emergency physician high pretest probability for acute coronary syndrome correlates with adverse...

Emergency Physician High Pretest Probabilityfor Acute Coronary Syndrome Correlateswith Adverse Cardiovascular OutcomesAbhinav Chandra, MD, Christopher J. Lindsell, PhD, Alexander Limkakeng, MD, Deborah B. Diercks,MD, James W. Hoekstra, MD, Judd E. Hollander, MD, J. Douglas Kirk, MD, W. Frank Peacock, MD,W. Brian Gibler, MD, and Charles V. Pollack, MD, on behalf of the EMCREG i*trACS Investigators

AbstractObjectives: The value of unstructured physician estimate of risk for disease processes, other than acutecoronary syndrome (ACS), has been demonstrated. The authors sought to evaluate the predictive valueof unstructured physician estimate of risk for ACS in emergency department (ED) patients without obvi-ous initial evidence of a cardiac event.

Methods: This was a post hoc secondary analysis of the Internet Tracking Registry for Acute CoronarySyndromes (i*trACS), a prospectively collected multicenter data registry of patients over the age of18 years presenting to the ED with symptoms of ACS between 1999 and 2001. In this registry, followingpatient history, physical exam, and electrocardiogram (ECG), the unstructured treating physician esti-mate of risk was recorded. A 30-day follow-up and a medical record review were used to determinerates of adverse cardiac events, death, myocardial infarction (MI), or revascularization procedure. Theanalysis included all patients with nondiagnostic ECG changes, normal initial biomarkers, and a non-MIinitial impression from the registry and excluded those without complete data or who were lost tofollow-up. Data were stratified by unstructured physician risk estimate: noncardiac, low risk, high risk,or unstable angina.

Results: Of 15,608 unique patients in the registry, 10,145 met inclusion ⁄ exclusion criteria. Patients weredefined as having unstable angina in 6.0% of cases; high risk, 23.5% of cases; low risk, 44.2%; and non-cardiac, 26.3% of cases. Adverse cardiac event rates had an inverse relationship, decreasing from 22.0%(95% confidence interval [CI] = 18.8% to 25.6%) for unstable angina, 10.2% (95% CI = 9.0% to 11.5%) forthose stratified as high risk, 2.2% (95% CI = 1.8% to 2.6%) for low risk, and to 1.8% (95% CI = 1.4% to2.4%) for noncardiac. The relative risk (RR) of an adverse cardiac event for those with an initial label ofunstable angina compared to those with a low-risk designation was 10.2 (95% CI = 8.0 to 13.0). The RRof an event for those with a high-risk initial impression compared to those with a low-risk initial impres-sion was 4.7 (95% CI = 3.8 to 5.9). The risk of an event among those with a low-risk initial impressionwas the same as for those with a noncardiac initial impression (RR = 0.83, 95% CI = 0.6 to 1.2).

Conclusions: In ED patients without obvious initial evidence of a cardiac event, unstructured emergencyphysician (EP) estimate of risk correlates with adverse cardiac outcomes.

ACADEMIC EMERGENCY MEDICINE 2009; 16:740–748 ª 2009 by the Society for Academic EmergencyMedicine

Keywords: chest pain, risk stratification, acute coronary syndrome

ISSN 1069-6563 ª 2009 by the Society for Academic Emergency Medicine740 PII ISSN 1069-6563583 doi: 10.1111/j.1553-2712.2009.00470.x

From the Division of Emergency Medicine, Duke University Medical Center (AC, AL), Durham, NC; the Department ofEmergency Medicine, University of Cincinnati (CJL, WBG), Cincinnati, OH; the Department of Emergency Medicine, Universityof California (DBD, JDK), Davis, Sacramento, CA; the Division of Emergency Medicine, Wake Forest University (JWH), Winston-Salem, NC; the Department of Emergency Medicine, University of Pennsylvania (JEH), Philadelphia, PA; the Department ofEmergency Medicine, Cleveland Clinic (WFP), Cleveland, OH; and the Department of Emergency Medicine, PennsylvaniaHospital (CVP), Philadelphia, PA.Received January 20, 2009; revision received April 17, 2009; accepted April 20, 2009.Presented at the American College of Emergency Physicians Scientific Assembly, New Orleans, LA, October 2006.The i*trACS project was funded in part by Millennium Pharmaceuticals and Schering-Plough Pharmaceuticals.Address for correspondence and reprints: Abhinav Chandra, MD; e-mail: [email protected].

T he Institute of Medicine reports that 113.9 millionemergency department (ED) visits occurred in2003,1 with chest pain as the second most com-

mon reason (5.1%).2 Cardiovascular disease is a leadingcause of all deaths in the United States, and missing amyocardial infarction (MI) is a leading medicolegal con-cern for emergency physicians (EPs).3,4 Therefore, thedetermination of which patients with symptoms of acutecoronary syndrome (ACS) require workup is an impor-tant decision for physicians.

Although chest pain may be due to a life-threateningdisease, benign etiologies such as chest wall pain, bron-chitis, or esophagitis are more common than an MI.5,6

However, research has shown that the informationmost commonly available at the time of presentation,namely patient history, physical exam, electrocardio-gram (ECG), and initial serum biomarkers, are insuffi-cient by themselves to rule out the entire spectrum ofACS.5,7–20 Because the ECG and biomarkers are ofteninitially normal, the physician’s estimate of a patient’srisk for ACS is usually the critical determinant ofpatient disposition. The desire to avoid missing ACShas increased rates of ACS evaluation so much thatsome chest pain centers now report that 98% of theirevaluations are negative.21 This has led some to ques-tion the value of ‘‘physician judgment’’ when it comesto ACS evaluation.

