Patients with cardiac diseases or conditions with high risk of developing cardiac diseases undergo risk assessment by cardiologists, primary care physicians, and scientists based on referral for more advanced risk assessment strategies, institution of preventive treatments, counselling of patients and their relatives, and selection of patients for scientific trials. The various methods used for risk assessment differ with respect to availability, complexity, and usefulness in different patient populations. Parameters associated with increased risk of e.g. death may also be associated with higher risk of other adverse outcomes. However, risk assessment strategies including specific methods for risk assessment and risk scores should be used only for the purposes for which they are validated. This expert consensus statement of the European Heart Rhythm Association (EHRA), Heart Rhythm Society (HRS), Asia Pacific Heart Rhythm Society (APHRS), and the Latin American Heart Rhythm Society (LAHRS) summarizes the consensus of the international writing group based on a thorough review of the medical literature regarding risk assessment in cardiac arrhythmias. To create a tool for clinicians to perform rational and evidence-based risk stratification, this task force was set down by EHRA, HRS, LAHRS, and APHRS, including representatives from each of the four societies. With this document, we intend to describe and review status of performing risk assessment in different patient populations with cardiac diseases or conditions with high risk of developing such. Our objectives are to raise awareness of using the right risk assessment tool for a given outcome in a given population, and to provide physicians with practical proposals that may lead to improvement of patient care in this regard. For quick reference, sub-chapters start with a short section on consensus statements. The document concludes with a summary of consensus statements. Members of the Task Force were asked to perform a detailed literature review using PubMed and EMBASE, weigh the strength of evidence for or against a particular treatment or procedure, and include estimates of expected health outcomes for which data exist. Patient-specific modifiers, comorbidities, and issues of patient preference that might influence the choice of particular tests are considered, as are frequency of follow-up and cost-effectiveness. In controversial areas, or with regard to issues without evidence other than usual clinical practice, consensus was achieved by agreement of the expert panel after thorough deliberations. This document was prepared by the Task Force and peer-reviewed by official external reviewers representing EHRA, HRS, APHRS, and LAHRS. Consensus statements are evidence-based and derived primarily from published data or determined through consensus opinion if no data available. Current systems of ranking level of evidence are becoming complicated in a way that might compromise their practical utility.1 In contrast to guidelines, we opted for an easier user-friendly system of ranking using 'coloured hearts' that should allow physicians to easily assess the current status of the evidence and consequent guidance (Table 1). This EHRA grading of consensus statements does not have separate definitions of the level of evidence. The categorization used for consensus statements must not be considered directly similar to the one used for official society guideline recommendations which apply a classification (Class I-III) and level of evidence (A, B, and C) to recommendations used in official guidelines. Thus, a green heart indicates a 'should do this' consensus statement or indicated risk assessment strategy based on at least one randomized trial or supported by strong observational evidence that it is beneficial and effective. A yellow heart indicates general agreement and/or scientific evidence favouring a 'may do this' statement or the usefulness/efficacy of a risk assessment strategy or procedure. A 'yellow heart' symbol may be supported by randomized trials based on a small number of patients or not widely applicable. Risk assessment strategies for which there is scientific evidence of no benefit or potential harm and should not be used ('do not do this') are indicated by a red heart. Finally, this consensus document includes evidence and expert opinions from several countries. The risk assessment approaches discussed may therefore include tests not approved by governmental regulatory agencies in all countries. All members of the writing group, as well as reviewers, have disclosed any potential conflicts of interest. Details are available in Supporting Information online. All consensus statements were voted upon by the writing committee independently and reached the predefined level of ≥75% consensus for inclusion in consensus statement tables. Each partner society officially reviewed the document, and all reviewer comments were addressed. Each partner society