American Journal of Emergency Medicine (2012) 30, 61-67

Original Contribution

Predicting unfavorable outcome in subjects with diagnosis of chest pain of undifferentiated origin^?,??,?

Andrea Fabbri MD ^a,?, Filippo Ottani MD ^b, Giulio Marchesini ^c,

Marcello Galvani MD ^b, Alberto Vandelli MD ^a

^aDipartimento dell‘Emergenza, Presidio Ospedaliero Morgagni-Pierantoni, Azienda Unita Sanitaria Locale di Forli,

I – 47100 Forli, Italy

^bUnita Operativa di Cardiologia, Presidio Ospedaliero Morgagni-Pierantoni, Azienda Unita Sanitaria Locale di Forli,

I – 47100 Forli, Italy

^cAlma Mater Studiorum, Universita di Bologna, I – 40138 Bologna, Italy

Received 26 July 2010; revised 10 September 2010; accepted 14 September 2010

Abstract

Background: Subjects with chest pain and a negative diagnostic workup constitute a problem for emergency physicians. We tested the usefulness of clinical variables in predicting 30-day and 6-month outcome in subjects with chest pain of undifferentiated origin after a negative workup.

Methods: Chest pain of undifferentiated origin was diagnosed by negative first-line (serial electrocardiograms, troponins assays, and 12- to 24-hour observation) and second-line evaluation (echocardiography, exercise tolerance test, stress scintigraphy, stress echocardiography, coronary angiography). Thirty-day and 6-month outcomes were considered unfavorable in the presence of any of the following: death, acute coronary syndrome, need for urgent coronary revascularization. The variables considered for risk stratification were age, sex, smoking, family history of coronary artery disease, presence of hypertension, high cholesterol levels, diabetes, chronic renal failure, cerebral vascular disease, and history of acute coronary syndrome, percutaneous transluminal angioplasty (PTA), coronary artery by pass graft, and heart failure.

Findings: Five items (diabetes, chronic renal failure, history of PTA or bypass, history of heart failure) were associated with 30-day unfavorable outcome (31 events/1262 cases; 2.5%). The receiver operating characteristic area of the selected items was 0.726 (95% confidence interval [CI], 0.654-0.798); sensitivity was 90.3% (73.1-95.8) and specificity was 54.8% (52.0-57.6). A similar panel of items (older age, diabetes, chronic renal failure, history of PTA) predicted an unfavorable 6-month outcome (90 subjects [7.1%], with lower accuracy (receiver operating characteristic area, 0.610 [95% CI, 0.594-

0.627, P b .05]; sensitivity, 98.9% [95% CI, 93.1-99.6]; specificity, 21.6% [95% CI, 19.4-23.9]).

^? Contributors: AF conceived the study, wrote the protocol, coordinated the data collection, contributed to interpretation of the results, and wrote the paper. OF contributed to interpretation of the results and critical review of the paper. GM contributed to study design, interpretation of the results, and co-wrote the paper. MG contributed to interpretation of the results and critical review of the paper and co-wrote the paper. AV contributed to study design, interpretation of the

results, and critical review of the paper. All authors approved the final version of the paper.

^?? Conflict of interest: All authors warrant to have no conflict of interest in connection with the article; they have access to all data in the study and they held final responsibility for the decision to submit for publication.

^? Funding. None.

* Corresponding author. Tel.: +39 0543 735801; fax: +39 0543 735881.

E-mail address: [email protected] (A. Fabbri).

0735-6757/$ – see front matter (C) 2012 doi:10.1016/j.ajem.2010.09.022

Interpretation: In subjects with chest pain of undifferentiated origin, the risk of unfavorable outcome cannot be accurately predicted by the selected clinical items.