We sought to determine the predictive value of clini-cians’ estimate of risk for ACS. We sought to study apatient population in which the physician’s estimate ofa patient’s risk for ACS is most critical: those with anondiagnostic ECG and normal initial cardiac markers.We hypothesized that in such a population, patientswho were estimated to have unstable angina or highrisk for ACS would have higher rates of a majoradverse cardiac event (defined as death, MI, or revascu-larization within 30 days) than those categorized as lowrisk for ACS or to have a noncardiac etiology of theirsymptoms.

METHODS

Study DesignThis was a post hoc secondary analysis of the InternetTracking Registry of Acute Coronary Syndromes(i*trACS). The details of the i*trACS registry have beendescribed in a previous publication.6 Briefly, i*trACS isa multicenter, prospective registry of ED patients withundifferentiated symptoms that may possibly representACS. Institutional review boards or ethics committeesapproved the study at all sites, waiving the need forconsent at all centers except one, where they requiredverbal consent prior to data collection.

Study Setting and PopulationThe i*trACS collaborative project enrolled ED patientsprospectively at eight centers in the United States andone in Singapore from June 1, 1999, to August 1, 2001.Six centers are academic teaching sites with an emer-gency medicine residency program, and two are com-munity hospitals. ED volumes at enrolling sites rangedfrom 10,000 to 160,000 visits per year. One center wasfrom the Pacific region, four from the Midwest region,

two from the Rocky Mountain region, one from theMid-Atlantic region, and one in Singapore.

Registry ProtocolA convenience sample of patients was prospectivelyenrolled into i*trACS if they were at least 18 years oldand received an ECG for symptoms of chest pain orangina equivalent (i.e., syncope, shortness of breath,palpitations, nausea, arm discomfort) in the ED. Datacollected included ECG interpretation, presenting signs,symptoms, coronary risk factors, and concomitant med-ications.22

An unstructured physician estimate of diagnosis orrisk for ACS was recorded based on clinical impressionby the physician for the patient. The options availablefor coding were acute MI, unstable angina, high-riskchest pain, low-risk chest pain, or noncardiac chestpain. Low risk was defined as the patient likely to besent home; high risk was defined as the patient forwhom serial ECG or marker tests were likely to beordered or for whom admission to a chest pain unitwas likely. The coding of the case report form wascompleted based upon the presenting history, physicalexamination, and ECG only at the time of the patientencounter by the treating physician. The pretest proba-bility coding was chosen from a designated list on thecase report form.

The ED and hospital course were followed by medi-cal record review or daily follow-up of patients admit-ted to the inpatient setting. The chart review recordedthe ED disposition, ED final diagnosis, results frominvasive or noninvasive provocative tests (exercise andpharmacologic stress tests, or cardiac catheterization),laboratory studies, vital signs, treatments, and hospitaldischarge diagnosis. A 30-day telephone follow-up wasattempted on all patients to determine outcomes andthe presence of cardiac events and continued to beattempted for 90 days, if initially unsuccessful. This wassupplemented with a structured chart review of theindex ED visit and associated hospitalization and deathregistry review by trained personnel if phone contactwas not possible.6 We were able to follow-up and havecomplete data on 93% of the registry.

Patient Selection for AnalysisWe sought to identify a patient population for whom thetreating physician’s unstructured estimate of risk is theprimary determinant of disposition decision-making.Therefore, for the present analysis, we excluded registrypatients with ECG evidence of ST-segment elevation ordepression, positive initial cardiac marker testing, or aninitial impression of acute MI. Patients with missingdemographic data, missing initial impression, unknowndisposition, or lost to follow-up were also excluded.Patients included in i*trACS more than once wereincluded for their last ED visit only; prior visits wereconsidered a component of medical history. Patientswere categorized by unstructured physician estimate ofunstable angina, high risk, low risk, or noncardiac.

Outcome MeasuresAll patients in the registry were evaluated for majoradverse cardiac events. Patients were considered positive

ACAD EMERG MED • August 2009, Vol. 16, No. 8 • www.aemj.org 741

for outcomes if any of the following were present onchart review or follow-up: greater than 70% stenosis inany coronary vessel at catheterization, MI, percutane-ous coronary intervention, coronary artery bypassgrafting, all-cause death, or a diagnosis-related group(DRG) code indicating revascularization (percutaneouscoronary intervention, coronary artery stent placement,or coronary artery bypass grafting), acute MI, non–Q-wave MI, or death. Death from a cardiac cause (asopposed to all-cause death) is reported if it occurredduring hospitalization and was verified.

Data AnalysisData are reported as means ± standard deviations(±SDs). Frequencies and proportions are reported with95% confidence intervals (CIs). Comparisons betweengroups used t-tests or chi-square tests as appropriate.Analyses were conducted using SPSS v16.0 (SPSS Inc.,Chicago, IL) and Microsoft Excel (Microsoft Corp., Red-mond, WA).