approved the final document and consensus statements. Clinical assessment of the patient with cardiac arrhythmias starts with a good clinical history and basic investigations for an underlying aetiological factor for the arrhythmia or its associated complication(s). In addition, an assessment of the risks and benefits of any therapeutic intervention should be made, and appropriate management initiated. Following on from clinical history and assessment, there is a proposal toward a more integrated and holistic approach to arrhythmia management, as evident in guidelines. Such an integrated approach requires multidisciplinary teams of healthcare professionals, patient involvement, access to treatment options, and decision-support tools to optimize the patient journey. Many proposals have been made towards the operationalization of such an integrated approach to risk assessment and practical management in cardiac arrhythmias, which has been of varying complexity. As an example, the management of atrial fibrillation (AF) has been simplified into the ABC pathway ('A' Avoid stroke with Anticoagulation; 'B' Better symptom management, with patient-centred and symptom-directed decisions on rate or rhythm control; 'C' Cardiovascular and comorbidity risk management), which has been shown to be associated with improved clinical outcomes and reduced healthcare costs.2-6 This makes a strong argument for using the right approaches and clinical tools for patient assessment, but using them appropriately for the reasons they were first proposed (e.g. stroke risk scores to assess stroke risk, and not other outcomes). Taking AF as an illustrative example with regard to using the right score for the right reason there are many risk factors for stroke (but the more common and validated ones have been used to formulate risk stratification tools).7 The most common in use is the CHA2DS2-VASc score8 but it is not meant to include every possible stroke risk factor, and was designed to be simple, reductionist and practical to help decision-making for stroke risk. As with all clinical scores based on clinical factors, the CHA2DS2-VASc score only performs modestly for predicting high-risk patients who sustain events. The use of more clinical factors and biomarkers improves prediction (at least statistically) but the practical added value is marginal, and less impressive in real-world cohorts.9, 10 Use of simplified scores to artificially categorize patients into low-, moderate- and high-risk strata can be problematic, as in the real-world patients do not necessarily fall into three neat categories of risk. Also, not all risk factors carry equal weight, hence, the move to focus the initial decision-making on identifying low-risk patients who do not need antithrombotic therapy first, following which stroke prevention can be offered to AF patients with ≥1 stroke risk factors.9 Stroke risk is also highly dynamic, and although logistically challenging, a clinical reassessment may be needed every 4-6 months to optimize risk re-assessment.11-13 As the CHA2DS2-VASc is a cluster of common cardiovascular risk factors, it is predictive of death, cardiovascular hospitalizations, and other adverse outcomes that the CHA2DS2-VASc score was not designed for. Also, given that many components of the CHA2DS2-VASc score are associated with incident AF, the CHA2DS2-VASc score is used to predict new onset AF, again something it was not designed for. Another misuse of the CHA2DS2-VASc score is the prediction of bleeding risk. Nevertheless, formal comparisons show that the CHA2DS2-VASc (and older CHA2DS2) score are inferior to a formal bleeding risk score such as the HAS-BLED score, for the prediction of major bleeding in AF patients.14 Indeed, bleeding risk is also highly dynamic, and the appropriate use of bleeding risk scores such as HAS-BLED is to address modifiable bleeding risk factors (e.g. uncontrolled hypertension, labile INR, concomitant aspirin, or NSAID use) then to schedule the 'high risk' patients for early and more frequent follow-up visits (e.g. 4 weeks rather than 4 months).15 Only focusing on modifiable bleeding risk factors is an inferior strategy for bleeding risk assessment, compared to the HAS-BLED score.8 We should use the scores only for the purposes they were designed for. Attention to appropriate methodology, statistics, etc.—as well as other clinical states merits consideration e.g. sudden death prediction (or failed ablation, device infection, etc.), Charlson Comorbidity Index, frailty etc.