(C) 2012

Introduction

Subjects attending the emergency department (ED) for chest pain without electrocardiographic signs of acute coronary syndrome (ACS) and a negative troponin assay are a special challenge for ED physicians [1]. The increased attention to emergency diagnostic workup has placed physicians under pressure for the cost of noncardiac chest pain management, to limit hospital admission only to high-risk subjects [2,3]. On the other hand, an increased tendency to defensive medicine limits their propensity to rapidly discharge subjects with chest pain of undifferentiated origin (CPUO), low-risk profile, and a negative diagnostic program [4,5]. The clinical risk profile is often the only tool for a final evaluation and decision making when both first-line (serial ECG and troponin assay) as well as second-line (echocardiography, exercise tolerance test [ETT], stress scintigraphy, stress echocardiography, or coronary angiography) assessment are negative [6].

This second-line evaluation is also subject to limitations because a relatively high proportion of subjects are not able to exercise, and the tests may produce false-positive or inconclusive results. In addition, cost, time, and space constraints may further limit the availability of additional Diagnostic procedures and their repeatability in individual subjects: in this case, an a priori stratification of the risk of coronary disease is of paramount importance.

The clinical characteristics of acute chest pain are per se of limited importance in triage decision making in subjects with nondiagnostic electrocardiogram [7]. The TIMI risk score was shown to improve the risk stratification of patients with suspected ACS [8] and as “front door” score in subjects presenting to the ED with Undifferentiated chest pain [9], but its usefulness in subjects with negative diagnostic workup has never been demonstrated.

In a retrospective study, only 4.4% of subjects discharged with a diagnosis of chest pain of undetermined origin from 1985 to 1992 showed adverse events within 12 months. The factors associated with adverse cardiac events were Abnormal ECG at admission, preexisting diabetes mellitus, and preexisting coronary artery disease [10,11].

We reviewed the records included in our large observa- tional database of subjects enrolled in a chest pain diagnostic program. The purpose of the study was to test the discriminating ability of the clinical items reported at history or the risk factors recorded during the initial evaluation in predicting 30-day and 6-month unfavorable outcomes in subjects discharged from the ED with a diagnosis of CPUO after negative chest pain diagnostic workup.

Materials and methods

Chest pain program

The ED of Forli, Italy, is the sole ED receiving all cardiac emergencies from an area of 1380 km². The ED is part of a 450 acute-bed general hospital providing general clinical services (except cardiac surgery and neurosur- gery). The ED treats nearly 50 000 children and adults per year from a population of 171 000 inhabitants, with 58% of the population living in an urban environment. The hospital admits more than 20 000 subjects, with 900 admissions to intensive care. The hospital database is directly connected with the General Registry Office of the District. The model of care of the chest pain program includes a short-stay multidisciplinary observation unit (26 beds) located inside the ED, caring for approximately 4000 subjects per year.

Approximately 1500 subjects per year with chest pain attend the ED, where they enter a diagnostic and therapeutic workup, as outlined in Fig. 1. The present report specifically addresses the subpopulation with chest pain and negative first-line and second-line evaluation tests, and exclusion of subjects with gastrointestinal, muscular, and anxiety dis- orders. For the purpose of the present report, all subjects revisited and discharged during the study period with a second diagnosis of CPUO were considered only once.

The clinical assessment of our chest pain population comprises the evaluation of risk factors for coronary disease including age, sex, family history of coronary artery disease (CAD), the presence of Arterial hypertension, high cholesterol levels, diabetes mellitus, smoking habits, chronic renal failure (CRF), cerebral vascular disease (CVD), history of ACS, percutaneous transluminal angio- plasty (PTA), coronary artery bypass graft (CABG), and heart failure .

During the whole study period, we followed the European guidelines on definitions and management of patients with chest pain and suspected ACS, published in 2002 [12] and revised in 2007 [13].

On the basis of history and clinical evaluation, all subjects were treated according to different diagnostic programs in relation to their clinical risk profile after negative ECG findings and serial troponin assays. Low-risk subjects (b3 coronary low-risk factors [older age [14], smoking, hyper- tension, cholesterol, family history of CAD, CRF, history of CVD]) were managed according to a short-term observation program (12 hours) and a second-line evaluation test (ETT or 99m Tc myocardial stress scintigraphy or stress echocardi- ography or coronary angiography) [15-17].

