Search

Computerized electrocardiography (ECG) has been widely used and allows linkage to electronic medical records. The present study describes the development and clinical applications of an electronic cohort derived from a digital ECG database obtained by the Telehealth Network of Minas Gerais, Brazil, for the period 2010–2017, linked to the mortality data from the national information system, the Clinical Outcomes in Digital Electrocardiography (CODE) dataset. From 2,470,424 ECGs, 1,773,689 patients were identified. A total of 1,666,778 (94%) underwent a valid ECG recording for the period 2010 to 2017, with 1,558,421 patients over 16 years old, 40.2% were men, with a mean age of 51.7 [SD 17.6] years. During a mean follow-up of 3.7 years, the mortality rate was 3.3%. ECG abnormalities assessed were: atrial fibrillation (AF), right bundle branch block (RBBB), left bundle branch block (LBBB), atrioventricular block (AVB), and ventricular pre-excitation. Most ECG abnormalities (AF: Hazard ratio [HR] 2.10, 95% CI 2.03–2.17, RBBB: HR 1.32, 95%CI 1.27–1.36, LBBB: HR 1.69, 95% CI 1.62–1.76, first degree AVB: Relative survival [RS]: 0.76, 95% CI0.71–0.81, 2:1 AVB: RS 0.21 95% CI0.09–0.52, and RS 0.36, third degree AVB: 95% CI 0.26–0.49) were predictors of overall mortality, except for ventricular pre-excitation (HR 1.41, 95% CI 0.56–3.57) and Mobitz I AVB (RS 0.65, 95% CI 0.34–1.24). In conclusion, a large ECG database established by a telehealth network can be a useful tool for facilitating new advances in the fields of digital electrocardiography, clinical cardiology and cardiovascular epidemiology.

Resumo Fundamento O bloqueio atrioventricular (BAV) descreve um comprometimento na condução dos átrios para os ventrículos. Embora o curso clínico do BAV tenha sido avaliado, os achados são de países de alta renda e, portanto, não podem ser extrapolados para a população latina. Objetivo Avaliar a associação entre BAV e mortalidade. Métodos Foram incluídos pacientes do estudo CODE (Clinical Outcomes in Digital Electrocardiology), maiores de 16 anos que realizaram eletrocardiograma (ECG) digital de 2010 a 2017. Os ECGs foram relatados por cardiologistas e por software automatizado. Para avaliar a relação entre BAV e mortalidade, foram utilizados o modelo log-normal e as curvas de Kaplan-Meier com valores de p bicaudais < 0,05 considerados estatisticamente significativos. Resultados O estudo incluiu 1.557.901 pacientes; 40,23% eram homens e a média de idade foi de 51,7 (DP ± 17,6) anos. Durante um seguimento médio de 3,7 anos, a mortalidade foi de 3,35%. A prevalência de BAV foi de 1,38% (21.538). Os pacientes com BAV de primeiro, segundo e terceiro graus foram associados a uma taxa de sobrevida 24% (taxa de sobrevida relativa [RS] = 0,76; intervalo de confiança [IC] de 95%: 0,71 a 0,81; p < 0,001), 55% (RS = 0,45; IC de 95%: 0,27 a 0,77; p = 0,01) e 64% (RS = 0,36; IC de 95%: 0,26 a 0,49; p < 0,001) menor quando comparados ao grupo controle, respectivamente. Os pacientes com BAV 2:1 tiveram 79% (RS = 0,21; IC de 95%: 0,08 a 0,52; p = 0,005) menor taxa de sobrevida do que o grupo controle. Apenas Mobitz tipo I não foi associado a maior mortalidade (p = 0,27). Conclusão BAV foi um fator de risco independente para mortalidade geral, com exceção do BAV Mobitz tipo I. Abstract Background Atrioventricular block (AVB) describes an impairment of conduction from the atria to the ventricles. Although the clinical course of AVB has been evaluated, the findings are from high-income countries and, therefore, cannot be extrapolated to the Latinx population. Objective Evaluate the association between AVB and mortality. Methods Patients from the CODE (Clinical Outcomes in Digital Electrocardiology) study, older than 16 years who underwent digital electrocardiogram (ECG) from 2010 to 2017 were included. ECGs were reported by cardiologists and by automated software. To assess the relationship between AVB and mortality, the log-normal model and the Kaplan-Meier curves were used with two-tailed p-values < 0.05 considered statistically significant. Results The study included 1,557,901 patients; 40.2% were men, and mean age was 51.7 (standard deviation ± 17.6) years. In a mean follow-up of 3.7 years, the mortality rate was 3.35%. The AVB prevalence was 1.38% (21,538). Patients with first-, second-, and third-degree AVB were associated with 24% (relative survival rate [RS] = 0.76; 95% confidence interval [CI]: 0.71-0.81; p < 0.001), 55% (RS = 0.45; 95% CI: 0.27-0.77; p = 0.01), and 64% (RS = 0.36; 95% CI: 0.26-0.49; p < 0.001) lower survival rate when compared to the control group, respectively. Patients with 2:1 AVB had 79% (RS = 0.21; 95% CI: 0.08-0.52; p = 0.005) lower survival rate than the control group. Only Mobitz type I was not associated with higher mortality (p = 0.27). Conclusion AVB was an independent risk factor for overall mortality, with the exception of Mobitz type I.