—but using the right score designed for that purpose. If appropriately used, some of these (simplified) tools help with clinical management. Indeed, the value of a medical test is measured by its accuracy as well as how it impacts medical decisions and ultimately patient health. As medical tests are considered and new ones emerge, they should be considered and evaluated in a framework of accuracy and patient impact.16 A test must not only be accurate, but also feasible. Tests that are difficult to reproduce, subject to technical failures, or difficult to interpret are likely to impact patient care as a consequence of a primary failure to produce a definitive and actionable result. The ECG is the gold standard for risk assessment in patients with or at risk of developing cardiac arrhythmias. The 12-lead ECG is inexpensive and widely available. Risk stratification with the ECG is limited in general by its low positive predictive value (PPV) determined to a large extent by the low prevalence of cardiovascular events in the general population. However, the prognostic significance of the ECG is enhanced in patients with heart disease. The prognostic value of P wave characteristics has been examined in subjects enrolled in clinical trials of AF for prediction of the development of AF, where maximum P wave duration was a significant independent risk marker for the development of AF over 10 years.20 This observation was confirmed by epidemiologic/population studies (including ARIC and the Copenhagen ECG studies) that showed increased risk of AF in patients with prolonged P wave duration and PR interval prolongation,21-23 and summarized in a review by Nikolaidou et al.24 Moreover, a prolonged P wave duration was determined as a sensitive predictor of post-operative AF in patients undergoing coronary artery bypass grafting (CABG).25 The definition of an abnormal P wave varies greatly depending on how it is measured, and definitions vary depending on whether P wave area, duration, terminal forces in lead V1 or signal-averaged P wave are analysed. Abnormal P wave morphology was associated with incident stroke in the Multi-Ethnic Study of Atherosclerosis.26 The prognostic significance of PR interval prolongation, which is variably defined as PR intervals greater than 196-220 milliseconds, is controversial and depends on the patient population studied. Most studies show that PR interval prolongation is not associated with increased mortality in healthy middle-aged individuals during medium term follow-up. On the other hand, a number of reports show worse survival in patients with suspected heart failure (acute and chronic) or heart disease (coronary artery disease [CAD]). Additionally, PR prolongation and P wave prolongation predict increased risk of AF and the greater degrees of PR prolongation and P wave duration predicted higher risks of AF.27, 28 An increased PR interval is also associated with poor cardiovascular outcomes in patients with AF.29 Several studies have shown that PR prolongation in patients undergoing cardiac pacing or receiving cardiac resynchronization therapy (CRT) is an independent predictor of worse prognosis and lower probability of reverse remodelling as well as an increased risk of AF, death, and hospitalization.30, 31 There are no data indicating whether the degree of PR prolongation portends a worse outcome compared to patients who have lesser degrees of PR prolongation, nor is there information on its prognostic value in acute inferior wall myocardial infarction (MI). Over the years, a number of ECG techniques have been developed to assess risk of ventricular tachyarrhythmias (VTs). These have the advantage of being non-invasive and, often, inexpensive. For almost all of these techniques, there are conflicting data, and not one technique has proven beneficial in patients with structural heart disease. Moreover, studies have varied in their reporting of sudden arrhythmic death vs. total mortality. Among the risk predictors shown to have value are QRS widening and fragmentation, QT prolongation, T-wave abnormalities, and ventricular ectopy. Although the prognostic value of each ECG parameter in isolation is limited, in combination with additional ECG, imaging, and genetic testing, these parameters can contribute to effective risk stratification. QRS prolongation has been associated with all-cause mortality in heart failure patients, implantable cardioverter-defibrillator (ICD) shocks, and inducibility of sustained VT. QRS prolongation in patients on Class IC antiarrhythmic drugs is a predictor of pro-arrhythmia, and should be monitored, particularly during exercise. QRS prolongation predicts risk in patients with myotonic dystrophy and in Brugada Syndrome. Additional prognostic information from the QRS is obtained from the signal-averaged ECG, which amplifies the QRS, averages multiple complexes to reduce noise, and filters out the T-wave in order to detect late potentials, and provides evidence of slow conduction substrate that associates with risk of re-entry tachyarrhythmias.17 The signal-averaged ECG has been used to detect risk of ventricular arrhythmias in post-infarction patients, arrhythmogenic cardiomyopathy, and Brugada Syndrome. Although its specificity is limited, its negative predictive value is high, particularly in survivors of inferior wall myocardial infarction. The signal-averaged ECG is not useful in patients with