Fig. 1 Flow chart of the diagnostic workup of subjects attending the Emergency Department of Forli for suspected acute coronory syndrome. The analyses were performed in subjects with negative first- and second-line evaluation tests.

High-risk subjects (presence of diabetes mellitus, CAD, history of CABG, PTA, or history of HF) were observed for a longer period (24 hours) in a short-term observation unit with serial 12-lead electrocardiography and serial troponin assay (at 0, 6, and 12 hours from entry) and chest radiography. Following a negative observation program, the indication to extend the evaluation to second-line tests was based on the assessment of cardiologists who were also in charge of a final evaluation before discharge.

After completion of a negative observation period and a negative extended diagnostic workup, all patients received recommendations for immediate referral to the ED for further evaluation in case of complaints. A 6-month appointment in the cardiology department was booked at time of discharge.

Outcome measures

The occurrence of any of the following events at 30-day and 6-month follow-up was considered as unfavorable outcome: death of Cardiac origin, ACS, need for urgent coronary revascularization (PTA, CABG). The ED of Forli is the sole hospital where subjects with cardiac events are referred to in the district area. Accordingly, we assumed that patients who were not readmitted either did not have events or died of events (in which case they would be traced in the Health District database). A systematic search of death certificates and the Health District database was periodically done by one of us (AF). All deaths were adjudicated with predefined criteria. Cardiovascular deaths included those with a definite cardiovascular cause and those of unknown

cause, in the absence of necropsy. In most of the cases, in the presence of a sudden death, a necropsy was performed and the cardiac origin was adjudicated or excluded. Deaths within 30 days for noncardiac causes were not considered.

The protocol was carried out according to the Helsinki Declaration and approved by the senior staff committee of the Azienda USL of Forli, an institutional review board regulating noninterventional studies. A consent to the anonymous treatment of personal data was signed by all patients at entry in ED.

Statistical analyses

A descriptive statistics was initially used to identify our population (median and interquartile range (IQR) or percentage for categorical variables]. The following covari- ates were tested as outcome predictors: age, sex, family history of CAD, arterial hypertension, high cholesterol levels (N200 mg/dL), diabetes mellitus, smoking habits, chronic renal failure, CVD, history of ACS, history of PTA, CABG, and HF.

Multivariable logistic regression analysis was used to test the predictive value of the variables significant at univariable analysis (odds ratios [ORs] and 95% confidence intervals [CIs]). The Hosmer-Lemeshow test was used to evaluate goodness of fit between study variables, whereas collinearity between variables was tested by the variance inflation factor (VIF). The receiver operating characteristics (ROC) curves of individual variables were calculated by reporting the proportion of subjects with all individual items stratified by

outcome. These areas, ranging from 0.5 to 1.0, represent a summary measure of the sensitivity and specificity of the individual variables [18]. The discriminating capacity for a variable was considered good when the ROC area was greater than 0.7. Proportions were compared by means of Fisher exact test. Two-tailed P values of less than .05 were considered statistically significant. A VIF of less than 2.0 was assumed to exclude collinearity. Statistical analyses were performed running the SPSS/PC+ statistical package (XVII Edition; SPSS, Chicago, Ill).

Results

Study population

The study population of the present study consisted of 1262 subjects who were discharged with diagnosis of CPUO after a negative chest pain diagnostic program in the period from October 1, 2003, to June 30, 2006. The analysis included 517 (40.9% of total) subjects with low-risk profile after a 12-hour observation program and 745 high-risk subjects after a negative 24-hour observation program. Their characteristics in relation to outcome are reported in Table

1. They are part of a larger group of 3449 consecutive subjects presenting with acute chest pain of possible coronary origin to the ED of our hospital (see Fig. 1).

