Misdiagnosis of atrial fibrillation and its clinical consequences

Stakeholder perspectives on contributors to delayed and inaccurate diagnosis of cardiovascular disease and their implications for digital health technologies: a UK-based qualitative study

OBJECTIVE

The aim of this study is to understand stakeholder experiences of diagnosis of cardiovascular disease (CVD) to support the development of technological solutions that meet current needs. Specifically, we aimed to identify challenges in the process of diagnosing CVD, to identify discrepancies between patient and clinician experiences of CVD diagnosis, and to identify the requirements of future health technology solutions intended to improve CVD diagnosis.

DESIGN

Semistructured focus groups and one-to-one interviews to generate qualitative data that were subjected to thematic analysis.

PARTICIPANTS

UK-based individuals (N=32) with lived experience of diagnosis of CVD (n=23) and clinicians with experience in diagnosing CVD (n=9).

RESULTS

We identified four key themes related to delayed or inaccurate diagnosis of CVD: symptom interpretation, patient characteristics, patient-clinician interactions and systemic challenges. Subthemes from each are discussed in depth. Challenges related to time and communication were greatest for both stakeholder groups; however, there were differences in other areas, for example, patient experiences highlighted difficulties with the psychological aspects of diagnosis and interpreting ambiguous symptoms, while clinicians emphasised the role of individual patient differences and the lack of rapport in contributing to delays or inaccurate diagnosis.

CONCLUSIONS

Our findings highlight key considerations when developing digital technologies that seek to improve the efficiency and accuracy of diagnosis of CVD.

Black Box Warning: Large Language Models and the Future of Infectious Diseases Consultation

Large language models (LLMs) are artificial intelligence systems trained by deep learning algorithms to process natural language and generate text responses to user prompts. Some approach physician performance on a range of medical challenges, leading some proponents to advocate for their potential use in clinical consultation and prompting some consternation about the future of cognitive specialties. However, LLMs currently have limitations that preclude safe clinical deployment in performing specialist consultations, including frequent confabulations, lack of contextual awareness crucial for nuanced diagnostic and treatment plans, inscrutable and unexplainable training data and methods, and propensity to recapitulate biases. Nonetheless, considering the rapid improvement in this technology, growing calls for clinical integration, and healthcare systems that chronically undervalue cognitive specialties, it is critical that infectious diseases clinicians engage with LLMs to enable informed advocacy for how they should-and shouldn't-be used to augment specialist care.

Artificial intelligence-enhanced electrocardiography for accurate diagnosis and management of cardiovascular diseases

Electrocardiography (ECG), improved by artificial intelligence (AI), has become a potential technique for the precise diagnosis and treatment of cardiovascular disorders. The conventional ECG is a frequently used, inexpensive, and easily accessible test that offers important information about the physiological and anatomical state of the heart. However, the ECG can be interpreted differently by humans depending on the interpreter's level of training and experience, which could make diagnosis more difficult. Using AI, especially deep learning convolutional neural networks (CNNs), to look at single, continuous, and intermittent ECG leads that has led to fully automated AI models that can interpret the ECG like a human, possibly more accurately and consistently. These AI algorithms are effective non-invasive biomarkers for cardiovascular illnesses because they can identify subtle patterns and signals in the ECG that may not be readily apparent to human interpreters. The use of AI in ECG analysis has several benefits, including the quick and precise detection of problems like arrhythmias, silent cardiac illnesses, and left ventricular failure. It has the potential to help doctors with interpretation, diagnosis, risk assessment, and illness management. Aside from that, AI-enhanced ECGs have been demonstrated to boost the identification of heart failure and other cardiovascular disorders, particularly in emergency department settings, allowing for quicker and more precise treatment options. The use of AI in cardiology, however, has several limitations and obstacles, despite its potential. The effective implementation of AI-powered ECG analysis is limited by issues such as systematic bias. Biases based on age, gender, and race result from unbalanced datasets. A model's performance is impacted when diverse demographics are inadequately represented. Potentially disregarded age-related ECG variations may result from skewed age data in training sets. ECG patterns are affected by physiological differences between the sexes; a dataset that is inclined toward one sex may compromise the accuracy of the others. Genetic variations influence ECG readings, so racial diversity in datasets is significant. Furthermore, issues such as inadequate generalization, regulatory barriers, and interpretability concerns contribute to deployment difficulties. The lack of robustness in models when applied to disparate populations frequently hinders their practical applicability. The exhaustive validation required by regulatory requirements causes a delay in deployment. Difficult models that are not interpretable erode the confidence of clinicians. Diverse dataset curation, bias mitigation strategies, continuous validation across populations, and collaborative efforts for regulatory approval are essential for the successful deployment of AI ECG in clinical settings and must be undertaken to address these issues. To guarantee a safe and successful deployment in clinical practice, the use of AI in cardiology must be done with a thorough understanding of the algorithms and their limits. In summary, AI-enhanced electrocardiography has enormous potential to improve the management of cardiovascular illness by delivering precise and timely diagnostic insights, aiding clinicians, and enhancing patient outcomes. Further study and development are required to fully realize AI's promise for improving cardiology practices and patient care as technology continues to advance.

An Artificial Intelligence Analysis of Electrocardiograms for the Clinical Diagnosis of Cardiovascular Diseases: A Narrative Review

Artificial intelligence (AI) applied to cardiovascular disease (CVD) is enjoying great success in the field of scientific research. Electrocardiograms (ECGs) are the cornerstone form of examination in cardiology and are the most widely used diagnostic tool because they are widely available, inexpensive, and fast. Applications of AI to ECGs, especially deep learning (DL) methods using convolutional neural networks (CNNs), have been developed in many fields of cardiology in recent years. Deep learning methods provide valuable support for rapid ECG interpretation, demonstrating a diagnostic capability overlapping with specialists in the diagnosis of CVD by a classical analysis of macroscopic changes in the ECG trace. Through photoplethysmography, wearable devices can obtain single-derivative ECGs for the recognition of AI-diagnosed arrhythmias. In addition, CNNs have been developed that recognize no macroscopic electrocardiographic changes and can predict, from a 12-lead ECG, atrial fibrillation, even from sinus rhythm; left and right ventricular function; hypertrophic cardiomyopathy; acute coronary syndromes; or aortic stenosis. The fields of application are many, but numerous are the limitations, mainly associated with the reliability of the acquired data, an inability to verify black box processes, and medico-legal and ethical problems. The challenge of modern medicine is to recognize the limitations of AI and overcome them.

Validation of an automated artificial intelligence system for 12‑lead ECG interpretation

BACKGROUND

The electrocardiogram (ECG) is one of the most accessible and comprehensive diagnostic tools used to assess cardiac patients at the first point of contact. Despite advances in computerized interpretation of the electrocardiogram (CIE), its accuracy remains inferior to physicians. This study evaluated the diagnostic performance of an artificial intelligence (AI)-powered ECG system and compared its performance to current state-of-the-art CIE.

METHODS

An AI-powered system consisting of 6 deep neural networks (DNN) was trained on standard 12‑lead ECGs to detect 20 essential diagnostic patterns (grouped into 6 categories: rhythm, acute coronary syndrome (ACS), conduction abnormalities, ectopy, chamber enlargement and axis). An independent test set of ECGs with diagnostic consensus of two expert cardiologists was used as a reference standard. AI system performance was compared to current state-of-the-art CIE. The key metrics used to compare performances were sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and F1 score.

RESULTS

A total of 932,711 standard 12‑lead ECGs from 173,949 patients were used for AI system development. The independent test set pooled 11,932 annotated ECG labels. In all 6 diagnostic categories, the DNNs achieved high F1 scores: Rhythm 0.957, ACS 0.925, Conduction abnormalities 0.893, Ectopy 0.966, Chamber enlargement 0.972, and Axis 0.897. The diagnostic performance of DNNs surpassed state-of-the-art CIE for the 13 out of 20 essential diagnostic patterns and was non-inferior for the remaining individual diagnoses.

CONCLUSIONS

Our results demonstrate the AI-powered ECG model's ability to accurately identify electrocardiographic abnormalities from the 12‑lead ECG, highlighting its potential as a clinical tool for healthcare professionals.

