on 19-Jul-2022 (Tue)

One of the primary goals that researchers look to achieve through customer base analysis is to leverage historical records of individual customer transactions and related context factors to forecast future behavior, and to link these forecasts with actionable characteristics of individuals, managerially significant customer sub-groups, and entire cohorts. This paper presents a new approach that helps firms leverage the automatic feature extraction capabilities of a specific type of deep learning models when applied to customer transaction histories in non-contractual business settings (i.e., when the time at which a customer becomes inactive is unobserved by the firm). We show how the proposed deep learning model improves on established models both in terms of individual-level accuracy and overall cohort-level bias. It also helps managers in capturing seasonal trends and other forms of purchase dynamics that are important to detect in a timely manner for the purpose of proactive customer-base management. We demonstrate the model performance in eight empirical real-life settings which vary broadly in transaction frequency, purchase (ir)regularity, customer attrition, availability of contextual information, seasonal variance, and cohort size. We showcase the flexibility of the approach and how the model further benefits from taking into account static (e.g., socio-economic variables, demographics) and dynamic context factors (e.g., weather, holiday seasons, marketing appeals)

Based on our initial discussion, an ideal model for customer base analysis in data-rich environments would combine a robust forecasting capability both in the short and in the long-term with limited engineering requirements at low computational cost and providing a direct link toward managerial decision-making. Recognizing that traditional statistical forecasting models often suffer from poor efficiency when increasing model complexity and heavily rely on manual feature engineering and data labeling, Table 1 picks up these issues and compares some of the key differences between stochastic BTYD models and the deep learning approach we present in this section

Yehud was a province of the Neo-Babylonian Empire established in the former territories of the Kingdom of Judah, which was destroyed by the Babylonians in the aftermath of the Jewish-Babylonian War, in 587/6 BCE.

Diagnostic evaluation emphasizes distinguishing between the following potential, life-threatening causes of chest pain:

● Acute coronary syndrome (ACS; myocardial infarction or unstable angina)

● Nonischemic chest pain, including potentially life-threatening conditions such as aortic dissection, pulmonary embolism, and esophageal rupture (table 1 and table 2A-B)

Clinicians should consider the possibility of ACS in any adult who presents to the emergency department complaining of chest discomfort or dyspnea. A personal or family history of ACS or other cardiovascular disease, older age, diabetes, dyslipidemia, cigarette smoking, hypertension, or cocaine use all increase the likelihood.

Females are more likely to have associated dyspnea than males, and patients who are older or have diabetes are more likely to present with dyspnea without chest pain.

Pain from coronary-related ischemia is more often characterized as non-focal chest discomfort or pressure rather than pain, is generally gradual in onset, and is exacerbated by activity. Discomfort that radiates, particularly to either or both arms, should increase suspicion for ACS.

If ACS is the leading diagnosis, initial assessment and interventions must be performed rapidly. The attached algorithm provides a simple approach to risk stratification while the table provides a concise outline of the immediate interventions needed to manage ST elevation myocardial infarction (STEMI), non-ST elevation myocardial infarction (NSTEMI), and unstable angina (algorithm 1 and table 3).

During the initial assessment phase, the following steps should be accomplished for any patient at significant risk for ACS:

• Airway, breathing, and circulation assessed

• Preliminary history and examination obtained

• 12-lead electrocardiogram (ECG) interpreted

• Resuscitation equipment brought to the bedside

• Cardiac monitor attached to patient

• Oxygen given as necessary

• Intravenous (IV) access and blood work (including high-sensitivity troponin, if available) obtained

• Aspirin 162 to 325 mg given

• Nitrates given (contraindications to nitrates include severe aortic stenosis, hypertrophic cardiomyopathy, suspected right ventricular infarct, hypotension, marked bradycardia or tachycardia, and recent use of phosphodiesterase 5 inhibitor [eg, Viagra])

Large observational studies report that ECG acquisition is frequently delayed and that females are significantly less likely to be assessed within the recommended 10 minutes or to receive testing with a cardiac biomarker [2-4].

Initial history and exam – Obtain a brief history and perform a focused physical examination. Important elements of the history include confirmation of the presenting symptoms, characteristics of the pain, exacerbating and alleviating factors, and important associated symptoms, past history of or risk factors for cardiovascular disease, and potential contraindications to thrombolytic therapy (table 4).

The history should address other potentially life-threatening, noncardiac causes of chest pain, such as acute aortic dissection, pulmonary embolism, tension pneumothorax, perforating peptic ulcer, and esophageal rupture (table 2A-B). Until the diagnosis of ACS is established, the clinician should keep these diagnoses in mind.

The initial ECG is often not diagnostic in patients with ACS. Thus, the ECG should be repeated at 15- to 30-minute intervals if the initial study is not diagnostic but the patient remains symptomatic and high clinical suspicion for ACS persists.

Supplemental oxygen – Supplemental oxygen should be given as necessary to maintain oxygen saturation above 90 percent.

Cardiac monitor – The patient should be placed on a cardiac monitor with emergency resuscitation equipment (including a defibrillator and airway equipment) nearby

Aspirin – All patients with suspected ACS should be given aspirin in a dose of 162 to 325 mg to chew and swallow unless there is a compelling contraindication (eg, history of anaphylactic reaction) or it has been taken prior to presentation. Despite its well-demonstrated benefit, aspirin remains underutilized in the setting of ACS.

Sublingual nitrates – In most cases of suspected ACS, sublingual nitroglycerin should be administered at a dose of 0.4 mg every five minutes for a total of three doses, after which an assessment of blood pressure and pain relief should guide the need for IV nitroglycerin.

Before this is done, all patients should be questioned about the use of phosphodiesterase-5 inhibitors, such as sildenafil (Viagra), vardenafil (Levitra), or tadalafil (Cialis); nitrates are contraindicated if these drugs have been used in the last 24 hours (or perhaps as long as 36 hours with tadalafil) because of the propensity to cause potentially severe hypotension.