RESULTS

Characteristics of Study SubjectsThe i*trACS registry include 17,737 patient visits, repre-senting 15,608 unique patients (Figure 1). The mean agewas 54 years (SD ± 16 years), with 51.3% femalepatients. Of these patients, 2.7% had a final ED diagno-sis of ST-segment elevation myocardial infarction(STEMI), and 18.3% of patients had a final ED diagnosisthat was related to coronary ischemia (STEMI, non–Q-wave MI, unstable angina, or stable angina).23

There were 10,713 patients meeting the study inclu-sion criteria of no ST-segment elevation or depressionon ECG, normal initial cardiac biomarkers in the ED(troponin, creatine kinase-myocardial band, or myoglo-bin), and no initial impression of MI. From these, 568

patients were excluded due to missing data (n = 124) orinability to follow-up (n = 444). There are differences indemographics among patients excluded for lack ofcomplete data, and the major adverse cardiac eventrates were higher among those included than thoseexcluded (Table 1).

Main ResultsThe final data set for this analysis consisted of 10,145patients. Of these, 604 patients (6.0%) were diagnosedas presenting with unstable angina, 2,383 (23.5%) ashigh risk, 4,487 (44.2%) as low risk, and the remaining2,671 (26.3%) as noncardiac. As the risk categoryincreased from noncardiac to unstable angina, the pro-portion of patients with traditional risk factors tendedto increase (Table 2).

Overall, there were 522 subjects (5.2%) with majoradverse cardiac events (MI, death, or revascularizationduring the index visit or within 30 days). The rates ofadverse cardiac events were 1.8% (95% CI = 1.4% to2.4%), 2.2% (95% CI = 1.8% to 2.6%), 10.2% (95%CI = 9.0% to 11.5%), and 22.0% (95% CI = 18.8% to25.6%) for the noncardiac, low risk, high risk, andunstable angina groups, respectively. Table 3 reportsthe individual numbers of events. These totals exceedthe before mentioned rates, as any patient may havehad several events. The relative risk (RR) of a majoradverse cardiac event for those with an initial diagnosisof unstable angina, compared to those with a low-riskdesignation, was 10.2 (95% CI = 8.0 to 13.0). The RR ofan event for those with a high-risk initial impressioncompared to those with a low-risk initial impressionwas 4.7 (95% CI = 3.8 to 5.9). The risk of an eventamong those with a low-risk initial impression was thesame as for those with a noncardiac initial impression(RR = 0.83, 95% CI = 0.6 to 1.2).

Finally, characteristics of chest pain patients dis-charged from the ED were described. In the group ofpatients designated as unstable angina, 43 (7.1%) weredischarged home from the ED, while 524 (22.0%) weredischarged home from the high-risk group, 2,627(58.5%) were discharged home from the low-riskgroup, and 2,031 (76.0%) were discharged from thenoncardiac group. Among those discharged from theED, 24 major adverse cardiac events occurred (0.5%,95% CI = 0.3% to 0.7%). Five events occurred amongthe 529 patients discharged after an initial estimate ofhigh risk (1.0%, 95% CI = 0.4% to 2.3%). Among the2,638 low-risk patients discharged, 11 (0.4%, 95%CI = 0.2% to 0.8%) had an event, and among the 2,039noncardiac patients discharged, 8 (0.4%, 95% CI = 0.2%to 0.8%) had an event (Table 4).

DISCUSSION

The magnitude of the challenge for chest pain patientrisk stratification in the ED is significant. It is estimatedthat 5 million ED visits are for chest pain or its equiva-lent annually, yet approximately 3.75 million of themwill be determined to have a noncardiac etiology fortheir symptoms.24 Most of these patients have a diversespectrum of historical features and a nondiagnosticECG.23 We examined the prognostic value of

Figure 1. Patient selection algorithm. The number of patientswithin each exclusion criterion is different from that of previousreports because the order in which the exclusion criteria wereapplied was different. ECG = electrocardiogram; MI = myocar-dial infarction; UA = unstable angina; Non-card. = noncardiacetiology. ‘‘outcome’’ refers to composite outcome of MI, revas-cularization, or death as defined under Methods.

742 Chandra et al. • CLINICAL STRATIFICATION ASSOCIATED WITH CARDIAC EVENTS

unstructured EP risk estimation in patients for whomthat estimate would be the critical component to dispo-sition decision-making. We found that patients deemedto have unstable angina, or who were considered to behigh risk, were at an increased risk for major adversecardiac events, including death, MI, or revascularizationwithin 30 days.

Patients with ST-segment deviation on the initialECG or positive cardiac markers are easy to identify,and the correct disposition of such patients is relativelyclear.25–30 The remaining patients are more difficult torisk stratify. Previous studies have demonstrated thatclinical gestalt and ECG alone are insufficient to iden-tify a patient population safe for discharge from theED.10–15,17,18,20,21,23,31,32 Our study likewise found a sig-nificant event rate even in the lowest risk group. Yetthe need to avoid sending any patient with ACS homefrom the ED has led to increasing hospital admissionsand use of vast resources on many patients who ulti-mately do not have a diagnosis of ACS.12,33 This chal-lenge has led to the development of new objectivetools, including new technologies (15 lead ECGs,34,35

body surface mapping,36–40 neural networks,41,42 serumbiomarkers,25,27,28,30,43–48 stress testing,5,33,49,50 com-puted tomography [CT] coronary angiography51,52) andrisk stratification tools11,31,53–55 to aid physicians indecision-making. Although promising, none of these

are as yet widely accepted as a means to identifypatients who can be safely discharged from the ED.