Impact of Computer-Interpreted ECGs on the Accuracy of Healthcare Professionals

The interpretation of electrocardiograms (ECGs) involves a dynamic interplay between computerized ECG interpretation (CEI) software and human overread. However, the impact of computer ECG interpretation on the performance of healthcare professionals remains largely unexplored. The aim of this study was to evaluate the interpretation proficiency of various medical professional groups, with and without access to the CEI report. Healthcare professionals from diverse disciplines, training levels, and countries sequentially interpreted 60 standard 12-lead ECGs, demonstrating both urgent and nonurgent findings. The interpretation process consisted of 2 phases. In the first phase, participants interpreted 30 ECGs with clinical statements. In the second phase, the same 30 ECGs and clinical statements were randomized and accompanied by a CEI report. Diagnostic performance was evaluated based on interpretation accuracy, time per ECG (in seconds [s]), and self-reported confidence (rated 0 [not confident], 1 [somewhat confident], or 2 [confident]). A total of 892 participants from various medical professional groups participated in the study. This cohort included 44 (4.9%) primary care physicians, 123 (13.8%) cardiology fellows-in-training, 259 (29.0%) resident physicians, 137 (15.4%) medical students, 56 (6.3%) advanced practice providers, 82 (9.2%) nurses, and 191 (21.4%) allied health professionals. The inclusion of the CEI was associated with a significant improvement in interpretation accuracy by 15.1% (95% confidence interval, 14.3-16.0; P < 0.001), decrease in interpretation time by 52 s (-56 to -48; P < 0.001), and increase in confidence by 0.06 (0.03-0.09; P = 0.003). Improvement in interpretation accuracy was seen across all professional subgroups, including primary care physicians by 12.9% (9.4-16.3; P = 0.003), cardiology fellows-in-training by 10.9% (9.1-12.7; P < 0.001), resident physicians by 14.4% (13.0-15.8; P < 0.001), medical students by 19.9% (16.8-23.0; P < 0.001), advanced practice providers by 17.1% (13.3-21.0; P < 0.001), nurses by 16.2% (13.4-18.9; P < 0.001), allied health professionals by 15% (13.4-16.6; P < 0.001), physicians by 13.2% (12.2-14.3; P < 0.001), and nonphysicians by 15.6% (14.3-17.0; P < 0.001).CEI integration improves ECG interpretation accuracy, efficiency, and confidence among healthcare professionals.

Assessment of real-time electrocardiogram effects on interpretation quality by emergency physicians

BACKGROUND

Electrocardiogram (ECG) is one of the most commonly performed examinations in emergency medicine. The literature suggests that one-third of ECG interpretations contain errors and can lead to clinical adverse outcomes. The purpose of this study was to assess the quality of real-time ECG interpretation by senior emergency physicians compared to cardiologists and an ECG expert.

METHODS

This was a prospective study in two university emergency departments and one emergency medical service. All ECGs were performed and interpreted over five weeks by a senior emergency physician (EP) and then by a cardiologist using the same questionnaire. In case of mismatch between EP and the cardiologist our expert had the final word. The ratio of agreement between both interpretations and the kappa (k) coefficient characterizing the identification of major abnormalities defined the reading ability of the emergency physicians.

RESULTS

A total of 905 ECGs were analyzed, of which 705 (78%) resulted in a similar interpretation between emergency physicians and cardiologists/expert. However, the interpretations of emergency physicians and cardiologists for the identification of major abnormalities coincided in only 66% (k: 0.59 (95% confidence interval (CI): 0.54-0.65); P-value = 1.64e-92). ECGs were correctly classified by emergency physicians according to their emergency level in 82% of cases (k: 0.73 (95% CI: 0.70-0.77); P-value ≈ 0). Emergency physicians correctly recognized normal ECGs (sensitivity = 0.91).

CONCLUSION

Our study suggested gaps in the identification of major abnormalities among emergency physicians. The initial and ongoing training of emergency physicians in ECG reading deserves to be improved.

A Deep Learning Approach for Atrial Fibrillation Classification Using Multi-Feature Time Series Data from ECG and PPG

Atrial fibrillation is a prevalent cardiac arrhythmia that poses significant health risks to patients. The use of non-invasive methods for AF detection, such as Electrocardiogram and Photoplethysmogram, has gained attention due to their accessibility and ease of use. However, there are challenges associated with ECG-based AF detection, and the significance of PPG signals in this context has been increasingly recognized. The limitations of ECG and the untapped potential of PPG are taken into account as this work attempts to classify AF and non-AF using PPG time series data and deep learning. In this work, we emploted a hybrid deep neural network comprising of 1D CNN and BiLSTM for the task of AF classification. We addressed the under-researched area of applying deep learning methods to transmissive PPG signals by proposing a novel approach. Our approach involved integrating ECG and PPG signals as multi-featured time series data and training deep learning models for AF classification. Our hybrid 1D CNN and BiLSTM model achieved an accuracy of 95% on test data in identifying atrial fibrillation, showcasing its strong performance and reliable predictive capabilities. Furthermore, we evaluated the performance of our model using additional metrics. The precision of our classification model was measured at 0.88, indicating its ability to accurately identify true positive cases of AF. The recall, or sensitivity, was measured at 0.85, illustrating the model's capacity to detect a high proportion of actual AF cases. Additionally, the F1 score, which combines both precision and recall, was calculated at 0.84, highlighting the overall effectiveness of our model in classifying AF and non-AF cases.

Gated CNN-Transformer Network for Automatic Cardiovascular Diagnosis using 12-lead Electrocardiogram

12-lead electrocardiogram (ECG) is a widely used method in the diagnosis of cardiovascular disease (CVD). With the increase in the number of CVD patients, the study of accurate automatic diagnosis methods via ECG has become a research hotspot. The use of deep learning-based methods can reduce the influence of human subjectivity and improve the diagnosis accuracy. In this paper, we propose a 12-lead ECG automatic diagnosis method based on channel features and temporal features fusion. Specifically, we design a gated CNN-Transformer network, in which the CNN block is used to extract signal embeddings to reduce data complexity. The dual-branch transformer structure is used to effectively extract channel and temporal features in low-dimensional embeddings, respectively. Finally, the features from the two branches are fused by the gating unit to achieve automatic CVD diagnosis from 12-lead ECG. The proposed end-to-end approach has more competitive performance than other deep learning algorithms, which achieves an overall diagnostic accuracy of 85.3% in the 12-lead ECG dataset of CPSC-2018.

Heartbeat classification based on single lead-II ECG using deep learning

The analysis and processing of electrocardiogram (ECG) signals is a vital step in the diagnosis of cardiovascular disease. ECG offers a non-invasive and risk-free method for monitoring the electrical activity of the heart that can assist in predicting and diagnosing heart diseases. The manual interpretation of the ECG signals, however, can be challenging and time-consuming even for experts. Machine learning techniques are increasingly being utilized to support the research and development of automatic ECG classification, which has emerged as a prominent area of study. In this paper, we propose a deep neural network model with residual blocks (DNN-RB) to classify cardiac cycles into six ECG beat classes. The MIT-BIH dataset was used to validate the model resulting in a test accuracy of 99.51%, average sensitivity of 99.7%, and average specificity of 98.2%. The DNN-RB method has achieved higher accuracy than other state-of-the-art algorithms tested on the same dataset. The proposed method is effective in the automatic classification of ECG signals and can be used for both clinical and out-of-hospital monitoring and classification combined with a single-lead mobile ECG device. The method has also been integrated into a web application designed to accept digital ECG beats as input for analyses and to display diagnostic results.

Rapid Atrial Fibrillation in the Emergency Department

Atrial fibrillation (AF) is the most common rhythm disorder seen in doctors' offices and emergency departments (EDs). In both settings, an AF holistic pathway including anticoagulation or stroke avoidance, better symptom management, and cardiovascular and comorbidity optimization should be followed. However, other considerations need to be assessed in the ED, such as haemodynamic instability, the onset of AF, the presence of acute heart failure and pre-excitation. Although the Advanced Cardiovascular Life Support guidelines (European Society of Cardiology guidelines, Acute Cardiac Care Association/European Heart Rhythm Association position statements) and several recent AF publications have greatly assisted physicians in treating AF with rapid ventricular response in the ED, further practical clinical guidance is required to improve physicians' skill and knowledge in providing the best treatment for patients. Herein, we combine multiple strategies with supporting evidence-based treatment and experiences encountered in clinical practice into practical stepwise approaches. We hope that the stepwise algorithm may assist residents and physicians in managing AF in the ED.