Pain that responds to sublingual nitroglycerin is frequently thought to have a cardiac etiology or to be due to esophageal spasm. However, pain relief with nitroglycerin in an acute care setting is not helpful in distinguishing cardiac from noncardiac chest pain [6]. In a study of 459 patients who presented to an emergency department with chest pain and were admitted to the hospital, the percentage of patients who had relief of chest pain with nitroglycerin was similar among the 141 patients with ACS and the 275 patients without active coronary disease (35 versus 41 percent experienced relief) [7]. As such, nitroglycerin administration should not be given as a diagnostic test in patients with chest pain or a presentation that is not consistent with ACS, as the response to treatment may cloud the clinical picture.

Intravenous morphine sulfate – IV morphine sulfate (initial dose of 2 to 4 mg, with increments of 2 to 8 mg, repeated at 5- to 15-minute intervals) may be given for the relief of severe, persistent chest pain not relieved by other means but should not be given routinely. Morphine can reduce sympathetic stimulation caused by pain and anxiety, thereby decreasing cardiac workload and risks associated with excess catecholamines. However, morphine may worsen outcomes in patients with acute myocardial infarction

One research group has proposed the following rule for identifying patients at high risk for ST elevation myocardial infarction (STEMI) who should be evaluated with an ECG [8]:

● Any patient over 30 with chest pain

● Any patient over 50 with any of the following: dyspnea, altered mental status, upper extremity pain, syncope, or weakness

● Any patient over 80 with abdominal pain, nausea, or vomiting

This proposed rule should not replace clinical judgement (in fact, the authors encourage clinicians to obtain an ECG based upon such judgement). As an example, a patient in whom a noncardiac diagnosis is likely (eg, younger patient with obvious zoster [shingles] lesions and no other symptoms) may not need an ECG.

Clinicians must be aware that the initial ECG is often not diagnostic in patients with ACS. Forty-one percent of patients with non-ST elevation myocardial infarction (NSTEMI) do not have the typical ST-depression or T-wave inversion [10]. In patients at intermediate to high risk for ACS without a clear diagnosis, ECGs should be repeated at frequent intervals until the patient's chest pain resolves or a definitive diagnosis is made.

Findings consistent with ST elevation myocardial infarction (STEMI):

• New or presumed new ST elevation at the J point in two anatomically contiguous leads using the following diagnostic thresholds:

≥0.1 mV (1 mm) in all leads other than V2 to V3, where the following diagnostic thresholds apply:

- ≥0.2 mV (2 mm) in males ≥ 40 years

- ≥0.25 mV (2.5 mm) in males <40 years

- ≥0.15 mV (1.5 mm) in females

Findings consistent with NSTEMI:

• New or presumed new horizontal or down-sloping ST depression ≥0.05 mV (0.5 mm) in two anatomically contiguous leads

and/or

• T wave inversion ≥0.1 mV (1 mm) in two anatomically contiguous leads with prominent R wave or R/S ratio >1

The anatomic location of a transmural infarct is determined by which ECG leads show ST elevation and/or increased T wave positivity:

● Anterior wall ischemia – Two or more of the precordial leads (V1 to V6) (waveform 2A-B)

● Anteroseptal ischemia – Leads V1 to V3 (waveform 3 and waveform 4)

● Apical or lateral ischemia – Leads aVL and I, and leads V4 to V6 (waveform 5 and waveform 6)

● Inferior wall ischemia – Leads II, III, and aVF (waveform 1)

● Right ventricular ischemia – Right-sided precordial leads (waveform 1)

● Posterior wall ischemia – Septal precordial leads (V1 to V2) (waveform 7) and posterior precordial leads (waveform 8)

The right-sided leads V4R, V5R, and V6R (figure 1) should be obtained if there is evidence of inferior wall ischemia demonstrated by ST elevation in leads II, III, and aVF.

The posterior leads V7, V8, and V9 (figure 2) may also be helpful if there is evidence of posterior wall ischemia as suggested by prominent R waves and ST depressions in leads V1 and V2.

If the initial ECG is not diagnostic but the patient remains symptomatic and clinical suspicion for ACS remains high, the ECG should be repeated at least every 15 to 30 minutes.

The initial ECG is often not diagnostic in patients with ACS. In two series, the initial ECG was nondiagnostic in 45 percent and normal in 20 percent of patients subsequently shown to have an acute myocardial infarction [1,11].

In the early hours of infarction, peaked, hyperacute T waves may be the only abnormality. In addition to evolution of the ECG, an uncommon source of error is pseudonormalization of baseline T wave inversion [ 12].

Some clinicians assume that an ECG obtained while the patient is experiencing chest pain that fails to show evidence of ischemia rules out the possibility of ACS. This assumption is false, as demonstrated by two prospective observational studies [13,14].

Ischemic pain has a number of features that tend to distinguish it from noncardiac pain. These characteristics are described below using the OPQRST mnemonic

Symptoms associated with the highest relative risk of myocardial infarction include radiation to an upper extremity, particularly when there is radiation to both arms, and pain associated with diaphoresis or with nausea and vomiting [ 19-21].

An important question is whether current pain is reminiscent of prior myocardial infarction. Of note, no one pain characteristic has sufficient positive or negative predictive value to definitively diagnose or exclude ACS

Onset – Ischemic pain is typically gradual in onset, although the intensity of the discomfort may wax and wane.

Provocation and palliation – Ischemic pain is generally provoked by an activity, such as exercise, that increases cardiac oxygen demand [20,22]. Ischemic pain does not change with respiration or position. It may or may not respond to nitroglycerin and, if there is improvement, this may only be temporary. Relief of pain following the administration of therapeutic interventions (eg, nitroglycerin, "gastrointestinal cocktail" of viscous lidocaine and antacid) does not reliably distinguish nonischemic from ischemic chest pain [23-25].