While previous work on the value of physician gestaltis available, it was done in an era when the decisionwas essentially bimodal: admit to the hospital or dis-charge from the ED.11,13,14,31,56,57 One possible reasonfor the more supportive results of this analysis is thatphysicians were allowed to label patients across a spec-trum of risk, which is more consistent with the patho-physiology of disease. With the advent of observationunits, rapid MI rule-out protocols, and CT coronaryangiography, the array of ACS risk-stratificationoptions has increased. Therefore, the focus has shiftedfrom finding patients who are likely safe for discharge,to picking the appropriate diagnostic test for patientswith possible ACS. The ability to identify patients withan increased risk for ACS is important for the optimaluse of resources, especially in an era of overcrowdedhospitals and EDs.58

Where do we go from here and how can we use thisinformation? Recently, several risk stratification strate-gies, the Thrombolysis In Myocardial Infarction(TIMI),53 Platelet Glycoprotein IIb ⁄ IIIa in UnstableAngina: Receptor Suppression Using Integrilin (PUR-SUIT)59 score, and the more recent Global Registry ofAcute Coronary Events (GRACE)60 scoring systemshave been developed from a cohort of patients with

Table 1Characteristics of Excluded and Included Patient Visits

Included patients(n = 10,145*)

Eligible butexcluded (n = 568) p-value

DemographicsAge (yr), mean (SD) 52.4 (15.5) 52.6 (17.4) 0.846Female 5,634 (55.5) 277 (49.5) 0.005White 3,545 (34.9) 273 (48.1) <0.001African American 3,730 (36.8) 161 (28.3)Other race 2,870 (28.3) 134 (23.6)

Medical historyCurrent or recent smoker 3,409 (33.6) 215 (37.9) 0.037Cocaine user 141 (1.4) 17 (3.0) 0.002Amphetamine user 27 (0.3) 3 (0.5) 0.250Chief complaint of chest pain 7,237 (71.3) 389 (68.5) 0.144Diabetes 1,824 (18.0) 91 (16.0) 0.236Hypertension 4,532 (44.7) 210 (37.0) <0.001Dyslipidemia 2,265 (22.3) 105 (18.5) 0.032History of angina 764 (7.5) 61 (10.7) 0.005CAD 1,846 (18.2) 82 (14.4) 0.023CHF 533 (5.3) 33 (5.8) 0.564Prior visit with possible ACS 1,004 (9.9) 24 (4.2) <0.001Family history of heart disease 3,261 (32.1) 217 (38.2) 0.003

DispositionDischarged home 5,249 (51.7) 292 (51.4) <0.000AMA 202 (2.0) 15 (2.6)ICU 653 (6.4) 31 (5.5)OR ⁄ catheterization lab 75 (0.7) 5 (0.9)Transferred 257 (2.5) 5 (0.9)Admitted 3,707 (36.5) 220 (38.7)Expired (cardiac death) 2 (0.0) 0 (0.0)

Composite outcome (%) 522 (5.1) 13 (2.3) 0.002

Percentages are for those with available data. Except where otherwise noted, data are presented as n (%).ACS = acute coronary syndrome; AMA = against medical advice; CAD = coronary artery disease; CHF = congestive heart failure;ECG = electrocardiogram; ICU = intensive care unit; OR = operating room.*Patients without ST changes on ECG or positive cardiac markers who were excluded due to either missing data or missing fol-low-up.


ACS to assist with prognosis prediction. They stratifypatients into low-, intermediate-, and high-risk catego-ries. These scoring systems are used to risk stratifyundifferentiated patients presenting for acute chest painto EDs. The one aspect lacking in these tools is physi-cian gestalt, which we report here as having prognosticvalue. Perhaps one strategy to incorporate physiciangestalt in risk stratification could parallel the strategyencouraged with the Pneumonia Severity Index tool.For example, the EP would use an objective tool tostratify a patient presenting with potential ACS andthen add his ⁄ her gestalt and determine a final posttest

probability of ACS. This model has not been studied orreported previously, but would seem to be reasonableas we use the same strategy in the risk stratification ofpatients with pneumonia and in the evaluation ofpatients for venous thrombus embolism.

With the increasing use of technology and a height-ened awareness of the costs of health care, it is bothprovocative and encouraging that clinical gestalt stillplays a role in the risk stratification of ACS. Indeed,unstructured physician estimate of risk has been foundto be one of the most important parts of structureddecision tools for other conditions such as pulmonary

Table 2Characteristics of Included Patients

Unstable Angina(n = 604)

High Risk(n = 2,383)

Low Risk(n = 4,487)

Noncardiac(n = 2,671)

DemographicsAge (yr), mean (SD) 61.0 (13.1) 58.5 (13.9) 51.3 (14.8) 47.0 (15.8)

Female 276 (45.7) 1,206 (50.6) 2,575 (57.4) 1,577 (59.0)White 212 (35.1) 934 (39.2) 1,562 (34.8) 837 (31.3)African American 150 (24.8) 787 (33.0) 1,526 (34.0) 1,267 (47.4)Other race 242 (40.1) 662 (27.8) 1,399 (31.2) 567 (21.2)

Medical historyCurrent or recent smoker 162 (26.8) 814 (34.2) 1,508 (33.6) 925 (34.6)Cocaine user 2 (0.3) 43 (1.8) 69 (1.5) 27 (1.0)Amphetamine user 0 (0.0) 2 (0.1) 18 (0.4) 7 (0.3)hief complaint of chest pain 491 (81.3) 1,802 (75.6) 3,287 (73.3) 1,657 (62.0)Diabetes 189 (31.3) 637 (26.7) 683 (15.2) 315 (11.8)Hypertension 378 (62.6) 1,411 (59.2) 1,870 (41.7) 873 (32.7)Dyslipidemia 252 (41.7) 805 (33.8) 862 (19.2) 346 (13.0)Angina 139 (23.0) 323 (13.6) 226 (5.0) 76 (2.8)CAD 327 (54.1) 862 (36.2) 464 (10.3) 193 (7.2)CHF 50 (8.3) 217 (9.1) 189 (4.2) 77 (2.9)Prior ED visit with possible ACS 84 (13.9) 295 (12.4) 380 (8.5) 245 (9.2)Family history of heart disease 189 (31.3) 927 (38.9) 1,446 (32.2) 699 (26.2)