The Patient Pathway Review for Atrial Fibrillation

The care pathway for patients with atrial fibrillation (AF) is variable and this variability is explored in a patient pathway review. This review describes events that may take place for a patient with AF considering the "ideal" and the "real-world" pathway and attempts to rationalize them by considering the patient, clinician, health service, and societal perspective. In the "ideal" pathway, AF in a patient is either identify before or after stroke. The "real-world" pathway introduces the concepts that symptoms may influence patient decision-making to seek help, AF may be identified incidentally, and healthcare professionals may fail to identify AF. The management of AF includes no treatment or treatment such as stroke prevention, rate or rhythm control, and comorbidity management. The overall outcomes for patient depend on the presence of symptoms and response to therapies. The two major priorities for patients are symptomatic relief and avoidance of stroke. While most clinicians will find that initial AF management is not challenging but there may be incidental opportunities for earlier identification. From the healthcare service perspective, noncardiologists and cardiologists care for patients with AF, which results in much heterogeneity management. From the societal perspective, the burden of AF is significant resulting in substantial cost from hospitalizations and treatments. People with AF can take on different paths, which depend on factors related to the patient's decision-making, clinical decision-making, and patient's response to the treatment. A streamlined approach to a holistic and integrated care pathway approach to AF management is needed.

mCardia: A Context-Aware ECG Collection System for Ambulatory Arrhythmia Screening

This article presents the design, technical implementation, and feasibility evaluation of mCardia —a context-aware, mobile electrocardiogram (ECG) collection system for longitudinal arrhythmia screening under free-living conditions. Along with ECG, mCardia also records active and passive contextual data, including patient-reported symptoms and physical activity. This contextual data can provide a more accurate understanding of what happens before, during, and after an arrhythmia event, thereby providing additional information in the diagnosis of arrhythmia. By using a plugin-based architecture for ECG and contextual sensing, mCardia is device-agnostic and can integrate with various wireless ECG devices and supports cross-platform deployment. We deployed the mCardia system in a feasibility study involving 24 patients who used the system over a two-week period. During the study, we observed high patient acceptance and compliance with a satisfactory yield of collected ECG and contextual data. The results demonstrate the high usability and feasibility of mCardia for longitudinal ambulatory monitoring under free-living conditions. The article also reports from two clinical cases, which demonstrate how a cardiologist can utilize the collected contextual data to improve the accuracy of arrhythmia analysis. Finally, the article discusses the lessons learned and the challenges found in the mCardia design and the feasibility study.

An artificial intelligence-enabled ECG algorithm for comprehensive ECG interpretation: Can it pass the 'Turing test'?

Objective

To develop an artificial intelligence (AI)–enabled electrocardiogram (ECG) algorithm capable of comprehensive, human-like ECG interpretation and compare its diagnostic performance against conventional ECG interpretation methods.

Methods

We developed a novel AI-enabled ECG (AI-ECG) algorithm capable of complete 12-lead ECG interpretation. It was trained on nearly 2.5 million standard 12-lead ECGs from over 720,000 adult patients obtained at the Mayo Clinic ECG laboratory between 2007 and 2017. We then compared the need for human over-reading edits of the reports generated by the Marquette 12SL automated computer program, AI-ECG algorithm, and final clinical interpretations on 500 randomly selected ECGs from 500 patients. In a blinded fashion, 3 cardiac electrophysiologists adjudicated each interpretation as (1) ideal (ie, no changes needed), (2) acceptable (ie, minor edits needed), or (3) unacceptable (ie, major edits needed).

Results

Cardiologists determined that on average 202 (13.5%), 123 (8.2%), and 90 (6.0%) of the interpretations required major edits from the computer program, AI-ECG algorithm, and final clinical interpretations, respectively. They considered 958 (63.9%), 1058 (70.5%), and 1118 (74.5%) interpretations as ideal from the computer program, AI-ECG algorithm, and final clinical interpretations, respectively. They considered 340 (22.7%), 319 (21.3%), and 292 (19.5%) interpretations as acceptable from the computer program, AI-ECG algorithm, and final clinical interpretations, respectively.

Conclusion

An AI-ECG algorithm outperforms an existing standard automated computer program and better approximates expert over-read for comprehensive 12-lead ECG interpretation.

Artificial Intelligence and Atrial Fibrillation

In the context of atrial fibrillation (AF), traditional clinical practices have thus far fallen short in several domains such as identifying patients at risk of incident AF or patients with concomitant undetected paroxysmal AF. Novel approaches leveraging artificial intelligence have the potential to provide new tools to deal with some of these old problems. In this review we focus on the roles of artificial intelligence-enabled ECG pertaining to AF, potential roles of deep learning (DL) models in the context of current knowledge gaps, as well as limitations of these models. One key area where DL models can translate to better patient outcomes is through automated ECG interpretation. Further, we overview some of the challenges facing AF screening and the harms and benefits of screening. In this context, a unique model was developed to detect underlying hidden AF from sinus rhythm and is discussed in detail with its potential uses. Knowledge gaps also remain regarding the best ways to monitor patients with embolic stroke of undetermined source (ESUS) and who would benefit most from oral anticoagulation. The AI-enabled AF model is one potential way to tackle this complex problem as it could be used to identify a subset of high-risk ESUS patients likely to benefit from empirical oral anticoagulation. Role of DL models assessing AF burden from long duration ECG data is also discussed as a way of guiding management. There is a trend towards the use of consumer-grade wristbands and watches to detect AF from photoplethysmography data. However, ECG currently remains the gold standard to detect arrythmias including AF. Lastly, role of adequate external validation of the models and clinical trials to study true performance is discussed. This article is protected by copyright. All rights reserved.

Inconsistent Methodology as a Barrier to Meaningful Research Outputs From Studies of Atrial Fibrillation After Cardiac Surgery

Atrial fibrillation after cardiac surgery (AFACS) is a serious postoperative complication. There is significant research interest in this field but also relevant heterogeneity in reported AFACS definitions and approaches used for its identification. Few data exist on the extent of this variation in clinical studies. The authors reviewed the literature since 2001 and included manuscripts reporting outcomes of AFACS in adults. They excluded smaller studies and studies in which patients did not undergo a sternotomy. The documented protocol in each manuscript was analyzed according to six different categories to determine how AFACS was defined, which techniques were used to identify it, and the inclusion and/or exclusion criteria. They also noted when a category was not described in the documented protocol. The authors identified 302 studies, of which 92 were included. Sixty-two percent of studies were randomized controlled trials. There was significant heterogeneity in the manuscripts, including the exclusion of patients with preoperative AF, the definition and duration of AF needed to meet the primary endpoint, the type of screening approach (continuous, episodic, or opportunistic), the duration of monitoring during the study period in days, the diagnosis with predefined electrocardiogram criteria, and the requirement for independent confirmation by study investigators. Furthermore, the definitions of these criteria frequently were not described. Consistent reporting standards for AFACS research are needed to advance scientific progress in the field. The authors here propose pragmatic standards for trial design and reporting standards. These include adequate sample size estimation, a clear definition of the AFACS endpoints, and a protocol for AFACS detection.

Screening for Atrial Fibrillation: Updated Evidence Report and Systematic Review for the US Preventive Services Task Force

IMPORTANCE

Atrial fibrillation (AF), the most common arrhythmia, increases the risk of stroke.

OBJECTIVE

To review the evidence on screening for AF in adults without prior stroke to inform the US Preventive Services Task Force.

DATA SOURCES

PubMed, Cochrane Library, and trial registries through October 5, 2020; references, experts, and literature surveillance through October 31, 2021.

STUDY SELECTION

Randomized clinical trials (RCTs) of screening among asymptomatic persons without known AF or prior stroke; test accuracy studies; RCTs of anticoagulation among persons with AF; systematic reviews; and observational studies reporting harms.

DATA EXTRACTION AND SYNTHESIS

Two reviewers assessed titles/abstracts, full-text articles, and study quality and extracted data; when at least 3 similar studies were available, meta-analyses were conducted.

MAIN OUTCOMES AND MEASURES

Detection of undiagnosed AF, test accuracy, mortality, stroke, stroke-related morbidity, and harms.

RESULTS

Twenty-six studies (N = 113 784) were included. In 1 RCT (n = 28 768) of twice-daily electrocardiography (ECG) screening for 2 weeks, the likelihood of a composite end point (ischemic stroke, hemorrhagic stroke, systemic embolism, all-cause mortality, and hospitalization for bleeding) was lower in the screened group over 6.9 years (hazard ratio, 0.96 [95% CI, 0.92-1.00]; P = .045), but that study had numerous limitations. In 4 RCTs (n = 32 491), significantly more AF was detected with intermittent and continuous ECG screening compared with no screening (risk difference range, 1.0%-4.8%). Treatment with warfarin over a mean of 1.5 years in populations with clinical, mostly persistent AF was associated with fewer ischemic strokes (pooled risk ratio [RR], 0.32 [95% CI, 0.20-0.51]; 5 RCTs; n = 2415) and lower all-cause mortality (pooled RR, 0.68 [95% CI, 0.50-0.93]) compared with placebo. Treatment with direct oral anticoagulants was also associated with lower incidence of stroke (adjusted odds ratios range, 0.32-0.44) in indirect comparisons with placebo. The pooled RR for major bleeding for warfarin compared with placebo was 1.8 (95% CI, 0.85-3.7; 5 RCTs; n = 2415), and the adjusted odds ratio for major bleeding for direct oral anticoagulants compared with placebo or no treatment ranged from 1.38 to 2.21, but CIs did not exclude a null effect.