Low-risk subject (517 cases; median age, 66 years; IQR, 56-77 years) completed the chest pain program within 12 hours after a negative first-line evaluation, with serum troponin assay obtained at entry and within 6 to 12 hours from admission. A second-line test was performed with negative results in only 239 subjects (46.2%) (ETT, 177

[34.2%]; stress scintigraphy, stress echocardiography, 53

[10.3%]; angiography, 9 [1.7%]).

Table 1 Clinical variables considered (number of cases and %, or median and IQR) in subjects with diagnosis of CPUO after the observation algorithm in relation to unfavorable event at 30 days of follow-up (% in parenthesis)

High-risk subjects (745 cases; median age, 70 years; IQR, 60-78 years) were discharged with a diagnosis of CPUO after 12- to 24-hour observation following negative serial troponin assays, ECG evaluations, and negative second-line tests in 451 subjects (60.5%) (ETT 153 cases [20.5%] and stress

scintigraphy, stress echocardiography, 210 [28.2%], angiog-

raphy 88 cases [11.8%]).

In summary, only 690 (54.7%) of 1262 subjects were discharged with diagnosis of CPUO after a negative second-line test (ETT 330 [26.1%], stress scintigraphy or stress echocardiography 263 [20.8%] coronary angiogra- phy, 97 (7.7%) subjects). In the remaining 572 cases, the diagnosis of CPUO was not substantiated by a negative second-line test.

Unfavorable outcome

A complete follow-up was obtained in 1245 subjects (98.7%). Five subjects died during the follow-up period because of complications unrelated to coronary disease (cancer, 3 cases; hemorrhagic stroke, 1; gastrointestinal hemorrhage, 1), whereas 12 were lost to follow-up.

At 30-day follow-up, the outcome was recorded as unfavorable in 31 subjects (2.5%) (ST elevation myocardial infarction , 14 cases; ACS, 12; need for urgent revascularization, 5). Of note, none of the subjects died within 30 days because of undiagnosed CAD; only 4 were in the group where second-line tests had not been performed during admission.

At 6-month follow-up, an unfavorable outcome was recorded in 90 subjects (7.1%) (cardiac arrest, 4 cases;

STEMI 12; ACS, 36; need for urgent revascularization, 38). Of note, only 5 of these subjects died of an acute cardiac event, and 25 were in the group who missed the second-line diagnostic workup.

Unfavorable Outcome (n = 31)		Favorable Outcome (n = 1262)		OR (95% CI)	LR	P
Age (y, decades)	63 (61-77)	63	(49-75)	–	–	.895
Sex (% males)	21 (67.7)	665	(54.0)	1.79 (0.83-3.83)	2.36	.147
Family history of CAD	4 (12.9)	221	(18.0)	0.68 (0.23-1.95)	0.57	.636
Hypertension	23 (74.2)	768	(62.4)	1.73 (0.77-3.91)	1.90	.194
Cholesterol	16 (51.6%)	495	(40.2)	1.59 (0.78-3.24)	1.60	.266
Diabetes	18 (58.1)	300	(24.1)	4.30 (2.08-8.87)	15.46	b.001
Smoking habits	8 (25.8)	333	(27.1)	0.94 (0.41-2.12)	0.02	1.000
CRF	9 (12.2)	65	(5.3)	7.33 (3.25-16.60)	17.15	b.001
CVD	4 (12.9)	106	(8.6)	1.57 (0.54-4.58)	0.62	.339
History of ACS	12 (38.7)	249	(20.2)	2.49 (1.19-5.20)	5.40	.022
History of PTA	12 (38.7)	174	(14.1)	3.84 (1.83-8.04)	11.00	.001
History of CABG	7 (22.6)	82	(6.7)	4.09 (1.71-9.77)	7.81	.004
History of HF	13 (41.9)	138	(11.2)	5.72 (2.74-11.93)	18.30	b.001
LR indicates likelihood ratio.

