical decision rule for triage of emergency department patients with chest pain: design and methodology. BMC Emerg Med. 2008;8:3. Hollander JE, Blomkalns AL, Brogan GX, et al. Standardized reporting guidelines for studies evaluating risk stratication of ED patients with potential acute coronary syndromes. Acad Emerg Med. 2004;11:1331-1340. Cullen L, Than M, Brown AF, et al. Comprehensive standardized data denitions for acute coronary syndrome research in emergency departments in Australasia. Emerg Med Australas. 2010;22:35-55. Forest RS, Shofer FS, Sease KL, et al. Assessment of the standardized reporting guidelines ECG classication system: the presenting ECG predicts 30-day outcomes. Ann Emerg Med. 2004;44:206-212. Morrow DA, Rifai N, Sabatine MS, et al. Evaluation of the AccuTnI cardiac troponin I assay for risk assessment in acute coronary syndromes. Clin Chem. 2003;49:1396-1398. Macrae AR, Kavsak PA, Lustig V, et al. Assessing the requirement for the 6-hour interval between specimens in the American Heart Association classication of myocardial infarction in epidemiology and clinical research studies. Clin Chem. 2006;52:812-818. Quinn J, Kramer N, McDermott D. Validation of the Social Security Death Index (SSDI): an important readily-available outcomes database for researchers. West J Emerg Med. 2008;9:6-8. Shrout PE, Fleiss JL. Intraclass correlations: uses in assessing rater reliability. Psychol Bull. 1979;86:420-428. Efron B. The Jacknife, the Bootstrap and Other Resampling Plans. Philadelphia, PA: Society for Industrial & Applied Mathematics; 1982.

A Prediction Rule for 30-Day Cardiac Events

28. Newcombe RG. Two-sided condence intervals for the single proportion: comparison of seven methods. Stat Med. 1998;17: 857-872. 29. Stiell IG, Wells GA, Vandemheen KL, et al. The Canadian C-Spine Rule for radiography in alert and stable trauma patients. JAMA. 2001;286:1841-1848. 30. Perry JJ, Stiell IG, Sivilotti MLA, et al. High risk clinical characteristics for subarachnoid haemorrhage in patients with acute headache: prospective cohort study. BMJ. 2010;341: c5204. 31. Anderson JL, Adams CD, Antman EM, et al. ACC/AHA 2007 guidelines for the management of patients with unstable angina/ non-ST-elevation myocardial infarction: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines (Writing Committee to Revise the 2002 Guidelines for the Management of Patients With Unstable Angina/Non-ST-Elevation Myocardial Infarction) developed in collaboration with the American College of Emergency Physicians, the Society for Cardiovascular Angiography and Interventions, and the Society of Thoracic Surgeons; endorsed by the American Association of Cardiovascular and Pulmonary Rehabilitation and the Society for Academic Emergency Medicine. J Am Coll Cardiol. 2007;50:e1-157. 32. Welch RD, Zalenski RJ, Frederick PD, et al. Prognostic value of a normal or nonspecic initial electrocardiogram in acute myocardial infarction. JAMA. 2001;286:1977-1984. 33. Hess EP, Agarwal D, Chandra S, et al. Diagnostic accuracy of the TIMI risk score in patients with chest pain in the emergency department: a meta-analysis. CMAJ. 2010;182:1039-1044. 34. Than M, Cullen L, Reid CM, et al. A 2-h diagnostic protocol to assess patients with chest pain symptoms in the Asia-Pacic region (ASPECT): a prospective observational validation study. Lancet. 2011. 35. Thygesen K, Mair J, Katus H, et al. Recommendations for the use of cardiac troponin measurement in acute cardiac care. Eur Heart J. 2010;31:2197-2204. 36. Six AJ, Backus BE, Kelder JC. Chest pain in the emergency room: value of the HEART score. Neth Heart J. 2008;16:191-196. 37. Backus BE, Six AJ, Kelder JC, et al. Chest pain in the emergency room: a multicenter validation of the HEART score. Crit Pathw Cardiol. 2010;9:164-169. 38. Collin MJ, Weisenthal B, Walsh KM, et al. Young patients with chest pain: 1-year outcomes. Am J Emerg Med. 2011;29:265270.

16. 17.

18.

19.

20.

21.

22.

23.

24.

25.

26. 27.

Volume , . : February

Annals of Emergency Medicine 125

A Prediction Rule for 30-Day Cardiac Events

Table E1. Frequency of acute myocardial infarction and death of cardiac or unknown cause at the 3 study sites.
Site Kingston, Ontario (n 496) Ottawa, Ontario (n 1,017) Rochester, MN (n 1,205) Acute Myocardial Infarction, No. (%) 36 (7.3) 68 (6.7) 63 (5.2) Death, No. (%) 2 (0.4) 2 (0.2) 2 (0.2)

Hess et al
Table E2. Characteristics of eligible enrolled and eligible missed patients.
Characteristic Mean age, y (SD) Range Male patient (%) Previous myocardial infarction (%) Known coronary artery disease (%) Congestive heart failure (%) Atrial brillation (%) Eligible Enrolled (n 2,718) 60.0 (14.9) 2599 1,439 (52.9) 630 (23.2) 892 (32.8) 185 (6.8) 232 (8.5) Missed Eligible (n 604) 62.0 (15.5) 2894 340 (56.3) 109 (18.0) 199 (32.9) 32 (5.3) 42 (7.0)

Table E3. Proportion of cohort who underwent cardiac diagnostic investigations.

Diagnostic Investigation Cardiac stress testing Exercise treadmill testing Nuclear myocardial perfusion imaging Stress echo Coronary computed tomography angiography Coronary angiography No. (%) (n 2,718) 896 (33.0) 230 (8.5) 455 (16.7) 211 (7.8) 47 (1.7) 510 (18.8)

125.e1 Annals of Emergency Medicine

Volume , . : February