CONCLUSIONS AND RELEVANCE

Although screening can detect more cases of unknown AF, evidence regarding effects on health outcomes is limited. Anticoagulation was associated with lower risk of first stroke and mortality but with increased risk of major bleeding, although estimates for this harm are imprecise; no trials assessed benefits and harms of anticoagulation among screen-detected populations.

Continuous ECG monitoring versus mobile telemetry: A comparison of arrhythmia diagnostics in human- versus algorithmic-dependent systems

Background

Clinicians rarely scrutinize the full disclosure of a myriad of FDA-approved long-term rhythm monitors, and they rely on manufacturers to detect and report relevant rhythm abnormalities.

Objective

The objective of this study is to compare the diagnostic accuracy between mobile cardiac telemetry (MCT), which uses an algorithm-based detection strategy, and continuous long-term electrocardiography (LT-ECG) monitoring, which uses a human-based detection strategy.

Methods

In an outpatient arrhythmia clinic, we enrolled 50 sequential patients ordered to wear a 30-day MCT, to simultaneously wear a continuous LT-ECG monitor. Periods of concomitant wear of both devices were examined using the associated report, which was over-read by 2 electrophysiologists.

Results

Forty-six of 50 patients wore both monitors simultaneously for an average of 10.3 ± 4.4 days (range: 1.2–14.8 days). During simultaneous recording, patients were more often diagnosed with arrhythmia by LT-ECG compared to MCT (23/46 vs 11/46), P = .018. Similarly, more arrhythmia episodes were detected during simultaneous recording with the LT-ECG compared to MCT (61 vs 19), P < .001. This trend remained consistent across arrhythmia subtypes, including ventricular tachycardia (13 patients by LT-ECG vs 7 by MCT), atrioventricular (AV) block (3 patients by LT-ECG vs 0 by MCT), and AV node reentrant tachycardia (2 patients by LT-ECG vs 0 by MCT). Atrial fibrillation (AF) was documented by both monitors in 2 patients; however, LT-ECG monitoring captured 4 additional AF episodes missed by MCT.

Conclusion

In a time-controlled, paired analysis of 2 disparate rhythm monitors worn simultaneously, human-dependent LT-ECG arrhythmia detection significantly outperformed algorithm-based MCT arrhythmia detection.

Expert-enhanced machine learning for cardiac arrhythmia classification

We propose a new method for the classification task of distinguishing atrial fibrillation (AFib) from regular atrial tachycardias including atrial flutter (AFlu) based on a surface electrocardiogram (ECG). Recently, many approaches for an automatic classification of cardiac arrhythmia were proposed and to our knowledge none of them can distinguish between these two. We discuss reasons why deep learning may not yield satisfactory results for this task. We generate new and clinically interpretable features using mathematical optimization for subsequent use within a machine learning (ML) model. These features are generated from the same input data by solving an additional regression problem with complicated combinatorial substructures. The resultant can be seen as a novel machine learning model that incorporates expert knowledge on the pathophysiology of atrial flutter. Our approach achieves an unprecedented accuracy of 82.84% and an area under the receiver operating characteristic (ROC) curve of 0.9, which classifies as "excellent" according to the classification indicator of diagnostic tests. One additional advantage of our approach is the inherent interpretability of the classification results. Our features give insight into a possibly occurring multilevel atrioventricular blocking mechanism, which may improve treatment decisions beyond the classification itself. Our research ideally complements existing textbook cardiac arrhythmia classification methods, which cannot provide a classification for the important case of AFib↔AFlu. The main contribution is the successful use of a novel mathematical model for multilevel atrioventricular block and optimization-driven inverse simulation to enhance machine learning for classification of the arguably most difficult cases in cardiac arrhythmia. A tailored Branch-and-Bound algorithm was implemented for the domain knowledge part, while standard algorithms such as Adam could be used for training.

Analyzing artificial intelligence systems for the prediction of atrial fibrillation from sinus-rhythm ECGs including demographics and feature visualization

Atrial fibrillation (AF) is an abnormal heart rhythm, asymptomatic in many cases, that causes several health problems and mortality in population. This retrospective study evaluates the ability of different AI-based models to predict future episodes of AF from electrocardiograms (ECGs) recorded during normal sinus rhythm. Patients are divided into two classes according to AF occurrence or sinus rhythm permanence along their several ECGs registry. In the constrained scenario of balancing the age distributions between classes, our best AI model predicts future episodes of AF with area under the curve (AUC) 0.79 (0.72–0.86). Multiple scenarios and age-sex-specific groups of patients are considered, achieving best performance of prediction for males older than 70 years. These results point out the importance of considering different demographic groups in the analysis of AF prediction, showing considerable performance gaps among them. In addition to the demographic analysis, we apply feature visualization techniques to identify the most important portions of the ECG signals in the task of AF prediction, improving this way the interpretability and understanding of the AI models. These results and the simplicity of recording ECGs during check-ups add feasibility to clinical applications of AI-based models.

ECG Interpretation: Clinical Relevance, Challenges, and Advances

Since its inception, the electrocardiogram (ECG) has been an essential tool in medicine. The ECG is more than a mere tracing of cardiac electrical activity; it can detect and diagnose various pathologies including arrhythmias, pericardial and myocardial disease, electrolyte disturbances, and pulmonary disease. The ECG is a simple, non-invasive, rapid, and cost-effective diagnostic tool in medicine; however, its clinical utility relies on the accuracy of its interpretation. Computer ECG analysis has become so widespread and relied upon that ECG literacy among clinicians is waning. With recent technological advances, the application of artificial intelligence-augmented ECG (AI-ECG) algorithms has demonstrated the potential to risk stratify, diagnose, and even interpret ECGs—all of which can have a tremendous impact on patient care and clinical workflow. In this review, we examine (i) the utility and importance of the ECG in clinical practice, (ii) the accuracy and limitations of current ECG interpretation methods, (iii) existing challenges in ECG education, and (iv) the potential use of AI-ECG algorithms for comprehensive ECG interpretation.

ECG Signal Classification Using Deep Learning Techniques Based on the PTB-XL Dataset

The analysis and processing of ECG signals are a key approach in the diagnosis of cardiovascular diseases. The main field of work in this area is classification, which is increasingly supported by machine learning-based algorithms. In this work, a deep neural network was developed for the automatic classification of primary ECG signals. The research was carried out on the data contained in a PTB-XL database. Three neural network architectures were proposed: the first based on the convolutional network, the second on SincNet, and the third on the convolutional network, but with additional entropy-based features. The dataset was divided into training, validation, and test sets in proportions of 70%, 15%, and 15%, respectively. The studies were conducted for 2, 5, and 20 classes of disease entities. The convolutional network with entropy features obtained the best classification result. The convolutional network without entropy-based features obtained a slightly less successful result, but had the highest computational efficiency, due to the significantly lower number of neurons.

Clinically contextualised ECG interpretation: the impact of prior clinical exposure and case vignettes on ECG diagnostic accuracy

BACKGROUND

ECGs are often taught without clinical context. However, in the clinical setting, ECGs are rarely interpreted without knowing the clinical presentation. We aimed to determine whether ECG diagnostic accuracy was influenced by knowledge of the clinical context and/or prior clinical exposure to the ECG diagnosis.

METHODS

Fourth- (junior) and sixth-year (senior) medical students, as well as medical residents were invited to complete two multiple-choice question (MCQ) tests and a survey. Test 1 comprised 25 ECGs without case vignettes. Test 2, completed immediately thereafter, comprised the same 25 ECGs and MCQs, but with case vignettes for each ECG. Subsequently, participants indicated in the survey when last, during prior clinical clerkships, they have seen each of the 25 conditions tested. Eligible participants completed both tests and survey. We estimated that a minimum sample size of 165 participants would provide 80% power to detect a mean difference of 7% in test scores, considering a type 1 error of 5%.