Quality – Ischemic pain is often characterized more as a discomfort than pain, and it may be difficult for the patient to describe. Terms frequently used by patients include squeezing, tightness, pressure, constriction, crushing, strangling, burning, heartburn, fullness in the chest, band-like sensation, knot in the center of the chest, lump in throat, ache, heavy weight on chest (elephant sitting on chest), like a bra too tight, and toothache (when there is radiation to the lower jaw). It is generally not described as sharp, fleeting, knife-like, stabbing, or like "pins and needles" [20].

Increased pain severity does not appear to correlate with an increased likelihood of acute myocardial infarction [26].

In some cases, the patient cannot qualify the nature of the discomfort but places his or her clenched fist in the center of the chest, known as the "Levine sign."

Radiation – Ischemic pain often radiates to other parts of the body including the upper abdomen (epigastrium), shoulders, arms (upper and forearm), wrist, fingers, neck and throat, lower jaw and teeth (but not upper jaw), and not infrequently to the back (specifically the interscapular region). The old dictum that "pain above the nose or below the navel is rarely cardiac in origin" still holds. Pain radiating to the upper extremities is highly suggestive of ischemic pain [19,20,22,27].

Site – Ischemic pain is not felt in one specific spot; rather, it is a diffuse discomfort that may be difficult to localize. The patient often indicates the entire chest, rather than localizing it to a specific area by pointing a single finger.

Time course – Angina is usually brief (two to five minutes) and is relieved by rest or with nitroglycerin. In comparison, patients with an ACS may have chest pain at rest, and the duration is variable but generally lasts longer than 30 minutes. Classic anginal pain lasting more than 20 minutes suggests ACS. Continuous pain that does not wax and wane and persists for over 24 hours is unlikely to be due to ACS [28].

Historical features that increase the likelihood of ACS include the following:

● Patients with a prior history of ACS have a significantly increased risk of recurrent ischemic events.

● A prior history of other vascular disease is associated with a risk of cardiac ischemic events comparable to that seen with a prior history of ACS.

● Risk factors for ACS, particularly age, male sex, diabetes, hypertension, dyslipidemia, and cigarette smoking [29]. (See "Overview of established risk factors for cardiovascular disease".)

● Recent use of cocaine or other sympathomimetic drugs (eg, methamphetamine).

Ischemic pain is often associated with other symptoms. The most common is shortness of breath, which may reflect mild pulmonary congestion resulting from ischemia-mediated diastolic dysfunction. Other symptoms may include belching, nausea, indigestion, vomiting, diaphoresis, dizziness, lightheadedness, clamminess, and fatigue

Systematic reviews have identified the following characteristics as more typical of nonischemic chest discomfort [19]:

● Pleuritic pain, sharp or knife-like pain related to respiratory movements or cough

● Primary or sole location in the mid or lower abdominal region

● Any discomfort localized with one finger

● Any discomfort reproduced by movement or palpation

● Constant pain lasting for days

● Fleeting pains lasting for a few seconds or less

● Pain radiating into the lower extremities or above the mandible

However, some patients with ACS present with so-called "atypical" chest pain. This was illustrated in the Multicenter Chest Pain Study in which acute ischemia was diagnosed in 22 percent of patients who presented with sharp or stabbing pain and 13 percent who presented with pleuritic-type pain [30].

In a review of over 430,000 patients with confirmed acute myocardial infarction from the National Registry of Myocardial Infarction II, one-third had no chest pain on presentation to the hospital [31]. These patients often present with symptoms such as dyspnea alone, weakness, nausea and/or vomiting, epigastric pain or discomfort, palpitations, syncope, or cardiac arrest. They are more likely to be older or have diabetes [20,31-33].

Focused neurologic examination – A screening neurologic examination should be performed to assess for focal lesions or cognitive deficits that might preclude safe use of thrombolytic therapy (table 4).

Anemia can exacerbate myocardial ischemia, and patients with a low hemoglobin may require blood transfusion. Transfusion thresholds and implementation are reviewed separately. (See "Indications and hemoglobin thresholds for red blood cell transfusion in the adult", section on 'ACS (including MI)'.)

Indices of coagulation should be obtained in patients at increased risk of coagulopathy due to warfarin or heparin use, a history of liver disease, or a history of excessive or spontaneous bleeding.

The management of patients who meet the criteria for ST elevation myocardial infarction (STEMI) is discussed separately. The accompanying table provides a concise summary of the immediate treatment interventions needed in these patients (table 3).

Selection and implementation of the optimal reperfusion strategy is the most important step in the management of STEMI and is discussed separately. Reperfusion therapy, whether percutaneous coronary intervention (PCI) or thrombolytics, should not await the result of cardiac biomarker measurement.

Unstable angina and NSTEMI comprise part of the spectrum of ACS. Angina is considered unstable if it presents in any of the following three ways:

● Rest angina, generally lasting longer than 20 minutes

● New-onset angina that markedly limits physical activity

● Increasing angina that is more frequent, lasts longer, or occurs with less exertion than previous angina

NSTEMI is distinguished from unstable angina by the presence of elevated serum biomarkers. ST segment elevations and Q waves are absent in both unstable angina and NSTEMI. As a result, unstable angina and NSTEMI can be indistinguishable at initial evaluation since an elevation in serum biomarkers may not be detectable for two to four hours after a myocardial infarction with contemporary troponin assays and one to three hours with high-sensitivity assays, and at least 12 hours are required to detect elevations in all patients.

Thrombolytic therapy should not be administered to patients with unstable angina or NSTEMI unless subsequent ECG monitoring documents ST segment elevations that persist. Repeat ECGs are an essential part of management in these patients.