outcome prediction“>All cases	OR	95% CI	P
Age (y, decades)	0.78	0.57-1.08	.130
Sex (% males)	1.57	0.69-3.58	.284
Family history of CAD	1.21	0.36-4.03	.758
Hypertension	1.21	0.50-2.91	.670
Cholesterol	1.37	0.62-3.01	.436
Diabetes	2.94	1.30-6.63	.009
Smoking habits	0.80	0.33-1.94	.619
CRF	5.20	2.00-13.55	.001
CVD	0.99	0.31-3.14	.984
History of ACS	1.07	0.44-2.58	.882
History of PTA	2.78	1.21-6.39	.016
History of CABG	3.43	1.21-9.68	.020
History of HF	3.16	1.30-7.65	.011

All cases	OR	95% CI	P
Age (y, decades)	1.57	1.24-1.99	b.001
Sex (% males)	1.61	0.99-2.62	.057
Family history of CAD	1.93	0.92-4.04	.080
Hypertension	1.28	0.76-2.18	.355
Cholesterol	1.12	0.68-1.82	.658
Diabetes	3.20	1.98-5.18	b.001
Smoking habits	0.65	0.35-1.20	.168
Chronic renal failure	0.90	0.41-1.97	.796
CVD	0.76	0.37-1.58	.465
History of ACS	1.22	0.71-2.09	.465
History of PTA	1.83	1.06-3.18	.031
History of CABG	1.85	0.91-3.76	.088
History of HF	3.94	2.32-6.70	b.001

Outcome prediction

Table 2 Multivariate logistic regression analysis: variables predicting a 30-day unfavorable outcome after diagnosis CPUO in relation to a priori risk profile

Table 4 Multivariate logistic analysis: variables predicting an unfavorable outcome at 6 months in relation to clinical profile

Five of the 13 variables considered (diabetes, CRF, history of ACS, history of PTA, and history of CABG) were independently associated with 30-day unfavorable outcome at multivariable stepwise logistic regression analysis (Table 2). The Hosmer-Lemeshow test showed a good fit between study variables used in the regression (P = .941), whereas VIF excluded collinearity in the model.

No item had a ROC area greater than 0.7 in outcome prediction. The most discriminating items were diabetes (ROC area, 0.67; 95% CI, 0.67-0.69) and history of HF (0.65; 95% CI, 0.63-0.68). The sensitivity of these variables was low (58% for diabetes and 41.9% for CRF), with a specificity of 75.3% and 99.8%, respectively.

The remaining items had less discriminating ROC areas (b0.65): CRF (0.62; 95% CI, 0.59-0.65; sensitivity 29.0%,

specificity 94.7%), history of ACS (0.59; 95% CI, 0.56-0.62; sensitivity 38.7%, specificity 79.8%), history of PTA (0.62;

95% CI 0.59-0.65; sensitivity 38.7%, specificity 85.8%),

history of CABG (0.57; 95% CI 0.55-0.60; sensitivity

22.6%, specificity 93.3%).

Table 3 Receiving operating characteristics areas with sensitivity, specificity, and NPV of items selected by the logistic model in the prediction of unfavorable outcome at 30 days

Unfavorable outcome Yes No

Specificity (%) ROC area NPV

54.8 (95% CI, 52.0-57.6)

0.726 (95% CI, 0.654-0.798)

99.6 %

NPV indicates negative predictive value.

Table 5 Receiving operating characteristics areas with sensitivity, specificity, and NPV of items selected by the logistic model in the prediction of unfavorable outcome at 6 months

Unfavorable outcome

Sensitivity (%) Specificity (%) ROC area NPV

98.9 (95% CI, 93.1-99.6)

21.6 (95% CI, 19.4-23.9)

0.610 (95% CI, 0.594-0.627)

99.6%

Table 3 shows the composite discriminating ROC area of the 5 variables selected by multivariate analysis for the 30-day prediction of unfavorable outcome. Only 3 subjects with unfavorable outcome (10% of subjects with unfavorable outcome, corresponding to b0.3% of total population) would remain undetermined by the selected variables.