DispositionDischarged home 43 (7.1) 529 (22.2) 2,638 (58.8) 2,039 (76.3)AMA 10 (1.7) 44 (1.8) 109 (2.4) 39 (1.5)ICU 118 (19.5) 315 (13.2) 170 (3.8) 50 (1.9)OR 0 (0.0) 3 (0.1) 2 (0.0) 2 (0.1)Catheterization lab 16 (2.6) 34 (1.4) 17 (0.4) 1 (0.0)Floor (monitor) 268 (44.4) 1,162 (48.8) 1,254 (27.9) 369 (13.8)Floor (no monitor) 127 (21.0) 210 (8.8) 187 (4.2) 130 (4.9)Transferred 22 (3.6) 84 (3.5) 110 (2.5) 41 (1.5)Cardiac death 0 (0.0) 2 (0.1) 0 (0.0) 0 (0.0)

Except where otherwise noted, data are presented as n (%).ACS = acute coronary syndrome; AMA = against medical advice; CAD = coronary artery disease; CHF = congestive heart failure;ICU = intensive care unit; OR = operating room.

Table 3Outcomes by Physician Estimates of Risk

OutcomeUnstable Angina

(n = 604)High Risk(n = 2,383)

Low Risk(n = 4,487)

Noncardiac(n = 2,671)

Total(n = 10,145)

Total revascularization 115 (19.0) 190 (8.0) 61 (1.4) 17 (0.6) 383 (3.8)Total MI 24 (4.0) 52 (2.2) 20 (0.4) 13 (0.5) 109 (1.1)All-cause death 5 (0.8) 23 (1.0) 25 (0.6) 24 (0.9) 77 (0.8)Composite outcome* 133 (22.0) 243 (10.2) 97 (2.2) 49 (1.8) 522 (5.1)

Data are presented as n (%).MI = myocardial infarction.*Numbers do not total to 100 because subjects could have had more than one event.


embolism.61 New technologies and decision supporttools for the diagnosis of ACS should support, not sup-plant, physician judgment.

Overall, the patients discharged in our cohort areyounger than admitted patients (Table 4). No differencein sex was observed. A greater proportion of admittedpatients were white, while the greatest proportion ofpatients discharged were nonwhite, non–African Amer-ican patients. Patients with events were observed to beolder and male. White patients had the highest rates ofevents. As the total number of events is small, we didnot perform statistical analysis on this subset ofpatients. The treating physician’s risk assessment influ-ences disposition decisions; patients assessed as highrisk should have a significant rate of admission. Yet, ofpatients diagnosed as having unstable angina or at highrisk for ACS, 19.1% were discharged (Table 2). Despitethe markedly higher discharge rate in the low-risk andnoncardiac groups, adverse events were similar com-pared to the unstable angina ⁄ high-risk cohort (0.4% vs.0.9%; Table 4). It is difficult to argue that these repre-sent alarmingly high rates of inappropriate discharges.Instead, it highlights once again the challenge of riskstratification and illustrates the cost of admitting everypatient with possible ACS.

Many factors influence disposition decisions. It isunclear, and this trial was not designed to define, whichfactors influenced the discharge disposition in the onein five patients considered to have unstable angina orhigh-risk ACS symptoms, but this does need to beinvestigated. Future work could reproduce our workand be directly focused on factors affecting physicianrisk assessment, such as level of experience or training.New technologies and decision support tools for thediagnosis of ACS should support, not supplant, physi-cian judgment.

LIMITATIONS

Analysis of a convenience cohort has inherent limita-tions, including inclusion bias. With multiple centersenrolling patients, it is expected that variability exists inthe enrollment of chest pain patients, especially thosewith perceived angina equivalents. Thus, it is uncertain

if patients with anginal equivalent were equally repre-sented in the data, although the enrollment of a rela-tively large number of patients at multiple differentsites may mitigate this effect.

There is also workup bias present, because patientsestimated to be low risk or to have a noncardiac etiol-ogy are less likely to be admitted or have other riskstratification tests. This may affect the rate of compositeendpoints, particularly revascularization. Indeed, in thisanalysis there is a notable difference in the rates of per-cutaneous interventions among risk groups (Table 3).This is unavoidable in a noninterventional trial. We aremore likely to miss events that occurred at other hospi-tals following the index visit in the low-risk and noncar-diac groups, since those patients were more likely to bedischarged. However, our low lost to follow-up ratemitigates the concern for missing events.

Next, we did not standardize which treating physi-cian performed the estimate of ACS risk. At times, thisestimate was performed by the estimates of residentphysicians, some of whom were not trained in emer-gency medicine. We do not have the data to analyzewhether or how physician training, background, oryears of experience might affect our results. Includingthe estimates of resident physicians, who are presum-ably less skilled in estimating ACS risk, only strength-ens the conclusion that unstructured clinical estimatehas some prognostic power.

Additionally, no formal definitions were provided todifferentiate the risk categories. Thus, some interratervariability may exist in classifying patients. This againreflects the bedside day-to-day challenges that EPs face.