● Sustained ventricular tachyarrhythmias in the peri-infarction period must be treated immediately because of their deleterious effect on cardiac output, possible exacerbation of myocardial ischemia, and the risk of deterioration into ventricular fibrillation (algorithm 2). (See "Ventricular arrhythmias during acute myocardial infarction: Incidence, mechanisms, and clinical features".)

● While supraventricular tachyarrhythmias in the peri-infarction period may pose less immediate risk of cardiac arrest, the management of such arrhythmias is important because any tachycardia can increase myocardial oxygen demand, thereby exacerbating ischemia and possibly decreasing cardiac output (algorithm 4). (See "Supraventricular arrhythmias after myocardial infarction".)

● Bradyarrhythmias occurring early in the setting of an inferior wall myocardial infarction (within the first 24 hours) may respond to treatment with atropine (algorithm 3). Later bradyarrhythmias, wide QRS-complex bradyarrhythmias, and those occurring in the setting of an anterior wall myocardial infarction may require temporary pacemaker placement. (See "Conduction abnormalities after myocardial infarction".)

The patient is at high risk for ACS if ST segment depression (≥0.05 mV [0.5 mm]) is present in two or more contiguous leads or serum biomarkers are elevated. A high score on a validated ACS risk stratification tool suggests the patient is at high risk. Such scores include the HEART Score (table 8) and EDACS Score [37].

Low- and moderate-risk patient — The management, including methods for risk assessment, of patients who have no ECG changes and normal serum biomarkers but are still considered to be at low or moderate risk for ACS is discussed separately.

With careful evaluation using effective risk scores, serial electrocardiograms (ECGs), and appropriate diagnostic testing, only a very small percentage of patients with an ACS are mistakenly discharged from the emergency department [38,39]. However, patients whose diagnosis is missed initially have an increase in short-term mortality [32,38,40,41]

The patients with missed myocardial infarction had the following characteristics:

● Females less than 55 years of age

● Not White Americans

● Shortness of breath as the major presenting symptom

● Normal or nondiagnostic ECG

Angina pectoris, or angina for short, is the term used when chest discomfort is thought to be attributable to myocardial ischemia.

Myocardial ischemia in the absence of chest discomfort or another anginal equivalent symptoms is termed "silent ischemia."

Myocardial ischemia can occur due to fixed epicardial coronary artery ("macrovascular") obstruction, coronary microvascular obstruction (ischemia with no obstructive coronary artery disease [INOCA]), or coronary artery spasm.

Myocardial oxygen demand — There are four major factors that determine myocardial work and therefore myocardial oxygen demand:

● Heart rate

● Systolic blood pressure (the clinical marker of afterload)

● Myocardial wall tension or stress (the product of ventricular end-diastolic volume or preload and myocardial muscle mass)

● Myocardial contractility

Myocardial oxygen supply — The major determinants of oxygen supply are the oxygen carrying capacity of the blood, which is affected by a variety of factors including oxygen tension and the hemoglobin concentration; the degree of oxygen unloading from hemoglobin to the tissues, which is related to 2,3 diphosphoglycerate levels; and the coronary artery blood flow delivered to the myocardium. The latter is influenced by:

● Coronary artery diameter and tone (resistance) [2,3].

● Collateral blood flow.

● Perfusion pressure. This is determined by the pressure gradients from the aorta to the coronary arteries. Coronary blood flow from the epicardium to endocardial capillaries is determined by the left ventricular end-diastolic pressure.

● Heart rate, which affects the duration of diastole; importantly, coronary artery flow primarily occurs during diastole. The percent of diastolic time decreases as the heart rate increases. Thus, heart rate is a determining factor for both oxygen demand and supply.

The nerve fibers travel along the sympathetic afferent pathways from the heart and enter the sympathetic ganglia in lower cervical and upper thoracic spinal cord (C5-6 and T1-T6). Impulses are then transmitted via the ascending spinothalamic pathways to the medial and lateral thalamus and ultimately activate several areas of the cerebral cortex [4].

Angina is a discomfort that is referred to the corresponding dermatomes that supply sympathetic afferent nerves to the same segments of the spinal cord as the heart [4]. Furthermore, stimulation of sensory receptors in different myocardial regions results in the transmission via the same neural pathway [13]. These characteristics account for two typical features of angina: It is often a diffuse discomfort felt in the chest, neck, lower jaw, and down the arm (typically the left arm, although some patients experience right arm discomfort). Most patients experience angina in the same distribution, regardless of which area of the myocardium is ischemic [13]. An exception is often post-cardiac surgery, as this may interrupt and alter the neural supply to the heart, which may affect the distribution of angina

Left ventricular hypertrophy may result in a reduction in subendocardial blood flow and oxygen supply; this may result in angina. This may be particularly important with the development of arterial hypertension, which may increase left ventricular end-diastolic pressure, resulting in impairment of capillary flow in the subendocardium.

Other examples of inadequate supply include shock (any cause), hypoxemia, anemia, and postprandial angina resulting from a redistribution of blood flow away from territories supplied by severely stenosed coronary arteries to those supplied by less diseased or normal arteries (ie, a steal phenomenon) [14,15]

The initial presentation of myocardial ischemia with angina may be one of a stable pattern or an acute coronary syndrome. Patients with recent onset of chest discomfort, episodes of rest chest discomfort, or one or more prolonged episodes (more than 20 minutes), should be evaluated in an acute care facility for the possibility of an acute coronary syndrome. However, every patient has their first episode of chest discomfort so that recent onset is not necessarily an acute coronary syndrome, particularly if the pattern of occurrence is stable and predictable (ie, with exertion)

Quality − Angina is usually characterized more as a discomfort rather than pain. Terms frequently used by patients include squeezing, tightness, pressure, constriction, strangling, burning, heart burn, fullness in the chest, band-like sensation, knot in the center of the chest, lump in throat, ache, heavy weight on chest (elephant sitting on chest), like a bra too tight, and toothache (when there is radiation to the lower jaw) [17]. In some cases, the patient cannot qualify the nature of the discomfort, but places his or her fist in the center of the chest, known as the "Levine sign."