At 6-month follow-up, nearly the same variables with the exception of CVD and older age were predictive of an unfavorable outcome (Table 4, goodness of fit by Hosmer- Lemeshow test, P = .073, collinearity excluded by VIF). The selected items had a lower accuracy of the ROC curve (6-month vs. 30-day follow-up), without differences in sensitivity (98.9%; 95% CI, 93.1-99.6) and negative predictive value (Table 5).

At 6-month follow-up, only 1 subject (1.1%) with unfavorable outcome would remain undetermined by any of the selected variables (b0.1% of the whole cohort).

Discussion

Our data show that subjects discharged with a diagnosis of CPUO have a chance of 2.5% at 30 days and 7.1% at

Variables present	28	556		Yes	No
Variables absent	3	675	Variables present	89	900
Sensitivity (%)	90.3 (95% CI,	73.1-95.8)	Variables absent	1	272

6 months follow-up of unfavorable outcome after a negative chest pain diagnostic program. A group of selected clinical items predicts the unfavorable outcome, but their discrim- inating ability remains fair for clinical decision-making.

Chest pain programs are based on the common belief that physician judgment alone is inadequate in distinguishing subjects with/without ACS [11,19,20]. A policy aimed at early diagnosis is likely to reduce admissions, health service costs, subject anxiety, and depression, as well as to improve outcome, quality of life, and satisfaction [2].

Any attempt to predict the outcome of CPUO subjects must consider several points. Firstly, the final prevalence of CPUO in our setting was 35.0%. Other studies conclude that a larger proportion (typically around 50%) of subjects with chest pain do not have ACS [21]. This means that any diagnostic program must be tailored on the clinical characteristics of the study population.

Secondly, in low-risk subjects the chance of ACS in subjects with CPUO after a predefined diagnostic program is very low [19,22]. In these subjects, the overuse of costly resources would carry a risk of inappropriate invasive procedures and serious morbidity [23], not to say false- positive results. In this setting, a point estimate pretest probability below the test threshold was proposed to exclude very-low-risk subjects from observation programs [24], thus reducing the overtesting rate. In addition, when the pretest probability of ACS is 2% or less, Mitchell et al [24] concluded that an observation program had to be warranted. These data imply that physicians either do not trust their own clinical judgment or are unaware of any appropriate or well- established test threshold [25]. According to the first hypothesis, subjects followed up for 6 months after a negative chest pain program had a non negligible rate of AMI or cardiac death (4.8% of subjects), the risk of major events being stratified by simple data available at presenta- tion [6]. Clinical data such as pain characteristics, diabetes, previous coronary surgery, and ST-segment depression allowed risk stratification; the greater the number of these factors, the higher the risk of major events. In a comparison of 3 widely used Cardiovascular risk stratification tools, Manini et al [26] showed a sensitivity far below the standard of 90% to predict ACS in different settings. We identified several variables associated with 30-day unfavorable out- come; the presence of at least one of them predicted outcome with a sensitivity greater than 90% and a nearly 100% negative predictive value.

Thirdly, new diagnostic tools are progressively added to increase the diagnostic and predictive accuracy. The N- terminal pro-brain natriuretic peptide assay showed incre- mental prognostic information above that given by the clinical history and early exercise testing [27]. By ROC curves, the assay significantly improved the accuracy in the prediction of death or acute myocardial infarction from 0.76 to 0.84 (P = .01), but not without effect on health costs.

Fourthly, risk factors may be different in relation to the observation period. In a 1-year follow-up, 4.4% of cases had

an event, and specifically subjects with preexisting coronary disease or diabetes [10]. In the absence of such factors, cardiac outcome is excellent.