In this registry, risk stratification was to be per-formed prior to biomarker results. However, this maynot have occurred in every instance. This could affectthe accuracy of some estimates. Furthermore, onecould argue that because the biomarkers were notavailable to the EP making the estimate, we should nothave excluded patients with positive biomarkers. How-ever, we wanted to select a population for whom thephysician estimate of risk was the key component ofdisposition decision making. Similar to ST-segmentdeviation on ECG, positive biomarkers have beendemonstrated to correlate with adverse cardiac events.

Table 4Characteristics of Patients by Disposition and Stratified by Event

Not Discharged Home Discharged Home

No Event (n = 4,398) Event (n = 498) No Event (n = 5,225) Event (n = 24)

DemographicsAge (yr), mean (SD) 57.0 (14.9) 60.9 (12.9) 47.8 (14.6) 59.5 (16.1)Female 2,485 (56.5) 186 (37.3) 2,954 (56.5) 9 (37.5)White 1,657 (37.7) 211 (42.4) 1,668 (31.9) 9 (37.5)African American 1,722 (39.2) 97 (19.5) 1,903 (36.4) 8 (33.3)Other 1,019 (23.2) 190 (38.2) 1,654 (31.7) 7 (29.2)

Risk categoryUnstable angina 428 (9.7) 133 (26.7) 43 (0.8) 0 (0.0)High risk 1,616 (36.7) 238 (47.8) 524 (10.0) 5 (20.8)Low risk 1,763 (40.1) 86 (17.3) 2,627 (50.3) 11 (45.8)Noncardiac 591 (13.4) 41 (8.2) 2,031 (38.9) 8 (33.3)

Data are presented as n (%).


Thus, the proper disposition of patients with positivebiomarkers is clear. However, an initial negative set ofbiomarkers does not effectively rule out coronary dis-ease, leaving the EP with only his or her clinical esti-mate of risk as a guide for the proper disposition of thepatient. We feel that by excluding patients with positivebiomarkers, this study more accurately reflects thedilemma most often faced by clinicians. Additionally,since the collection of this data, cardiac biomarkershave been improved, and some of the patients withnormal biomarkers then may have had abnormal bio-markers now. However, the struggle of risk stratifica-tion of patients with normal ECG and new generationcardiac biomarkers stays the same.

Finally, it should be noted that even in the lowest riskgroup, there was still a significant event rate. This isconsistent with previous findings. Therefore, our find-ings do not suggest that clinical estimate identifies apatient population safe for discharge from the ED.

CONCLUSIONS

Emergency physician unstructured diagnostic impres-sion of high risk correlates with major adverse cardiacevents in patients with nondiagnostic ECG and normalinitial cardiac biomarkers. Clinical treatment paradigmsshould support the treating physician’s judgment indetermining disposition of these patients.

References

1. Warden G. The Future of Emergency Care in theUnited States Health System. Washington, DC:National Academies Press, 2006.

2. McCaig L, Burt C. National hospital ambulatorymedical care survey: 2003 emergency departmentsummary. Adv Data. 2005; 358:1–38.

3. American Heart Association. Heart Disease andStroke Statistics-2006 Update. Dallas, TX: AmericanHeart Association, 2006, p 46.

4. Rusnak RA, Stair TO, Hansen K, Fastow JS. Litiga-tion against the emergency physician: common fea-tures in cases of missed myocardial infarction. AnnEmerg Med. 1989; 18:1029–34.

5. American College of Emergency Physicians ClinicalPolicies Committee; Clinical Policies CommitteeSubcommittee on Suspected Pulmonary Embolism.Clinical policy: critical issues in the evaluation andmanagement of adult patients presenting with sus-pected pulmonary embolism. Ann Emerg Med.2003; 41:257–70.

6. Lindsell CJ, Anantharaman V, Diercks D, et al. TheInternet Tracking Registry of Acute Coronary Syn-dromes (i*trACS): a multicenter registry of patientswith suspicion of acute coronary syndromesreported using the standardized reporting guide-lines for emergency department chest pain studies.Ann Emerg Med. 2006; 48:666–77.

7. American College of Emergency Physicians. Clinicalpolicy for management of adult patients presentingwith a chief complaint of chest pain, with no historyof trauma. Dallas TX: American College of Emer-gency Physicians, 1990.

8. American College of Emergency Physicians. Clini-cal policy for the initial approach to adults pre-senting with a chief complaint of chest pain, withno history of trauma. Ann Emerg Med. 1995;25:274–99.

9. American College of Emergency Physicians. Clinicalpolicy: critical issues in the evaluation and manage-ment of adult patients presenting with suspectedAMI or unstable angina. Ann Emerg Med. 2000;35:521–44.

10. Blomkalns AL, Lindsell CJ, Chandra A, et al. Canelectrocardiographic criteria predict adverse car-diac events and positive cardiac markers? AcadEmerg Med. 2003; 10:205–10.

11. Goldman L, Cook EF, Brand DA, et al. A computerprotocol to predict myocardial infarction in emer-gency department patients with chest pain. N EnglJ Med. 1988; 318:797–803.

12. Karlson BW, Herlitz J, Wiklund O, Richter A,Hjalmarson A. Early prediction of acute myocardialinfarction from clinical history, examination andelectrocardiogram in the emergency room. Am JCardiol. 1991; 68:171–5.

13. Lee TH, Cook EF, Weisberg M, Sargent RK, WilsonC, Goldman L. Acute chest pain in the emergencyroom. Identification and examination of low-riskpatients. Arch Intern Med. 1985; 145:65–9.