It is generally not described as sharp, dull-aching, knife-like, stabbing, or pins and needles-like. In a report of patients presenting to the emergency department, "sharp" or "stabbing" pain was a low risk description, particularly when the pain was pleuritic or positional, was fully reproducible by palpation, and the patient had no history of angina or myocardial infarction [18].

Angina is typically gradual in onset and offset, with the intensity of the discomfort increasing and decreasing over several minutes. In contrast, noncardiac pain is often of greatest intensity at its onset and often has an abrupt onset and offset.

Since angina is a referred discomfort, patients tend to have the same quality of chest discomfort with recurrent ischemic episodes [13]. Generally, it is felt in the same location. The discomfort is generally the same prior to or with a myocardial infarction and is the same quality as prior to revascularization by either surgery (although the location may be different due to disruption of neural innervation of the heart) or percutaneous coronary intervention.

Angina is a constant discomfort that does not change with respiration or most changes in position (one exception is lying down, which increases venous return). It is also not provoked or worsened with palpation of the chest wall. However, the presence of a change in pain with respiration (or position) or pain elicited by palpation does not exclude angina as the cause.

The patient often indicates the entire chest when asked where the discomfort is felt. Pain that localizes to one small area of the chest is more likely of chest wall or pleural origin rather than visceral.

Angina is referred to the corresponding dermatomes (C5-6 and T1-T6) that supply afferent nerves to the same segments of the spinal cord as the heart. Thus, angina often radiates to other parts of the body, including the upper abdomen (epigastric), shoulders, arms (upper and forearm), wrist, fingers, neck and throat, lower jaw and teeth (but not upper jaw), and rarely to the back (specifically the interscapular region) [19,20]

Radiation to both arms is a stronger predictor of acute myocardial infarction. The location and radiation of angina is usually the same each time. Occasionally, the location and radiation, but not quality, may be different after bypass surgery due to the disruption of the neural innervation of the heart.

Isolated back pain is unusual in patients with angina. However, it may be seen with an aortic dissection that also involves the coronary arteries.

Provoking factors − Angina is often elicited by activities and situations that increase myocardial oxygen demand, including physical activity, cold, emotional stress, sexual intercourse, meals, or lying down (which results in an increase in venous return and increase in wall stress) [21-23]. Patients should be questioned about the use of cocaine or other recreational drugs, as they may trigger myocardial ischemia.

Postprandial pain is generally considered to be gastrointestinal in origin. However, it may also be anginal, especially in patients with severe ischemia (eg, left main or three vessel coronary disease) [15].

Timing − Angina occurs more commonly in the morning due to a diurnal increase in sympathetic tone. Enhanced sympathetic activity raises heart rate, blood pressure, vessel tone and resistance (resulting in a reduced vessel diameter that causes any fixed lesion to be more occlusive), and promotes platelet aggregation (resulting in the release of vasoactive substances, such as serotonin and thromboxane A2) [8,9].

Duration and relief − Classic angina is often relieved with termination of the provoking factor. Angina generally lasts for two to five minutes. It is not a fleeting discomfort, which lasts only for a few seconds or less than a minute, and it generally does not last for 20 to 30 minutes, unless the patient is experiencing an acute coronary syndrome, especially myocardial infarction.

Factors that reduce oxygen demand or increase oxygen supply will result in relief of angina. These include cessation of activity or termination of the provoking factor, use of nitroglycerin (which is a venodilator, reducing venous return, and a coronary artery vasodilator that increases coronary blood flow), and sitting up (which reduces venous return and preload).

However, some patients with an acute coronary syndrome present with atypical types of chest pain. In one study, acute ischemia was diagnosed in 22 percent of patients who presented with sharp or stabbing pain and 13 percent who presented with pleuritic-type pain [18].

Angina is often associated with other symptoms. Dyspnea is a common anginal equivalent, and its presence is associated with a higher cardiovascular death rate compared with patients with typical angina. Dyspnea in the setting of angina may reflect pulmonary congestion due to an elevation in left ventricular end diastolic pressure related to failure of the myocardium to relax normally in diastole (as relaxation or lusitropy is energy dependent). The resulting diastolic "stiffness" or diastolic dysfunction results in an increase in left ventricular end diastolic pressure, left atrial pressure, and pulmonary venous pressure, which is transmitted to the pulmonary vessels.

Other symptoms may include belching, nausea, indigestion, diaphoresis, dizziness, lightheadedness, clamminess, and fatigue. These have been referred to as "angina equivalent" symptoms and appear to be more common in women compared with men. However, these symptoms may be seen with other etiologies for chest pain, especially gastrointestinal causes.

It is common for patients with diabetes mellitus, who often have autonomic (sympathetic) dysfunction, to experience "silent ischemia" or best termed "discomfortless ischemia." (see "Silent myocardial ischemia: Epidemiology, diagnosis, treatment, and prognosis"). They may also present with anginal equivalent symptoms.

Physical examination — Ischemia can produce impairment in myocardial function, which may result in the following findings on physical examination. All disappear with resolution of the ischemia. Some patients have none of these features.

Increase in heart rate — Ischemia can raise the heart rate even if the patient is receiving a beta blocker or calcium channel blocker. The increase in heart rate is induced by reflex sympathetic nervous system activation as a response to ischemia.

Elevation in blood pressure — Ischemia often causes a hypertensive blood pressure response. The elevation in blood pressure is induced by both sympathetic activation in response to ischemia and stimulation of the left anterior descending coronary artery chemoreceptor. This chemoreceptor is stimulated by serotonin secreted as a result of platelet aggregation, which often occurs in association with angina.