RESULTS

This study comprised 176 participants (67 [38.1%] junior students, 55 [31.3%] senior students, 54 [30.7%] residents). Prior ECG exposure depended on their level of training, i.e., junior students were exposed to 52% of the conditions tested, senior students 63.4% and residents 96.9%. Overall, there was a marginal improvement in ECG diagnostic accuracy when the clinical context was known (Cohen's d = 0.35, p < 0.001). Gains in diagnostic accuracy were more pronounced amongst residents (Cohen's d = 0.59, p < 0.001), than senior (Cohen's d = 0.38, p < 0.001) or junior students (Cohen's d = 0.29, p < 0.001). All participants were more likely to make a correct ECG diagnosis if they reported having seen the condition during prior clinical training, whether they were provided with a case vignette (odds ratio [OR] 1.46, 95% confidence interval [CI] 1.24-1.71) or not (OR 1.58, 95% CI 1.35-1.84).

CONCLUSION

ECG interpretation using clinical vignettes devoid of real patient experiences does not appear to have as great an impact on ECG diagnostic accuracy as prior clinical exposure. However, exposure to ECGs during clinical training is largely opportunistic and haphazard. ECG training should therefore not rely on experiential learning alone, but instead be supplemented by other formal methods of instruction.

Interpretable deep learning for automatic diagnosis of 12-lead electrocardiogram

Electrocardiogram (ECG) is a widely used reliable, non-invasive approach for cardiovascular disease diagnosis. With the rapid growth of ECG examinations and the insufficiency of cardiologists, accurate and automatic diagnosis of ECG signals has become a hot research topic. In this paper, we developed a deep neural network for automatic classification of cardiac arrhythmias from 12-lead ECG recordings. Experiments on a public 12-lead ECG dataset showed the effectiveness of our method. The proposed model achieved an average F1 score of 0.813. The deep model showed superior performance than 4 machine learning methods learned from extracted expert features. Besides, the deep models trained on single-lead ECGs produce lower performance than using all 12 leads simultaneously. The best-performing leads are lead I, aVR, and V5 among 12 leads. Finally, we employed the SHapley Additive exPlanations method to interpret the model's behavior at both the patient level and population level.

•

We develop a deep learning model for the automatic diagnosis of ECG

•

We present benchmark results of 12-lead ECG classification

•

We find out the top performance single lead in diagnosing ECGs

•

We employ the SHAP method to enhance clinical interpretability

Effectiveness of blended learning versus lectures alone on ECG analysis and interpretation by medical students

BACKGROUND

Most medical students lack confidence and are unable to accurately interpret ECGs. Thus, better methods of ECG instruction are being sought. Current literature indicates that the use of e-learning for ECG analysis and interpretation skills (ECG competence) is not superior to lecture-based teaching. We aimed to assess whether blended learning (lectures supplemented with the use of a web application) resulted in better acquisition and retention of ECG competence in medical students, compared to conventional teaching (lectures alone).

METHODS

Two cohorts of fourth-year medical students were studied prospectively. The conventional teaching cohort (n = 67) attended 4 hours of interactive lectures, covering the basic principles of Electrocardiography, waveform abnormalities and arrhythmias. In addition to attending the same lectures, the blended learning cohort (n = 64) used a web application that facilitated deliberate practice of systematic ECG analysis and interpretation, with immediate feedback. All participants completed three tests: pre-intervention (assessing baseline ECG competence at start of clinical clerkship), immediate post-intervention (assessing acquisition of ECG competence at end of six-week clinical clerkship) and delayed post-intervention (assessing retention of ECG competence 6 months after clinical clerkship, without any further ECG training). Diagnostic accuracy and uncertainty were assessed in each test.

RESULTS

The pre-intervention test scores were similar for blended learning and conventional teaching cohorts (mean 31.02 ± 13.19% versus 31.23 ± 11.52% respectively, p = 0.917). While all students demonstrated meaningful improvement in ECG competence after teaching, blended learning was associated with significantly better scores, compared to conventional teaching, in immediate (75.27 ± 16.22% vs 50.27 ± 17.10%, p <  0.001; Cohen's d = 1.58), and delayed post-intervention tests (57.70 ± 18.54% vs 37.63 ± 16.35%, p <  0.001; Cohen's d = 1.25). Although diagnostic uncertainty decreased after ECG training in both cohorts, blended learning was associated with better confidence in ECG analysis and interpretation.

CONCLUSION

Blended learning achieved significantly better levels of ECG competence and confidence amongst medical students than conventional ECG teaching did. Although medical students underwent significant attrition of ECG competence without ongoing training, blended learning also resulted in better retention of ECG competence than conventional teaching. Web applications encouraging a stepwise approach to ECG analysis and enabling deliberate practice with feedback may, therefore, be a useful adjunct to lectures for teaching Electrocardiography.

Clinical classification and the subclinical atrial fibrillation challenge: a position paper of the European Cardiac Arrhythmia Society

Symptomatic atrial fibrillation (AF) or clinical AF is associated with impaired quality of life, higher risk of stroke, heart failure, and increased mortality. Current clinical classification of AF is based on the duration of AF episodes and the recurrence over time. Appropriate management strategy should follow guidelines of Scientific Societies. The last decades have been marked by the advances in mechanism comprehension, better management of symptomatic AF, particularly regarding stroke prevention with the use of direct oral anticoagulants and a wider use of AF catheter or surgical ablations. The advent of new tools for detection of asymptomatic AF including continuous monitoring with implanted electronic devices and the use of implantable cardiac monitors and recently wearable devices or garments have identified what is called "subclinical AF" encompassing atrial high-rate episodes (AHREs). New concepts such as "AF burden" have resulted in new management challenges. Oral anticoagulation has proven to reduce substantially stroke risk in patients with symptomatic clinical AF but carries the risk of bleeding. Management of detected asymptomatic atrial arrhythmias and their relation to clinical AF and stroke risk is currently under evaluation. Based on a review of recent literature, the validity of current clinical classification has been reassessed and appropriate updates are proposed. Current evidence supporting the inclusion of subclinical AF within current clinical classification is discussed as well as the need for controlled trials which may provide responses to current therapeutic challenges particularly regarding the subsets of asymptomatic AF patients that might benefit from oral anticoagulation.

A comprehensive artificial intelligence–enabled electrocardiogram interpretation program

Background

Automated computerized electrocardiogram (ECG) interpretation algorithms are designed to enhance physician ECG interpretation, minimize medical error, and expedite clinical workflow. However, the performance of current computer algorithms is notoriously inconsistent. We aimed to develop and validate an artificial intelligence–enabled ECG (AI-ECG) algorithm capable of comprehensive 12-lead ECG interpretation with accuracy comparable to practicing cardiologists.

Methods

We developed an AI-ECG algorithm using a convolutional neural network as a multilabel classifier capable of assessing 66 discrete, structured diagnostic ECG codes using the cardiologist’s final annotation as the gold-standard interpretation. We included 2,499,522 ECGs from 720,978 patients ≥18 years of age with a standard 12-lead ECG obtained at the Mayo Clinic ECG laboratory between 1993 and 2017. The total sample was randomly divided into training (n = 1,749,654), validation (n = 249,951), and testing (n = 499,917) datasets with a similar distribution of codes. We compared the AI-ECG algorithm’s performance to the cardiologist’s interpretation in the testing dataset using receiver operating characteristic (ROC) and precision recall (PR) curves.

Results

The model performed well for various rhythm, conduction, ischemia, waveform morphology, and secondary diagnoses codes with an area under the ROC curve of ≥0.98 for 62 of the 66 codes. PR metrics were used to assess model performance accounting for category imbalance and demonstrated a sensitivity ≥95% for all codes.

Conclusions

An AI-ECG algorithm demonstrates high diagnostic performance in comparison to reference cardiologist interpretation of a standard 12-lead ECG. The use of AI-ECG reading tools may permit scalability as ECG acquisition becomes more ubiquitous.

Automatic Triage of 12-Lead ECGs Using Deep Convolutional Neural Networks

BACKGROUND

The correct interpretation of the ECG is pivotal for the accurate diagnosis of many cardiac abnormalities, and conventional computerized interpretation has not been able to reach physician‐level accuracy in detecting (acute) cardiac abnormalities. This study aims to develop and validate a deep neural network for comprehensive automated ECG triage in daily practice.