Fifthly, second-line evaluation tests cannot be per- formed and/or correctly evaluated in all subjects. In 21% of subjects, second-line cardiac testing was not diagnostic because of inability to exercise or technical problems. In a recent postal questionnaire, less than 16% of subjects with negative ECG and troponin assay had an exercise stress test before discharge from hospital [28]. In our study, 67% (844/1262) of subjects with negative first-line evaluation program performed a second-line test, 476 (56%) with ETT completed. Notably, subjects with a history of CAD resulted more prone to cardiac testing than subjects without (67% vs 31%; P b .001), but the performance of second-line tests was not associated with an adverse event at 30 days and at 6 months.

Finally, subjects with CPUO and a recent negative chest pain workup who re-attend for chest pain represent a particular challenge. When history of cardiac disease, diabetes, or both is present, the suspicion of ACS remains high, and observation with serial ECG and cardiac markers is mandatory. However, at 6 months, the Death rate for cardiac events resulted only 0.4%, much lower than that reported in 12-month follow-up studies (3.5% and 1.7%) [10,27]. Also, in our series, the final event rate is increased by the presence of diabetes, responsible for most of the long-term events. By excluding these cases, an unfavorable outcome at 6 months occurred only in 4 subjects (0.3%) and was unpredictable.

Several limitations should be acknowledged. Firstly, the clinical characteristics of our cohort were obtained from a prospectively collected database, but all analyses were performed post hoc. Secondly, no phone follow-ups were done to ascertain that no events were missed. The ED of Forli is the sole facility in the area where subjects with cardiac events could be referred to, and only events when travelling or on holidays outside our Health District could be sources of error. This does not include death, which would have been traced in the Health District database. Finally, our database is derived from a rural population, which cannot be generalized to an urban population. However, the demographic char- acteristics of the Forli district inhabitants resemble those of the Italian population, an unselected, old population seen in a general ED, at variance with published studies.

Conclusions

In our old population with a diagnosis of CPUO following a negative diagnostic program, the Composite outcome (death of cardiac origin, ACS, need for urgent coronary revascularization) occurs in 1 of 40 cases at 30 days and in 1 of 14 at 6 months. Unfavorable outcomes are associated with a group of clinical variables, but their predictive value remains low for clinical decision making.