New heart sounds — Ischemia-induced myocardial dysfunction can lead to changes in the normal heart sounds. The second heart sound may become paradoxically split due to delayed relaxation of the left ventricular myocardium and delayed closure of the aortic valve. There may also be a third or fourth heart sound.

New/changed murmurs — Impaired myocardial function may result in a new mitral regurgitation murmur, which appears to be due to papillary muscle dysfunction causing apical tethering or tenting of the leaflets, or changes in the intensity or timing of pre-existing murmurs.

Precordial pulsation — Palpation of the chest wall may reveal abnormal pulsations that correlate with transient left ventricular dysfunction. An area of dyskinesis may develop, especially at the apex of the left ventricle or at the anterior axillary line (location of the left ventricular wall), reflecting disease of the left anterior descending coronary artery. (See "Examination of the precordial pulsation".)

Palpation of the left anterior chest wall at the anterior axillary line may reveal an abnormal tapping in systole, which reflects the presence of an area of dyskinetic contraction or aneurysm. Transient right ventricular dysfunction may lead to a transient right ventricular heave or sternal pulsation.

Cardiac biomarkers (eg, troponin) are often obtained in patients with angina. They are unlikely to be elevated in patients with intermittent and relatively brief angina episodes. However, they may be useful when the anginal episode is more prolonged and for prognostication.

The term "ACS" is applied to patients in whom there is evidence of myocardial ischemia or infarction. There are three types of ACS: ST-elevation myocardial infarction (STEMI), non-ST-elevation MI (NSTEMI), and unstable angina (UA). The first two are characterized by a typical rise and/or fall in serum troponin [2]. UA is characterized by new or worsening myocardial ischemia without elevated biomarkers and is often a clinical diagnosis based on history, dynamic ECG changes, or inducible ischemia on stress testing. UA accounts for a smaller proportion of ACS now that there is widespread use of highly sensitive troponin, which can detect very small amounts of myocardial injury

The primary focus in the very early evaluation (generally within the first 15 minutes after presentation) of patients with possible ACS is to reasonably confirm or exclude in an expeditious manner ACS as the cause for the patient's symptoms (algorithm 1). In this early time period, reasonable confirmation of ACS is generally secured by the finding of diagnostic ECG changes. ACS is reasonably excluded when the probability of ACS is very low (generally considered below a 1 to 2 percent threshold [3,4]).

An initial 12-lead ECG within 10 minutes of arrival (even if one has been performed in an ambulance, unless that ECG showed ST-segment elevation) in patients ≥25 years of age with chest pain, or age ≥50 years with chest pain or other symptoms that could represent myocardial ischemia like shortness of breath, altered mental status, upper extremity pain, syncope, or generalized weakness. Very elderly (≥80 years of age) patients with abdominal pain or nausea/vomiting should have an immediate 12-lead ECG as well [2,5].

While patients who are assessed to be at very low risk of an ACS do not need a troponin test, many institutions have protocols that direct nursing staff to order them based on their assessment, which may be prior to physician evaluation. We believe this is a reasonable strategy.

The results of the first high sensitivity troponin should be available within 60 minutes.

We attempt to assign patients to low, intermediate, or high risk for ACS category, or definite ACS (table 1), based on information obtained from the history, physical examination, ECG, and continuous ECG monitoring (algorithm 1) but before the first troponin level is available.

Multiple scoring systems have been evaluated. We prefer the History, ECG, Age, Risk factors, and a single sensitive Troponin (HEART) score, which is used in the HEART Pathway (see 'Care pathways' below). Using this score, individuals with a score of 0 to 3 are considered low risk. There is an element of subjectivity in the HEART score, so internal discussions to improve consistency of use are suggested [7].

Low-risk patients generally have none of the risk factors found in the table (table 1). Specifically, low-risk patients cannot have a history of coronary artery disease or have ST-segment elevation, significant ST-segment depression, or new T wave inversion.

At some lower threshold of pretest probability, the clinician is more likely to cause harm by further testing. It has been suggested that an acceptable lower limit of pretest probability for ACS, above which further testing should be done, is 1 to 2 percent [3,4]. We believe patients below this threshold are very low risk and are unlikely to benefit from additional testing for an ACS. All other patients should receive further testing and risk stratification.

Providers commonly identify very low-risk patients using an experienced clinician's unstructured clinical estimate. In an evaluation of this method, approximately 25 percent of chest pain patients were classified as very low risk (<2 percent risk), and these patients experienced an event rate of 0.7 percent (95% CI 0-2.4) [8].

When formulating the clinician's pretest (before the troponin) probability, it is important to avoid eliminating the possibility of an ACS even when an alternative diagnosis is possible. In patients being evaluated for chest pain, it has been shown that those with a "clear-cut" alternative noncardiac diagnosis still have a 4 percent rate of ACS at 30 days [9].

As an example, among patients with ACS, a common misdiagnosis is gastroesophageal reflux disease. In two different datasets, it has been shown that patients with ACS commonly describe their pain as burning [11,12]. A gastrointestinal etiology should be assigned only after the diagnosis of a cardiac emergency has been excluded.

While we have identified specific times in this process when evaluation must take place rapidly, evaluation is a continuous process, and new information at any time point can alter a patient's risk category. For example, worsening of symptoms, arrhythmias, or new ECG changes on continuous monitoring can make the possibility of an ACS more likely.