METHODS AND RESULTS

We developed a 37‐layer convolutional residual deep neural network on a data set of free‐text physician‐annotated 12‐lead ECGs. The deep neural network was trained on a data set with 336.835 recordings from 142.040 patients and validated on an independent validation data set (n=984), annotated by a panel of 5 cardiologists electrophysiologists. The 12‐lead ECGs were acquired in all noncardiology departments of the University Medical Center Utrecht. The algorithm learned to classify these ECGs into the following 4 triage categories: normal, abnormal not acute, subacute, and acute. Discriminative performance is presented with overall and category‐specific concordance statistics, polytomous discrimination indexes, sensitivities, specificities, and positive and negative predictive values. The patients in the validation data set had a mean age of 60.4 years and 54.3% were men. The deep neural network showed excellent overall discrimination with an overall concordance statistic of 0.93 (95% CI, 0.92–0.95) and a polytomous discriminatory index of 0.83 (95% CI, 0.79–0.87).

CONCLUSIONS

This study demonstrates that an end‐to‐end deep neural network can be accurately trained on unstructured free‐text physician annotations and used to consistently triage 12‐lead ECGs. When further fine‐tuned with other clinical outcomes and externally validated in clinical practice, the demonstrated deep learning–based ECG interpretation can potentially improve time to treatment and decrease healthcare burden.

Interpretations of and management actions following ECGs in programmatic cardiovascular care in primary care: A retrospective dossier study

Background

The usefulness of routine electrocardiograms (ECGs) in cardiovascular risk management (CVRM) and diabetes care is doubted.

Objectives

To assess the performance of general practitioners (GPs) in embedding ECGs in CVRM and diabetes care.

Methods

We collected 852 ECGs recorded by 20 GPs (12 practices) in the context of CVRM and diabetes care. Of all abnormal (n = 265) and a sample of the normal (n = 35) ECGs, data on the indications, interpretations and management actions were extracted from the corresponding medical records. An expert panel consisting of one cardiologist and one expert GP reviewed these 300 ECG cases.

Results

GPs found new abnormalities in 13.0% of all 852 ECGs (12.0% in routinely recorded ECGs versus 24.3% in ECGs performed for a specific indication). Management actions followed more often after ECGs performed for specific indications (17.6%) than after routine ECGs (6.0%). The expert panel agreed with the GPs’ interpretations in 67% of the 300 assessed cases. Most often misinterpreted relevant ECG abnormalities were previous myocardial infarction, R‑wave abnormalities and typical/atypical ST-segment and T‑wave (ST-T) abnormalities. Agreement on patient management between GP and expert panel was 74%. Disagreement in most cases concerned additional diagnostic testing.

Conclusions

In the context of programmatic CVRM and diabetes care by GPs, the yield of newly found ECG abnormalities is modest. It is higher for ECGs recorded for a specific reason. Educating GPs seems necessary in this field since they perform less well in interpreting and managing CVRM ECGs than in ECGs performed in symptomatic patients.

Electronic supplementary material

The online version of this article (10.1007/s12471-020-01376-3) contains supplementary material, which is available to authorized users.

Erroneous computer-based interpretations of atrial fibrillation and atrial flutter in a Swedish primary health care setting

Objective: To describe the incidence of incorrect computerized ECG interpretations of atrial fibrillation or atrial flutter in a Swedish primary care population, the rate of correction of computer misinterpretations, and the consequences of misdiagnosis.Design: Retrospective expert re-analysis of ECGs with a computer-suggested diagnosis of atrial fibrillation or atrial flutter.Setting: Primary health care in Region Kronoberg, Sweden.Subjects: All adult patients who had an ECG recorded between January 2016 and June 2016 with a computer statement including the words 'atrial fibrillation' or 'atrial flutter'.Main outcome measures: Number of incorrect computer interpretations of atrial fibrillation or atrial flutter; rate of correction by the interpreting primary care physician; consequences of misdiagnosis of atrial fibrillation or atrial flutter.Results: Among 988 ECGs with a computer diagnosis of atrial fibrillation or atrial flutter, 89 (9.0%) were incorrect, among which 36 were not corrected by the interpreting physician. In 12 cases, misdiagnosed atrial fibrillation/flutter led to inappropriate treatment with anticoagulant therapy. A larger proportion of atrial flutters, 27 out of 80 (34%), than atrial fibrillations, 62 out of 908 (7%), were incorrectly diagnosed by the computer.Conclusions: Among ECGs with a computer-based diagnosis of atrial fibrillation or atrial flutter, the diagnosis was incorrect in almost 10%. In almost half of the cases, the misdiagnosis was not corrected by the overreading primary-care physician. Twelve patients received inappropriate anticoagulant treatment as a result of misdiagnosis.Key pointsData regarding the incidence of misdiagnosed atrial fibrillation or atrial flutter in primary care are lacking. In a Swedish primary care setting, computer-based ECG interpretations of atrial fibrillation or atrial flutter were incorrect in 89 of 988 (9.0%) consecutive cases.Incorrect computer diagnoses of atrial fibrillation or atrial flutter were not corrected by the primary-care physician in 47% of cases.In 12 of the cases with an incorrect computer rhythm diagnosis, misdiagnosed atrial fibrillation or flutter led to inappropriate treatment with anticoagulant therapy.

Wearing Your Heart on Your Sleeve: the Future of Cardiac Rhythm Monitoring

PURPOSE OF REVIEW

This review describes the novel category of wearable ECG monitors and identifies where patients, healthcare providers, and device manufacturers should focus efforts to maximize the clinical benefit of these devices.

RECENT FINDINGS

Notable wearable ECG monitors include the AliveCor Kardia devices, Apple Watch Series 4, and several others. The most common use case is monitoring for atrial fibrillation. The available evidence validates the ability of the Kardia devices and Apple Watch to distinguish atrial fibrillation from sinus rhythm. Key questions for manufacturers include how to calibrate each device's algorithms and streamline workflows for healthcare providers. Wearable ECG monitors are currently most useful to detect atrial fibrillation. Further study is needed to demonstrate whether wearable ECG monitors improve patient outcomes, and to expand their use into other indications. Device manufacturers and healthcare providers must work together to establish new workflows to process and act on wearable ECG data.

A deep neural network for 12-lead electrocardiogram interpretation outperforms a conventional algorithm, and its physician overread, in the diagnosis of atrial fibrillation

Background

Automated electrocardiogram (ECG) interpretations may be erroneous, and lead to erroneous overreads, including for atrial fibrillation (AF). We compared the accuracy of the first version of a new deep neural network 12-Lead ECG algorithm (Cardiologs®) to the conventional Veritas algorithm in interpretation of AF.

Methods

24,123 consecutive 12-lead ECGs recorded over 6 months were interpreted by 1) the Veritas® algorithm, 2) physicians who overread Veritas® (Veritas® + physician), and 3) Cardiologs® algorithm. We randomly selected 500 out of 858 ECGs with a diagnosis of AF according to either algorithm, then compared the algorithms' interpretations, and Veritas® + physician, with expert interpretation. To assess sensitivity for AF, we analyzed a separate database of 1473 randomly selected ECGs interpreted by both algorithms and by blinded experts.

Results

Among the 500 ECGs selected, 399 had a final classification of AF; 101 (20.2%) had ≥1 false positive automated interpretation. Accuracy of Cardiologs® (91.2%; CI: 82.4–94.4) was higher than Veritas® (80.2%; CI: 76.5–83.5) (p < 0.0001), and equal to Veritas® + physician (90.0%, CI:87.1–92.3) (p = 0.12). When Veritas® was incorrect, accuracy of Veritas® + physician was only 62% (CI 52–71); among those ECGs, Cardiologs® accuracy was 90% (CI: 82–94; p < 0.0001). The second database had 39 AF cases; sensitivity was 92% vs. 87% (p = 0.46) and specificity was 99.5% vs. 98.7% (p = 0.03) for Cardiologs® and Veritas® respectively.

Conclusion

Cardiologs® 12-lead ECG algorithm improves the interpretation of atrial fibrillation.

SPICED-ACS: Study of the potential impact of a computer-generated ECG diagnostic algorithmic certainty index in STEMI diagnosis: Towards transparent AI

BACKGROUND

Computerised electrocardiogram (ECG) interpretation diagnostic algorithms have been developed to guide clinical decisions like with ST segment elevation myocardial infarction (STEMI) where time in decision making is critical. These computer-generated diagnoses have been proven to strongly influence the final ECG diagnosis by the clinician; often called automation bias. However, the computerised diagnosis may be inaccurate and could result in a wrong or delayed treatment harm to the patient. We hypothesise that an algorithmic certainty index alongside a computer-generated diagnosis might mitigate automation bias. The impact of reporting a certainty index on the final diagnosis is not known.