underlying bundle branch block. QRS fragmentation, which includes abnormally notched narrow and wide QRS complexes, is associated with the presence of myocardial scar and is also associated with mortality in patients with cardiomyopathy and with Brugada Syndrome.32 The presence of an unprovoked type 1 Brugada Syndrome pattern is associated with increased risk as is discussed later in the document. Measurement of the QT interval can be complicated by QRS prolongation and by the need to correct for heart rate, as has been described elsewhere.33 Despite these limitations, prolongation of the heart rate-corrected QT interval (QTc) has been associated with mortality in several population studies.34, 35 In congenital long QT syndrome (LQTS), the length of the QT interval is a major predictor of risk of cardiac events, including sudden cardiac death (SCD). When initiating QT-prolonging drugs such as sotalol or dofetilide, a QT interval of 500 milliseconds or higher should prompt reduction or discontinuation of the offending drug(s). This measure of ventricular repolarization heterogeneity is typically defined from the 12-lead ECG as the QTmax − QTmin. It has been used to predict a wide variety of events, including ventricular pro-arrhythmia, VTs, although the sensitivity, specificity, and accuracy are poorly defined and highly dependent on the patient population studied.36 T wave inversions are common and may be non-specific or may signal important abnormalities such as ischaemia or hypertrophy. Widespread deep T wave inversions in combination with QT prolongation, such as may occur in acute stress cardiomyopathy, can be associated with torsades de pointes. Abnormal T wave notching can be a clue to abnormal repolarization and is often seen in patients with QT prolongation. Computerized T-wave analytic techniques such as principal component analysis, T-wave residuum, flatness, asymmetry, and notching have been developed in an effort to detect and quantify abnormal repolarization and may have particular value in identifying patients with LQTS.37, 38 Moreover, it has been shown that adding T-wave morphology characterizations to age, gender, and QTc in a support vector machine model can improve LQTS diagnosis.39 However, these additional analytic techniques are not used in routine clinical practice. The Tpeak-end interval, measured from the peak to the end of the T-wave, thought to reflect heterogeneity of repolarization in the heart, has been associated with arrhythmic risk in various populations.40 However, considerable controversy remains as to how it should be measured and applied.41 T-wave alternans is a beat-to-beat alternation of T wave morphology. When seen with the naked eye, it usually accompanies marked QT prolongation and is a harbinger of immanent torsades de pointes. Analysis of more subtle T-wave alternans has been used for assessing abnormal and heterogeneous repolarization to predict mortality and arrhythmic risk. Abnormal microvolt T-wave alternans assessed using the spectral method during graded exercise has a high negative predictive value and has been used to identify a subgroup of patients with reduced ejection fraction who are not likely to benefit from defibrillator implantation.18 Microvolt T-wave alternans analysis cannot be performed when the rhythm is AF, and patients with ventricular pacing have not been studied extensively. Early repolarization pattern, highly prevalent in the overall population, defined as an elevation of the J point of at least 0.1 mV, may occur in the anteroseptal or inferolateral leads. In 2008, Haissaguerre reported an association of inferolateral early repolarization with increased risk of idiopathic ventricular fibrillation (VF) in a case–control study42 and subsequently confirmed in other case–control studies. Exercise testing or isoproterenol testing improved the pattern of repolarization, and the pattern was accentuated with exposure to beta-adrenergic blockers. In a meta-analysis of population-based studies, inferolateral early repolarization was associated with increased risk of arrhythmic death, but the risk was still quite low in general (70/100 000 patient-years).43 It appears that individuals at highest risk have early repolarization in multiple (especially inferior) leads, with high voltage (at least 0.2 mV), and with notching or horizontal/down-sloping ST segments. Early repolarization is especially prevalent in young men, particularly young black men, and in athletes.44 Because the absolute risk of arrhythmic death is so low, asymptomatic individuals with early repolarization, even those with higher risk ECG patterns, do not require further evaluation except when there is a strong family history of sudden cardiac death or when the J point elevation is associated with Brugada syndrome (discussed later in this document) or short QT syndrome (SQT). In 1984, Bigger et al. showed that ventricular ectopy recorded on a Holter monitor, especially when combined