Compared with a standard troponin assay, hs-cTn leads to the diagnosis of acute MI in a modest additional number of chest pain patients and may do so earlier in the patients' emergency department (ED) stay [14]. This occurs because hs-cTn detects biochemical evidence of myocardial injury at lower concentrations, and hs-cTn assays allow for discrimination of small changes in concentration even within the normal reference range. However, MI is not ruled in unless there is both a rising and/or falling pattern of values and one value above the upper reference limit (URL) [15]

With hs-cTn, serial values <99^th percentile of the URL make the diagnosis of acute MI quite unlikely, but the value does not exclude the possibility of unstable angina [17,18]

If the first troponin returns clearly elevated in a patient with clinical findings consistent with ACS, the management for NSTEMI likely begins. (See "Overview of the acute management of non-ST elevation acute coronary syndromes".)

However, this paradigm may be changing with the routine use of increasingly sensitive troponin assays. Data from outside the United States suggest that 11 to 15 percent of patients will be classified as having NSTEMI based on that initial hs-cTn result, of whom 68 to 79 percent will ultimately be adjudicated as having NSTEMI [22]. As these hs-cTn assays are applied to broader patient populations with concomitant medical conditions such as sepsis and renal failure, this positive predictive value is likely to be reduced. This fact needs to be considered by clinicians.

If the troponin value is less than the upper reference limit, further evaluation is directed by the clinical presentation and risk factors. (See 'Risk scores' below.)

● For patients whose symptoms remain suspicious for an ACS or who have a low or intermediate risk of ACS (Heart Score ≥1), obtain a second troponin (algorithm 1). For most patients, we obtain a second troponin three hours after the first troponin and use the HEART Pathway approach (see 'Care pathways' below). Monitoring continues until a second troponin returns.

● For selected patients whose symptoms are best explained by a noncardiac diagnosis (eg, costochondritis, pneumonia), and who have a risk for ACS less than 2 percent by clinical evaluation or a Heart Score of zero, the evaluation for ACS can stop [23].

On occasion, some patients will require a third troponin. Examples include those with stuttering pain, new return of pain while in the ED, patients presenting very soon after the onset of symptoms, and those in whom the clinician has a high degree of concern.

For low-risk patients or those with normal ECGs who present more than three hours after the onset of chest pain [21,24], a value of hs-cTn below the limit of detection of the assay is a sensitive way to rule out acute MI.

With hs-cTn assays, one- to three-hour rule-outs using assay specific change criteria are usually adequate. There are only rare patients who rule in thereafter [34].

Thus, any patient who is >3 to 6 hours after the onset of symptoms (assuming no recurrences) and who has a normal cTn value (either hs-cTn or contemporary cTn) can be considered to have ruled out for acute MI [36].

Late presenters — Patients who present 12 hours or more after the onset of symptoms can be diagnostically challenging. If the first hs-cTn is elevated in these patients, clinicians should be careful in their interpretation of changes or deltas between first and second values. Patients with adjudicated NSTEMI with less acute presentations, and those with longer ischemic times, are more likely to present near their peak hs-cTn value. Subsequent values may downtrend or stay about the same, providing false reassurance or misleading the clinician away from the diagnosis of NSTEMI, but still leaving the patient with substantial risk for mortality from ACS [37]. These patients potentially require additional testing.

The TIMI score (calculator 1) was derived and validated in patients enrolled in clinical trials with ACS, but has since been validated in patients suspected to have ACS [40]. Advantages include simplicity, lack of a need for a nomogram or computer algorithm, and its ability to predict a risk of short-term (30 days) outcome [40].

Among the elements of the TIMI risk score, troponin elevation was the most powerful factor. However, even TIMI 0 patients had a significant 2.1 percent incidence of the composite outcome. As such, the TIMI score has limited utility as a tool to determine who can be discharged as very low risk with no further testing. Once ACS has been documented, then the GRACE risk score performs better than the TIMI score [ 33]

The HEART score (table 1) uses similar components to the TIMI risk score. The HEART score has been widely validated [23,41,42]. However, validation studies have shown the need to add a three-hour troponin to achieve acceptable sensitivity [7]. A score of 0 to 3 identifies a patient at low risk of major adverse cardiovascular events; patients with a score of 4 to 6 have intermediate risk, and those with a score of 7 or greater are high risk. A meta-analysis of 30 studies of the HEART risk score found that a score of 4 or greater had a sensitivity of about 96 percent for the prediction of short-term (39 days or 6 weeks) incidence of major adverse cardiovascular events [43].

If the second troponin value is also normal, we use the HEART score to reestimate risk (table 1) and the HEART Pathway (algorithm 1) to determine the next step in management.

It is generally agreed upon by the medical community that protocols designed to facilitate early discharge of patients who are initially suspected to be at low or moderate risk of ACS should have a negative predictive value of greater than 99 percent; that is, 99 percent of individuals chosen for early discharge will not have an adverse cardiovascular event within 30 days.

The HEART Pathway consists of a modified HEART score (table 1), which is estimated from clinical data, and two troponin values at zero and three hours (algorithm 1) [7]. (See 'Risk assessment' above.)

• If the HEART score is ≤3 points and two troponin measurements are negative (ie, below the 99^th percentile), and no new ischemic ECG changes are present, the risk of incident ischemic events is considered sufficiently low for discharge without further observation or diagnostic testing [23,45-48].

• For HEART scores >3 points, the patient is at intermediate or high risk and should be appropriately managed for ACS [46].

Similar to the 0/2- and 0/3-hour pathways, the very low troponin value at presentation in a 0/1-hour pathway is not helpful for patients with recent onset (<3 hours) chest pain [33].

Patients identified as low risk with these approaches can be discharged home with primary care follow-up. However, with all decision aids, it is important to incorporate the provider's clinical gestalt into this decision

Some patients will have been assessed to be at high risk, despite a normal biomarker, because of the presence of ongoing chest pain, new ECG changes, or heart failure. These patients should be admitted and managed as having unstable angina and/or heart failure.

In patients classified as not low risk, subsequent noninvasive provocative testing remains an important part of the evaluation. Among patients with normal or nondiagnostic ECG and normal troponin results, up to 10 percent may still have ACS [3], and 2 to 4 percent may have early adverse events [4,60]. Such patients may be cared for in the ED, in an observation unit, or are admitted.