PURPOSE

To ascertain whether knowledge of the computer-generated ECG algorithmic certainty index influences operator diagnostic accuracy.

METHODOLOGY

Clinicians who regularly analyse ECGs such as cardiology or acute care doctors, cardiac nurses and ambulance staff were invited to complete an online anonymous survey between March and April 2019. The survey had 36 ECGs with a clinical vignette of a typical chest pain and which were either a STEMI, normal, or borderline (but do not fit the STEMI criteria) along with an artificially created certainty index that was either high, medium, low or none. Participants were asked whether the ECG showed a STEMI and their confidence in the diagnosis. The primary outcomes were whether a computer-generated certainty index influenced interpreter's diagnostic decisions and improved their diagnostic accuracy. Secondary outcomes were influence of certainty index between different types of clinicians and influence of certainty index on user's own-diagnostic confidence.

RESULTS

A total of 91 participants undertook the survey and submitted 3262 ECG interpretations of which 75% of ECG interpretations were correct. Presence of a certainty index significantly increased the odds ratio of a correct ECG interpretation (OR 1.063, 95% CI 1.022-1.106, p = 0.004) but there was no significant difference between correct certainty index and incorrect certainty index (OR 1.028, 95% CI 0.923-1.145, p = 0.615). There was a trend for low certainty index to increase odds ratio compared to no certainty index (OR 1.153, 95% CI 0.898-1.482, p = 0.264) but a high certainty index significantly decreased the odds ratio of a correct ECG interpretation (OR 0.492, 95% CI 0.391-0.619, p < 0.001). There was no impact of presence of a certainty index (p = 0.528) or correct certainty index (p = 0.812) on interpreters' confidence in their ECG interpretation.

CONCLUSIONS

Our results show that the presence of an ECG certainty index improves the users ECG interpretation accuracy. This effect is not seen with differing levels of confidence within a certainty index, with reduced ECG interpretation success with a high certainty index compared with a trend for increased success with a low certainty index. This suggests that a certainty index improves interpretation when there is an increased element of doubt, possibly forcing the ECG user to spend more time and effort analysing the ECG. Further research is needed looking at time spent analysing differing certainty indices with alternate ECG diagnoses.

Interpretations of and management actions following electrocardiograms in symptomatic patients in primary care: a retrospective dossier study

BACKGROUND

The electrocardiogram (ECG) has become a popular tool in primary care. The clinical value of the ECG depends on the appropriateness of the indication and the interpretation skills of the general practitioner (GP).

OBJECTIVES

To describe the use of electrocardiography in primary care and to assess the performance of GPs in interpreting ECGs and making subsequent management decisions.

METHODS

Three hundred ECGs, recorded during daily practice in symptomatic patients by 14 GPs who regularly perform electrocardiography, were selected. Corresponding data of the indications, interpretations and subsequent management actions were extracted from the associated medical records. A panel consisting of an expert GP and a cardiologist reviewed the ECGs and specified their agreement with the findings and actions of the study GPs.

RESULTS

The most common indications were suspicion of a rhythm abnormality (43.7%), ischaemic heart disease (42.7%) and patient reassurance (14.3%). The study GPs interpreted 53.3% of the ECGs as showing no (new or acute) abnormality, whereas supraventricular rhythm disorders (12.3%), conduction disorders (7.7%) and repolarisation disorders (7.0%) were the most frequently reported pathological findings. Overall, the expert panel disagreed with the interpretations of the study GPs in 16.2% of cases, and with the GPs' management actions in 11.7%. Learning goals for GPs performing electrocardiography could be formulated for acute coronary syndrome, rhythm disorders, pulmonary embolism, reassurance, left ventricular hypertrophy and premature ventricular complexes.

CONCLUSION

GPs who feel competent in electrocardiography performed well in the opinion of the expert panel. We formulated various learning objectives for GPs performing electrocardiography.

ECG interpretation skill acquisition: A review of learning, teaching and assessment

The recording of 12 lead electrocardiograms (ECG) is one of the most useful and commonly performed medical procedures. ECGs are used in diagnosis, risk-stratification management decision-making, and assessment in response to therapy. The correct interpretation of 12 lead ECG recordings is complex and clinically challenging with misinterpretation having the potential to result in poor outcomes or even patient fatality. Despite its widespread use, several studies have highlighted deficiencies in ECG interpretation skills among health professionals. The literature suggests that up to 33% of ECG interpretations have some error when compared to the expert reference and up to 11% resulted in inappropriate management. The pedagogy of ECG interpretation lacks universal establishment; time allocation, faculty training and teaching format vary considerably within the literature. This review of the literature reports how a lack of established ECG reporting methods may contribute to the variation in reported ECG interpretation competence across many healthcare professionals. The ubiquity of the ECG in clinical practice and an over reliance on computer assisted ECG interpretation are additionally explored as factors affecting acquisition and retention of this clinical skill.

The Computerized ECG: Friend and Foe

Computerized interpretation of the electrocardiogram (ECG) began in the 1950s when conversion of its analog signal to digital form became available. Since then, automatic computer interpretations of the ECG have become routine, even at the point of care, by the addition of interpretive algorithms to portable ECG carts. Now, more than 100 million computerized ECG interpretations are recorded yearly in the United States. These interpretations have contributed to medical care by reducing physician reading time and accurately interpreting most normal ECGs. But errors do occur. The computer cannot be held responsible for misinterpretations due to recording errors, such as muscle artifacts or lead reversal. But, in many abnormal ECGs, the computer makes its own errors-sometimes critical-in its incorrect detection of arrhythmias, pacemakers, and myocardial infarctions. These errors require that all computerized statements be over-read by trained physicians who have the advantage of clinical context, unavailable to the computer. Together, the computer and over-readers now provide the most accurate ECG interpretations available.

A deep neural network learning algorithm outperforms a conventional algorithm for emergency department electrocardiogram interpretation

BACKGROUND

Cardiologs® has developed the first electrocardiogram (ECG) algorithm that uses a deep neural network (DNN) for full 12‑lead ECG analysis, including rhythm, QRS and ST-T-U waves. We compared the accuracy of the first version of Cardiologs® DNN algorithm to the Mortara/Veritas® conventional algorithm in emergency department (ED) ECGs.

METHODS

Individual ECG diagnoses were prospectively mapped to one of 16 pre-specified groups of ECG diagnoses, which were further classified as "major" ECG abnormality or not. Automated interpretations were compared to blinded experts'. The primary outcome was the performance of the algorithms in finding at least one "major" abnormality. The secondary outcome was the proportion of all ECGs for which all groups were identified, with no false negative or false positive groups ("accurate ECG interpretation"). Additionally, we measured sensitivity and positive predictive value (PPV) for any abnormal group.

RESULTS

Cardiologs® vs. Veritas® accuracy for finding a major abnormality was 92.2% vs. 87.2% (p < 0.0001), with comparable sensitivity (88.7% vs. 92.0%, p = 0.086), improved specificity (94.0% vs. 84.7%, p < 0.0001) and improved positive predictive value (PPV 88.2% vs. 75.4%, p < 0.0001). Cardiologs® had accurate ECG interpretation for 72.0% (95% CI: 69.6-74.2) of ECGs vs. 59.8% (57.3-62.3) for Veritas® (P < 0.0001). Sensitivity for any abnormal group for Cardiologs® and Veritas®, respectively, was 69.6% (95CI 66.7-72.3) vs. 68.3% (95CI 65.3-71.1) (NS). Positive Predictive Value was 74.0% (71.1-76.7) for Cardiologs® vs. 56.5% (53.7-59.3) for Veritas® (P < 0.0001).

CONCLUSION

Cardiologs' DNN was more accurate and specific in identifying ECGs with at least one major abnormal group. It had a significantly higher rate of accurate ECG interpretation, with similar sensitivity and higher PPV.

Relative contribution of modifiable risk factors for incident atrial fibrillation in Hispanics, African Americans and non-Hispanic Whites

BACKGROUND

Contribution of modifiable risk factors for the risk of new onset atrial fibrillation (AF) in minority populations is poorly understood. Our objective was to compare the population attributable risk (PAR) of various risk factors for incident AF between Hispanic, African American and non-Hispanic Whites.

METHODS

An ECG/EMR database was interrogated for individuals free of AF for development of subsequent AF from 2000 to 2013. Cox regression analysis controlled for age > 65, male gender, body mass index > 40 kg/m², systolic blood pressure > 140 mm Hg, diabetes mellitus, heart failure, socioeconomic status less than the first percentile in New York State, and race/ethnicity. PAR was calculated as (prevalence of X) ∗ (HR - 1)/HR, where HR is the hazard ratio, and X is the risk factor.