with a low left ventricular ejection fraction (LVEF), predicted a higher risk of mortality in post-infarction patients compared to those without ectopy.45 Non-sustained VT is also associated with increased risk in patients with arrhythmogenic and hypertrophic cardiomyopathy (HCM). Other data that can be extracted from ambulatory monitoring include heart rate, heart rate variability, and heart rate turbulence measurements, which can predict mortality risk at least in ischaemic cardiomyopathy, but have not been incorporated into clinical practice.19, 46 Evaluation for the presence of structural heart disease (SHD) is important for patients suspected of being at risk for sudden cardiac death. Left ventricular ejection fraction remains the key independent parameter for risk stratification of sudden cardiac death and to guide implantation of an ICD.47, 48 Randomized controlled trials have shown a survival benefit from ICDs in patients with SHD and an EF ≤35%.54-56 Although EF is currently the only proven imaging modality demonstrated to risk stratify for sudden cardiac death, only 1%-5% of patients with ICDs, implanted based upon a low EF, require therapies each year and the large majority of patients who receive ICDs will not have ICD therapies over the 3-year period after implantation.57, 58 In addition, up to 70% of all sudden cardiac deaths in the community occur in individuals with EF >35%.58-60 Although the Efficacy of ICDs in Patients with Non-ischaemic Systolic Heart Failure (DANISH) trial showed that primary prevention ICD in the setting of severe non-ischaemic cardiomyopathy did not reduce all-cause mortality in patients on optimal medical therapy for heart failure, ICD implantation was associated with a 50% reduction in arrhythmic death. Of note, within this non-ischaemic cardiomyopathy population, younger patients (<68 years old) experienced a mortality benefit of 36% if treated with an ICD.61 Ejection fraction is most readily evaluated with echocardiography (recommendation level: green), given both lower cost, availability of equipment, and available expertise; however, cardiac MRI or CT can also be used to evaluate EF and SHD, particularly if obtained in combination of other assessment aims, such as CAD or if there is controversy over the quantified EF with echo (recommendation level: green). The imaging modality used to estimate EF has not been shown to determine benefit from ICD.48 Additional parameters beyond EF remain to be tested in large studies. Cardiac MRI with late gadolinium enhancement (LGE) can provide important prognostic information and may allow for more accurate assessment of scar. Presence and location of scar can portend a higher risk of sustained VT.49-51, 62, 63 In a study of 452 non-ischaemic cardiomyopathy patients with New York Heart Association Class II or II and EF <35%, ICD implantation was only associated with reduced mortality in the population that had presence of scar on cardiac MRI.64 Cardiac positron emission tomography (PET) may elucidate areas of inflammation which may identify inflammatory cardiomyopathies and sarcoidosis, a condition that is associated with higher risk of ventricular arrhythmias in patients without CAD (increased F-2-fluorodeoxyglucose uptake) or can be used to identify sympathetic denervation (carbon-11-metahydroxyephedrine imaging) or regions of inflammation. Greater sympathetic denervation on PET in a prospective study of ischaemic cardiomyopathy patients was a better predictor of ICD shocks than EF.65 Uptake of iodine-123 meta-iodobenzylguanidine (MIBG) to evaluate heart to mediastinum ration (H/M ratio) has shown mixed results in predicting arrhythmic death with some studies suggesting additional prognostic benefit for this parameter, while others have not demonstrated additional value.66, 67 Importantly, the value of these additional parameters in determining risk of sustained VT, VF, or benefit from ICD in various population remains to be clarified. Finally, routine use of viability assessment using PET to guide revascularization in order to reduce risk of SCD remains an area of investigation. In patients with an EF <35% and CAD amenable to revascularization, routine use of PET to guide revascularization was not beneficial in reducing overall mortality.68 Echocardiography (transthoracic or transoesophageal) is a valuable tool in patients who present with atrial arrhythmias, specifically atrial flutter and AF, to evaluate for the presence of structural heart disease, left atrial enlargement, and valvular heart disease in order to better define treatment options. Cardiac MRI or CT may also be used if images obtained at echocardiography are not reliable. However, routine use of echocardiography, including atrial strain or atrial function in patients who do not have atrial arrhythmias to assess risk for the development of AF or atrial flutter is not warranted, unless other structural cardiac abnormalities are suspected. Currently, there are