References

1. Schriger DL, Brown TB. Decisions, decisions: emergency physician evaluation of low probability-high morbidity conditions. Ann Emerg Med 2005;46(6):534-5.
2. Goodacre S, Nicholl J, Dixon S, et al. Randomised controlled trial and economic evaluation of a chest pain observation unit compared with routine care. BMJ 2004;328(7434):254.
3. Goodacre S, Cross E, Lewis C, Nicholl J, Capewell S. Effectiveness and safety of chest pain assessment to prevent emergency admissions: ESCAPE cluster randomised trial. BMJ 2007;335(7621):659.
4. 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):1-10.
5. Katz DA, Williams GC, Brown RL, et al. Emergency physicians’ fear of malpractice in evaluating patients with possible Acute cardiac ischemia. Ann Emerg Med 2005;46(6):525-33.
6. Sanchis J, Bodi V, Llacer A, et al. Risk stratification of patients with acute chest pain and normal troponin concentrations. Heart 2005;91 (8):1013-8.
7. Goodacre S, Locker T, Morris F, Campbell S. How useful are clinical features in the diagnosis of acute, undifferentiated chest pain? Acad Emerg Med 2002;9(3):203-8.
8. Antman EM, Cohen M, Bernink PJ, et al. The TIMI risk score for unstable angina/non-ST elevation MI: A method for prognostication and therapeutic decision making. JAMA 2000;284(7):835-42.
9. Conway Morris A, Caesar D, Gray S, Gray A. TIMI risk score accurately risk stratifies patients with undifferentiated chest pain presenting to an emergency department. Heart 2006;92(9):1333-4.
10. Prina LD, Decker WW, Weaver AL, et al. Outcome of patients with a final diagnosis of chest pain of undetermined origin admitted under the suspicion of acute coronary syndrome: a report from the Rochester Epidemiology Project. Ann Emerg Med 2004;43(1):59-67.
11. Sekhri N, Feder GS, Junghans C, Hemingway H, Timmis AD. How effective are rapid access chest pain clinics? Prognosis of incident angina and non-cardiac chest pain in 8762 consecutive patients. Heart 2007;93(4):458-63.
12. Bertrand ME, Simoons ML, Fox KA, et al. Management of acute coronary syndromes in patients presenting without persistent ST- segment elevation. Eur Heart J 2002;23(23):1809-40.
13. Bassand JP, Hamm CW, Ardissino D, et al. Guidelines for the diagnosis and treatment of non-ST-segment elevation acute coronary syndromes. Eur Heart J 2007;28(13):1598-660.
14. Diamond GA, Forrester JS. Analysis of probability as an aid in the clinical diagnosis of coronary-artery disease. N Engl J Med 1979;300: 1350-8.
15. Sanchis J, Bodi V, Nunez J, et al. Usefulness of early exercise testing and clinical risk score for prognostic evaluation in chest pain units without preexisting evidence of myocardial ischemia. Am J Cardiol 2006;97(5):633-5.
16. Alpert JS, Thygesen K, Antman E, Bassand JP. Myocardial infarction redefined–a consensus document of the Joint European Society of Cardiology/American College of Cardiology Committee for the redefini- tion of myocardial infarction. J Am Coll Cardiol 2000;36(3):959-69.
17. Geleijnse ML, Elhendy A, Kasprzak JD, et al. Safety and prognostic value of early Dobutamine-atropine stress echocardiography in patients with spontaneous chest pain and a non-diagnostic electrocardiogram. Eur Heart J 2000;21(5):397-406.
18. Linden A. Measuring diagnostic and predictive accuracy in disease management: an introduction to receiver operating characteristic (ROC) analysis. J Eval Clin Pract 2006;12(2):132-9.
19. Reilly BM, Evans AT, Schaider JJ, Wang Y. Triage of patients with chest pain in the emergency department: a comparative study of physicians’ decisions. Am J Med 2002;112(2):95-103.
20. Conti A, Paladini B, Toccafondi S, et al. Effectiveness of a multidisciplinary chest pain unit for the assessment of coronary syndromes and risk stratification in the Florence area. Am Heart J 2002;144(4):630-5.
21. Sanchez M, Lopez B, Bragulat E, et al. Triage flowchart to rule out acute coronary syndrome. Am J Emerg Med 2007;25(8):865-72.
22. Graff LG, Dallara J, Ross MA, et al. Impact on the care of the emergency department chest pain patient from the chest pain evaluation registry (CHEPER) study. Am J Cardiol 1997;80(5):563-8.
23. Davis C, VanRiper S, Longstreet J, Moscucci M. Vascular complica- tions of coronary interventions. Heart Lung 1997;26(2):118-27.
24. Mitchell AM, Garvey JL, Chandra A, Diercks D, Pollack CV, Kline JA. Prospective multicenter study of quantitative pretest probability assessment to exclude acute coronary syndrome for patients evaluated in emergency department Chest pain units. Ann Emerg Med 2006;47 (5):447.
25. Swap CJ, Nagurney JT. Value and limitations of chest pain history in the evaluation of patients with Suspected acute coronary syndromes. JAMA 2005;294(20):2623-9.
26. Manini AF, Dannemann N, Brown DF, et al. Limitations of risk score models in patients with acute chest pain. Am J Emerg Med 2009;27(1): 43-8.
27. Sanchis J, Bosch X, Bodi V, et al. Combination of clinical risk profile, early exercise testing and circulating biomarkers for evaluation of patients with acute chest pain without ST-segment deviation or Troponin elevation. Heart 2008;94(3):311-5.
28. Dunham M, Challen K, Walter D. Risk stratification of patients with acute chest pain without a rise in troponin: current practice in England. Emerg Med J 2010;27(6):461-4.