14. Lee TH, Rouan GW, Weisberg MC, et al. Sensitivityof routine clinical criteria for diagnosing myocardialinfarction within 24 hours of hospitalization. AnnIntern Med. 1987; 106:181–6.

15. Limkakeng A Jr, Gibler WB, Pollack C, et al. Com-bination of Goldman risk and initial cardiac tropo-nin I for emergency department chest pain patientrisk stratification. Acad Emerg Med. 2001; 8:696–702.

16. Ornato JP, Selker HP, Zalenski RJ. Overview: diag-nosing acute cardiac ischemia in the emergencydepartment. A report from the National HeartAttack Alert Program. Ann Emerg Med. 2001;37:450–2.

17. Swap CJ, Nagurney JT. Value and limitations ofchest pain history in the evaluation of patients withsuspected acute coronary syndromes. JAMA. 2005;294:2623–9.

18. Zalenski RJ, Shamsa FH. Diagnostic testing of theemergency department patient with chest pain.Curr Opin Cardiol. 1998; 13:248–53.

19. Goldman L, Kirtane AJ. Triage of patients withacute chest pain and possible cardiac ischemia: theelusive search for diagnostic perfection. Ann InternMed. 2003; 139:987–95.

20. Goodacre S, Locker T, Morris F, Campbell S. Howuseful are clinical features in the diagnosis of acute,undifferentiated chest pain? Acad Emerg Med.2002; 9:203–8.

21. Wilkinson K, Severance H. Identification of chestpain patients appropriate for an emergency depart-ment observation unit. Emerg Med Clin North Am.2001; 19:35–66.

22. Hollander JE, Blomkalns AL, Brogan GX, et al.Standardized reporting guidelines for studies evalu-ating risk stratification of emergency department


patients with potential acute coronary syndromes.Ann Emerg Med. 2004; 44:589–98.

23. Miller CD, Lindsell CJ, Khandelwal S, et al. Is theinitial diagnostic impression of ‘‘noncardiac chestpain’’ adequate to exclude cardiac disease? AnnEmerg Med. 2004; 44:565–74.

24. Nawar EW, Niska RW, Jianmin X. National HospitalAmbulatory Medical Care Survey: 2005 emergencydepartment summary. Adv Data. 2007; 386:1–32.

25. Apple FS, Quist HE, Murakami MM. Diagnostic andprognostic value of cardiac troponin I assays inpatients admitted with symptoms suggestive ofacute coronary syndrome. Arch Pathol Lab Med.2004; 128:430–4.

26. Apple FS, Wu AH, Jaffe AS. European Society ofCardiology and American College of Cardiologyguidelines for redefinition of myocardial infarction:how to use existing assays clinically and for clinicaltrials. Am Heart J. 2002; 144:981–6.

27. Christenson RH, Duh SH, Newby LK, et al. Cardiactroponin T and cardiac troponin I: relative values inshort-term risk stratification of patients with acutecoronary syndromes. GUSTO-IIa Investigators. ClinChem. 1998; 44:494–501.

28. Hamm CW. Troponin T: a new marker for myocar-dial cell injury. Ann Med. 1994; 26:319–20.

29. Heeschen C, Goldmann BU, Terres W, Hamm CW.Cardiovascular risk and therapeutic benefit of coro-nary interventions for patients with unstable anginaaccording to the troponin T status. Eur Heart J.2000; 21:1159–66.

30. Ohman EM, Armstrong PW, Christenson RH, et al.Cardiac troponin T levels for risk stratification inacute myocardial ischemia. GUSTO IIA Investiga-tors. N Engl J Med. 1996; 335:1333–41.

31. Pozen MW, D’Agostino RB, Selker HP, SytkowskiPA, Hood WB Jr. A predictive instrument toimprove coronary-care-unit admission practices inacute ischemic heart disease. A prospective multi-center clinical trial. N Engl J Med. 1984; 310:1273–8.

32. Selker HP, Zalenski RJ, Antman EM, et al. An eval-uation of technologies for identifying acute cardiacischemia in the emergency department: executivesummary of a National Heart Attack Alert ProgramWorking Group Report. Ann Emerg Med. 1997;29:1–12.

33. Zalenski RJ, Grzybowski M. The chest pain centerin the emergency department. Emerg Med ClinNorth Am. 2001; 19:469–81.

34. Zalenski RJ, Cooke D, Rydman R, Sloan EP, Mur-phy DG. Assessing the diagnostic value of an ECGcontaining leads V4R, V8, and V9: the 15-lead ECG.Ann Emerg Med. 1993; 22:786–93.

35. Zalenski RJ, Rydman RJ, Sloan EP, et al. Value ofposterior and right ventricular leads in comparisonto the standard 12-lead electrocardiogram in evalua-tion of ST-segment elevation in suspected acute myo-cardial infarction. Am J Cardiol. 1997; 79:1579–85.

36. Carley SD, Jenkins M, Mackway Jones K. Body sur-face mapping versus the standard 12 lead ECG inthe detection of myocardial infarction amongstemergency department patients: a Bayesianapproach. Resuscitation. 2005; 64:309–14.

37. Drew BJ, Schindler DM, Zegre JK, Fleischmann KE,Lux RL. Estimated body surface potential maps inemergency department patients with unrecognizedtransient myocardial ischemia. J Electrocardiol.2007; 40(6 Suppl):S15–20.

38. Lefebvre C, Hoekstra J. Early detection and diagno-sis of acute myocardial infarction: the potential forimproved care with next-generation, user-friendlyelectrocardiographic body surface mapping. Am JEmerg Med. 2007; 25:1063–72.