If the results of stress testing or imaging are negative for a cardiac cause, and other life-threatening components of the differential diagnosis have been ruled out, discharge is appropriate for stable patients. These patients are labeled as having noncardiac chest pain. Many of these patients will carry the diagnosis of nonspecific chest pain while some will ultimately be diagnosed with cardiac causes such as myocarditis/pericarditis, stress cardiomyopathy, or heart failure [67].

One of the primary goals that researchers look to achieve through customer base analysis is to leverage historical records of individual customer transactions and related context factors to forecast future behavior, and to link these forecasts with actionable characteristics of individuals, managerially significant customer sub-groups, and entire cohorts.

We suggest the following at the time of discharge:

• Provide care for the most likely cause of pain.

• Inform the patient about the basis for the presumptive diagnosis and discuss areas of uncertainty in the diagnosis.

• Establish a plan for follow-up with a care provider, ideally within 72 hours.

• Provide clear and concise written discharge instructions that describe specific reasons for reevaluation and where and when to return.

We end up with the following attributes for the parent and child components:

| | | | | --- | --- | --- |Parent attributes | Name | Type | Description | | Employee_id | Categorical | 78 levels | | order_date | Datetime | 1996-1998 | | required_date | Datetime | 1996-1998 | | shipped_date | Datetime | 1996-1998 | | ship_via | Categorical | 3 levels | | freight | Numerical | [0.02 - 1007] | | ship_region | Categorical | 19 levels | | ship_country | Categorical | 21 levels | | customer_city | Categorical | 70 levels | | customer_region | Categorical | 19 levels | | customer_country | Categorical | 21 levels | | order_lenght | Numerical | [1 - 22] | | | | | | --- | --- | --- |Table 3: Child attributes Name Type Description product_id Categorical 77 levels supplier_id Categorical 29 levels category_id Categorical 8 levels unit_price Numerical [2 - 263.5] quantity_per_unit Numerical [0 - 70]

All data is converted into a one-hot encoding, including datetime and numerical types (previously binned into 20 levels).

Based on our initial discussion, an ideal model for customer base analysis in data-rich environments would combine a robust forecasting capability both in the short and in the long-term with limited engineering requirements at low computational cost and providing a direct link toward managerial decision-making.

● Normal first troponin – If the troponin value is less than the upper reference limit, further evaluation is directed by the clinical presentation and risk factors. (See 'Normal first troponin' above and 'Late presenters' above and 'Risk scores' above.)

• For most patients whose symptoms are best explained by a noncardiac diagnosis (eg, costochondritis, pneumonia), and who have a risk for ACS less than 2 by clinical evaluation or a Heart Score of zero, the evaluation for ACS can stop (algorithm 1).

• For patients whose symptoms remain suspicious for an ACS or who have a low or intermediate risk of ACS (HEART Score ≥1), obtain a second troponin (algorithm 1). For most patients, we obtain a second troponin three hours after the first troponin and use the HEART Pathway approach. (See 'Troponin testing' above and 'Care pathways' above.)

● Normal second troponin – If the second troponin value is also normal, we use the HEART score to reestimate risk (table 1) and the HEART Pathway (algorithm 1) to determine the next step in management (see 'Care pathways' above):

• For HEART Pathway score values ≤3 and two negative troponin measurements three hours apart, the patient can be discharged with follow-up in the next week.

• For HEART Pathway score values >3 and two negative troponin measurements three hours apart, the patient should remain under observation and undergo non-invasive ischemia evaluation prior to discharge.

Cardiac disease is the leading cause of death in the United States, yet only 2 to 4 percent of patients presenting to a primary care office with chest pain will have unstable angina or an acute myocardial infarction [2-4].

The most common causes of chest pain in the primary care population are chest wall pain (20 to 50 percent), reflux esophagitis (10 to 20 percent), and costochondritis (13 percent).

In this specific domain of customer base analysis, probabilistic approaches from the ‘‘Buy ’Till You Die” (BTYD) model family represent the gold standard, leveraging easily observable Recency and Frequency (RF, or RFM when including also the monetary value) metrics together with a latent attrition process to deliver accurate predictions (Schmittlein, Morrison, & Colombo, 1987; Fader, Hardie, & Lee, 2005; Fader & Hardie, 2009)

Aortic dissection – Aortic dissection is rare but may be a surgical emergency. Patients with acute aortic dissection typically present with acute chest and back pain that is severe and sharp and may have a ripping or tearing quality [5]. Pain can radiate anywhere in the chest or into the abdomen.

Pulmonary embolism – The most common symptoms of pulmonary embolism include dyspnea followed by pleuritic chest pain, cough, and symptoms of deep venous thrombosis. Risk factors that may predispose to thromboembolic disease are listed in the table (table 1). In addition, the Wells score is useful for calculating the probability of pulmonary embolism (calculator 1).

Tension pneumothorax – Patients with spontaneous pneumothorax present with sudden onset of pleuritic chest pain and dyspnea. Evidence of labored breathing, or accessory muscle use suggest a sizeable pneumothorax. Hemodynamic compromise (eg, tachycardia, hypotension) is an ominous sign and suggests a tension pneumothorax and/or impending cardiopulmonary collapse.

Esophageal rupture, perforation – Spontaneous perforation of the esophagus (Boerhaave syndrome) caused by straining or vomiting presents as excruciating retrosternal chest pain [6].

Cardiac tamponade – In a patient with pericarditis, development of cardiac tamponade can be life-threatening. Symptoms are sudden in onset and include chest pain, tachypnea, and dyspnea. The jugular venous pressure is markedly elevated and may be associated with venous distension in the forehead and scalp. The heart sounds are often muted.