a few indications to perform an electrophysiological study (EPS) to further assess the risk of arrhythmias in at-risk cardiac patients. Such patients include those with structural heart disease, LVEF >35%, pre-syncope, syncope, palpitations, or markedly abnormal ECG suggesting severe conduction disease. These patients can be considered for an EPS to assess the risk of ventricular arrhythmias and sudden cardiac death to decide on need of an ICD, or to identify conduction disturbances or supraventricular tachycardias that can be treated with ablation or pacing.70, 71 Patients with ischaemic cardiomyopathy without a primary indication for an ICD, EF ≤40%, and non-sustained VT on ambulatory cardiac monitoring are candidates for an EPS according to the findings in the MUSTT trial,73 in which, 35% of patients with inducible sustained VT had a significantly lower risk of death with an ICD.66 The MADIT trial initially also utilized an EPS in post-MI patients with an EF ≤30%, and non-sustained VT events to implant an ICD, and showed survival benefit with the ICD.54 However, MADIT-II subsequently eliminated the need for an EPS in post-MI patients with an EF ≤30% and similarly showed the life-saving benefit of the ICD in a broader patient cohort.55 Therefore, post-MI patients with an EF ≤30% do not currently need to undergo an EPS to guide decisions on whether to implant an ICD. In patients with heart failure and EF ≤35%, an EPS is not recommended for risk assessment for the decision on ICD indication. Some centres perform an EPS for inducibility to better characterize induced, sustained VT events, and their response to antitachycardia pacing (ATP), which may potentially help to tailor ICD programming. Furthermore, in patients who have syncope of uncertain origin, an EPS could identify ventricular arrhythmias or document electrical conduction disorders.67, 70, 71, 74 In the case of channelopathies, there is no indication for an EPS, except for Brugada syndrome. In Brugada syndrome, EPS may be useful in asymptomatic patients with spontaneous or drug-induced type 1 pattern, especially when there is a family history of sudden death.75-77 The implantable loop recorder (ILR) provides long-term continuous monitoring and improves the diagnosis in patients with unexplained syncope.81 In a meta-analysis of 49 studies that included 4381 participants, the diagnostic yield for the detection of arrhythmogenic syncope was 26.5%.78 Moreover, the CRYSTAL-AF trial80 revealed that the ILR can detect subclinical AF following cryptogenic stroke. Still, any benefit of these findings needs to be confirmed in large randomized trials. Early use of the ILR has been advocated by the European guidelines82 and in the American guidelines following inconclusive non-invasive monitoring.83 The indications for ILR have been expanded in the current guidelines (Table 2). While the ILR can be useful to detect atrial and ventricular arrhythmias, a large cohort study indicated that most of the current use of ILRs is primarily in patients with unexplained syncope (84%), followed by palpitations (13%), and suspected AF (12%).79 Another smaller study specifically studying the risk of SCD and arrhythmias in patients with haemodialysis, found that 20% of these patients had SCD or bradyarrhythmia events necessitating pacemaker implantation, and 33% of these patients had an arrhythmic endpoint. Interestingly, the median time to event was 2.6 years, confirming the need for long-term monitoring. Surprisingly however, bradyarrhythmias were very commonly diagnosed in this cohort suspected to be at high risk for ventricular arrhythmias and sudden cardiac death.84 Further studies are needed to establish the role of ILR in risk stratification. The direct to consumer or wearable technology market, comprised of devices that monitor physiological parameters such as heart rate and sleep pattern, is anticipated to grow to 929 million connected devices by 2021.87 These devices encompass wristbands, glasses, in-ear monitors, chest straps, and smart phone-enabled recording electrode systems or electronic shirts, with varying capacity to monitor heart rate, heart rhythm, blood pressure, physical activity, respiratory rate, blood glucose, and sleep patterns.88-90 For heart rate monitoring, most wearable devices use photoplethysmography (PPG) technology, meaning they are inherently less accurate than conventional electrocardiography monitoring techniques. Accuracy of various devices varies, with correlation to reference standard ECG monitoring ranging from 0.76 to 0.99.91 Recent advances in wearable ECG acquisition include use of direct electrode recording that represents a regulatory approved medical device generating a lead I like rhythm strip, blurring the lines between consumer and medical devices.92 A growing body of evidence suggests that these technologies can be harnessed to facilitate arrhythmia detection in the appropriate context. Although marketed as consumer devic