RESULTS

47,722 persons free of AF (43% Hispanic, 37% Black and 20% White) were followed for subsequent incident AF. Hypertension in African Americans and Hispanics had a 7.93% and 7.66% greater PAR compared with non-Hispanics Whites. Similar findings existed for the presence of heart failure, with a higher PAR in non-Whites compared to Whites.

CONCLUSION

In conclusion, modifiable risk factors play an important role in the risk of incident AF. Higher PAR estimates in African Americans and Hispanics were observed for elevated systolic blood pressure and heart failure. Identification of these modifiable risk factors for atrial fibrillation in non-White minorities may assist in targeting better prevention therapies and planning from a public health perspective. No funding sources were used for this study.

Accurate detection of atrial fibrillation and atrial flutter using the electrocardiomatrix technique

BACKGROUND

Atrial fibrillation (AFIB) and atrial flutter (AFL) are two common cardiac arrhythmias that predispose patients to serious medical conditions. There is a need to accurately detect these arrhythmias to prevent diseases and reduce mortality. Apart from accurately detecting these arrhythmias, it is also important to distinguish between AFIB and AFL due to differing clinical treatments.

METHODS

In this study, we applied a new technology, the electrocardiomatrix (ECM) invented in our lab, in detecting AFIB and AFL in human patients. ECM converts 2D ECG signals into a 3D color matrix, which renders arrhythmia detection intuitive, fast, and accurate. Using ECM, we analyzed the ECG signals from the online MIT-BIH Atrial Fibrillation Database (PhysioNet), and compared our ECM-based results to manual annotations based on ECG by physicians.

RESULTS

Results demonstrate that ECM and PhysioNet annotations of AFIB and AFL agree more than 99% of the time. The sensitivities of the ECM for AFIB and AFL detection were 99.2% and 98.0%, respectively, and the specificities of the ECM for AFIB and AFL were both at 99.8% and 99.8%.

CONCLUSIONS

This study demonstrates that ECM is a reliable method for accurate identification of AFIB and AFL.

Screening for Atrial Fibrillation With Electrocardiography: Evidence Report and Systematic Review for the US Preventive Services Task Force

IMPORTANCE

Atrial fibrillation is the most common arrhythmia and increases the risk of stroke.

OBJECTIVE

To review the evidence on screening for nonvalvular atrial fibrillation with electrocardiography (ECG) and stroke prevention treatment in asymptomatic adults 65 years or older to inform the US Preventive Services Task Force.

DATA SOURCES

MEDLINE, Cochrane Library, and trial registries through May 2017; references; experts; literature surveillance through June 6, 2018.

STUDY SELECTION

English-language randomized clinical trials (RCTs), prospective cohort studies evaluating detection rates of atrial fibrillation or harms of screening, and systematic reviews evaluating stroke prevention treatment. Eligible treatment studies compared warfarin, aspirin, or novel oral anticoagulants (NOACs) with placebo or no treatment. Studies were excluded that focused on persons with a history of cardiovascular disease.

DATA EXTRACTION AND SYNTHESIS

Dual review of abstracts, full-text articles, and study quality. When at least 3 similar studies were available, random-effects meta-analyses were conducted.

MAIN OUTCOMES AND MEASURES

Detection of previously undiagnosed atrial fibrillation, mortality, stroke, stroke-related morbidity, and harms.

RESULTS

Seventeen studies were included (n = 135 300). No studies evaluated screening compared with no screening and focused on health outcomes. Systematic screening with ECG identified more new cases of atrial fibrillation than no screening (absolute increase, from 0.6% [95% CI, 0.1%-0.9%] to 2.8% [95% CI, 0.9%-4.7%] over 12 months; 2 RCTs, n = 15 803), but a systematic approach using ECG did not detect more cases than an approach using pulse palpation (2 RCTs, n = 17 803). For potential harms, no eligible studies compared screening with no screening. Warfarin (mean, 1.5 years) was associated with a reduced risk of ischemic stroke (relative risk [RR], 0.32 [95% CI, 0.20-0.51]) and all-cause mortality (RR, 0.68 [95% CI, 0.50-0.93]) and with increased risk of bleeding (5 trials, n = 2415). Participants in treatment trials were not screen detected, and most had long-standing persistent atrial fibrillation. A network meta-analysis reported that NOACs were associated with a significantly lower risk of a composite outcome of stroke and systemic embolism (adjusted odds ratios compared with placebo or control ranged from 0.32-0.44); the risk of bleeding was increased (adjusted odds ratios, 1.4-2.2), but confidence intervals were wide and differences between groups were not statistically significant.

CONCLUSIONS AND RELEVANCE

Although screening with ECG can detect previously unknown cases of atrial fibrillation, it has not been shown to detect more cases than screening focused on pulse palpation. Treatments for atrial fibrillation reduce the risk of stroke and all-cause mortality and increase the risk of bleeding, but trials have not assessed whether treatment of screen-detected asymptomatic older adults results in better health outcomes than treatment after detection by usual care or after symptoms develop.

Relation of Body Mass Index to Development of Atrial Fibrillation in Hispanics, Blacks, and Non-Hispanic Whites

No previous studies have examined the interaction between body mass index (BMI) and race/ethnicity with the risk of atrial fibrillation (AF). We retrospectively followed 48,323 persons free of AF (43% Hispanic, 37% black, and 20% white; median age 60 years) for subsequent incident AF (ascertained from electrocardiograms). BMI categories included very severely underweight (BMI <15 kg/m²), severely underweight (BMI 15.1 to 15.9 kg/m²), underweight (BMI 16 to 18.4 kg/m²), normal (BMI 18.5 to 24.9 kg/m²), overweight (BMI 25.0 to 29.9 kg/m²), moderately obese (BMI 30 to 34.9 kg/m²), severely obese (BMI 35 to 39.9 kg/m²), and very severely obese (BMI >40 kg/m²). Cox regression analysis controlled for baseline covariates: heart failure, gender, age, treatment for hypertension, diabetes, PR length, systolic blood pressure, left ventricular hypertrophy, socioeconomic status, use of β blockers, calcium channel blockers, and digoxin. Over a follow-up of 13 years, 4,744 AF cases occurred. BMI in units of 10 was associated with the development of AF (adjusted hazard ratio 1.088, 95% confidence interval 1.048 to 1.130, p <0.01). When stratified by race/ethnicity, non-Hispanic whites compared with blacks and Hispanics had a higher risk of developing AF, noted in those whom BMI classes were overweight to severely obese. In conclusion, our study demonstrates that there exists a relation between obesity and race/ethnicity for the development of AF. Non-Hispanic whites had a higher risk of developing AF compared with blacks and Hispanics.

Is computer-assisted instruction more effective than other educational methods in achieving ECG competence among medical students and residents? Protocol for a systematic review and meta-analysis

INTRODUCTION

Although ECG interpretation is an essential skill in clinical medicine, medical students and residents often lack ECG competence. Novel teaching methods are increasingly being implemented and investigated to improve ECG training. Computer-assisted instruction is one such method under investigation; however, its efficacy in achieving better ECG competence among medical students and residents remains uncertain.

METHODS AND ANALYSIS

This article describes the protocol for a systematic review and meta-analysis that will compare the effectiveness of computer-assisted instruction with other teaching methods used for the ECG training of medical students and residents. Only studies with a comparative research design will be considered. Articles will be searched for in electronic databases (PubMed, Scopus, Web of Science, Academic Search Premier, CINAHL, PsycINFO, Education Resources Information Center, Africa-Wide Information and Teacher Reference Center). In addition, we will review citation indexes and conduct a grey literature search. Data extraction will be done on articles that met the predefined eligibility criteria. A descriptive analysis of the different teaching modalities will be provided and their educational impact will be assessed in terms of effect size and the modified version of Kirkpatrick framework for the evaluation of educational interventions. This systematic review aims to provide evidence as to whether computer-assisted instruction is an effective teaching modality for ECG training. It is hoped that the information garnered from this systematic review will assist in future curricular development and improve ECG training.

ETHICS AND DISSEMINATION

As this research is a systematic review of published literature, ethical approval is not required. The results will be reported according to the Preferred Reporting Items for Systematic Review and Meta-Analysis statement and will be submitted to a peer-reviewed journal. The protocol and systematic review will be included in a PhD dissertation.

PROSPERO REGISTRATION NUMBER

CRD42017067054; Pre-results.