39. Self WH, Mattu A, Martin M, Holstege C, Preuss J,Brady WJ. Body surface mapping in the ED evalua-tion of the patient with chest pain: use of the 80-lead electrocardiogram system. Am J Emerg Med.2006; 24:87–112.

40. Shusterman V, Goldberg A, Schindler DM, Fleisch-mann KE, Lux RL, Drew BJ. Dynamic tracking ofischemia in the surface electrocardiogram. J Elec-trocardiol. 2007; 40(6 Suppl):S179–86.

41. Baxt WG, Shofer FS, Sites FD, Hollander JE. A neu-ral network aid for the early diagnosis of cardiacischemia in patients presenting to the emergencydepartment with chest pain. Ann Emerg Med. 2002;40:575–83.

42. Baxt WG, Shofer FS, Sites FD, Hollander JE. A neuralcomputational aid to the diagnosis of acute myo-cardial infarction. Ann Emerg Med. 2002; 39:366–73.

43. Fesmire FM. Delta CK-MB outperforms delta tropo-nin I at 2 hours during the ED rule out of acute myo-cardial infarction. Am J Emerg Med. 2000; 18:1–8.

44. Fesmire FM. Improved identification of acute coro-nary syndromes with delta cardiac serum markermeasurements during the emergency departmentevaluation of chest pain patients. Cardiovasc Toxi-col. 2001; 1:117–23.

45. McCord J, Nowak RM, McCullough PA, et al.Ninety-minute exclusion of acute myocardial infarc-tion by use of quantitative point-of-care testing ofmyoglobin and troponin I. Circulation. 2001; 104:1483–8.

46. Mitchell AM, Garvey JL, Kline JA. Multimarkerpanel to rule out acute coronary syndromes in low-risk patients. Acad Emerg Med. 2006; 13:803–6.

47. Peacock F, Morris DL, Anwaruddin S, et al. Meta-analysis of ischemia-modified albumin to rule outacute coronary syndromes in the emergencydepartment. Am Heart J. 2006; 152:253–62.

48. Wu AH. Markers for early detection of cardiac dis-eases. Scand J Clin Lab Invest Suppl. 2005; 240:112–21.

49. Chandra A, Rudraiah L, Zalenski RJ. Stress testingfor risk stratification of patients with low to moder-ate probability of acute cardiac ischemia. EmergMed Clin North Am. 2001; 19:87–103.

50. Kontos MC, Fratkin MJ, Jesse RL, Anderson FP,Ornato JP, Tatum JL. Sensitivity of acute rest myo-cardial perfusion imaging for identifying patientswith myocardial infarction based on a troponin def-inition. J Nucl Cardiol. 2004; 11:12–9.

51. Chase M, Brown AM, Robey JL, et al. Clinicalimplementation of computed tomography in emer-gency department patients with low risk chest pain[abstract]. Acad Emerg Med. 2006; 13:s103–4.


52. Hollander JE, Chang AM, Shofer FS, McCuskerCM, Baxt WG, Litt HI. Coronary computed tomo-graphic angiography for rapid discharge of low-riskpatients with potential acute coronary syndromes.Ann Emerg Med. 2009; 53:295–304.

53. Antman EM, Cohen M, Bernink PJLM, et al. TheTIMI Risk Score for unstable angina ⁄ non-ST Eleva-tion MI: a method for prognostication and thera-peutic decision making. JAMA. 2000; 284:835–42.

54. Christenson J, Innes G, McKnight D, et al. A clinicalprediction rule for early discharge of patients withchest pain. Ann Emerg Med. 2006; 47:1–10.

55. Fesmire FM. A rapid protocol to identify andexclude acute myocardial infarction: continuous 12-lead ECG monitoring with 2-hour delta CK-MB.Am J Emerg Med. 2000; 18:698–702.

56. Goldman L, Cook EF, Johnson PA, Brand DA, Rou-an GW, Lee TH. Prediction of the need for intensivecare in patients who come to the emergencydepartments with acute chest pain. N Engl J Med.1996; 334:1498–504.

57. Lee KL, Pryor DB, Harrell FE Jr, et al. Predictingoutcome in coronary disease. Statistical models ver-sus expert clinicians. Am J Med. 1986; 80:553–60.

58. Diercks DB, Roe MT, Chen AY, et al. Prolongedemergency department stays of non ST-segment-elevation myocardial infarction patients are associ-ated with worse adherence to the American Collegeof Cardiology ⁄ American Heart Association guide-lines for management and increased adverse events.Ann Emerg Med. 2007; 50:489–96.

59. Boersma E, Pieper KS, Steyerberg EW, et al. forthe PURSUIT Investigators. Predictors of outcomein patients with acute coronary syndromes withoutpersistent ST-segment elevation. Results from aninternational trial of 9461 patients. Circulation.2000; 101:2557–67.

60. Granger CB, Goldberg RJ, Dabbous OH, et al. forthe Global Registry of Acute Coronary EventsInvestigators. Predictors of hospital mortality in theglobal registry of acute coronary events. ArchIntern Med. 2003; 163:2345–53.

61. Runyon MS, Webb WB, Jones AE, Kline JA. Com-parison of the unstructured clinician estimate ofpretest probability for pulmonary embolism to theCanadian score and the Charlotte rule: a prospec-tive observational study. Acad Emerg Med. 2005;12:587–93.


emergency physician high pretest probability for acute coronary syndrome correlates with adverse...

Documents