Atrioventricular block (AVB) describes an impairment of conduction from the atria to the ventricles. Although the clinical course of AVB has been evaluated, the findings are from high-income countries and, therefore, cannot be extrapolated to the Latinx population.Evaluate the association between AVB and mortality.Patients from the CODE (Clinical Outcomes in Digital Electrocardiology) study, older than 16 years who underwent digital electrocardiogram (ECG) from 2010 to 2017 were included. ECGs were reported by cardiologists and by automated software. To assess the relationship between AVB and mortality, the log-normal model and the Kaplan-Meier curves were used with two-tailed p-valueslt; 0.05 considered statistically significant.The study included 1,557,901 patients; 40.2% were men, and mean age was 51.7 (standard deviation ± 17.6) years. In a mean follow-up of 3.7 years, the mortality rate was 3.35%. The AVB prevalence was 1.38% (21,538). Patients with first-, second-, and third-degree AVB were associated with 24% (relative survival rate [RS] = 0.76; 95% confidence interval [CI]: 0.71-0.81; plt; 0.001), 55% (RS = 0.45; 95% CI: 0.27-0.77; p = 0.01), and 64% (RS = 0.36; 95% CI: 0.26-0.49; plt; 0.001) lower survival rate when compared to the control group, respectively. Patients with 2:1 AVB had 79% (RS = 0.21; 95% CI: 0.08-0.52; p = 0.005) lower survival rate than the control group. Only Mobitz type I was not associated with higher mortality (p = 0.27).AVB was an independent risk factor for overall mortality, with the exception of Mobitz type I.O bloqueio atrioventricular (BAV) descreve um comprometimento na condução dos átrios para os ventrículos. Embora o curso clínico do BAV tenha sido avaliado, os achados são de países de alta renda e, portanto, não podem ser extrapolados para a população latina.Avaliar a associação entre BAV e mortalidade.Foram incluídos pacientes do estudo CODE (Clinical Outcomes in Digital Electrocardiology), maiores de 16 anos que realizaram eletrocardiograma (ECG) digital de 2010 a 2017. Os ECGs foram relatados por cardiologistas e por software automatizado. Para avaliar a relação entre BAV e mortalidade, foram utilizados o modelo log-normal e as curvas de Kaplan-Meier com valores de p bicaudaislt; 0,05 considerados estatisticamente significativos.O estudo incluiu 1.557.901 pacientes; 40,23% eram homens e a média de idade foi de 51,7 (DP ± 17,6) anos. Durante um seguimento médio de 3,7 anos, a mortalidade foi de 3,35%. A prevalência de BAV foi de 1,38% (21.538). Os pacientes com BAV de primeiro, segundo e terceiro graus foram associados a uma taxa de sobrevida 24% (taxa de sobrevida relativa [RS] = 0,76; intervalo de confiança [IC] de 95%: 0,71 a 0,81; plt; 0,001), 55% (RS = 0,45; IC de 95%: 0,27 a 0,77; p = 0,01) e 64% (RS = 0,36; IC de 95%: 0,26 a 0,49; plt; 0,001) menor quando comparados ao grupo controle, respectivamente. Os pacientes com BAV 2:1 tiveram 79% (RS = 0,21; IC de 95%: 0,08 a 0,52; p = 0,005) menor taxa de sobrevida do que o grupo controle. Apenas Mobitz tipo I não foi associado a maior mortalidade (p = 0,27).BAV foi um fator de risco independente para mortalidade geral, com exceção do BAV Mobitz tipo I.

Location-aware information is now commonplace, as the ubiquity and pervasiveness of technology enabled its generation and storage at large scale. These data constitute a rich representation of entities’ whereabouts and behavior as they move on the map. Although several studies reported considerable predictability of such mobility patterns, several factors may impose significant changes on moving behavior. Being able to detect these changes can benefit several applications. In this article, we formalize and address the problem of detecting mobility drifts in mobility patterns. This problem is particularly challenging due to the noisy and incomplete nature of the data. We design non-parametric tests and present two algorithms to detect mobility drifts when the putative drift point is known in advance and there is no previous knowledge about the existence of potential changes, and we need to search for the most likely drift point rigorously. To evaluate our algorithms, we perform an extensive experimental study with real-world data coming from a variety of scenarios, such as geo-tagged social media data and GPS traces of connected vehicles. The results show the effectiveness of our algorithms, being able to identify existing drift points on spatial mobility patterns correctly.

Left bundle branch block is recognized as a marker of higher risk of death, but the prognostic value of the right bundle branch block in the general population is still controversial. Our aim is to evaluate the risk of overall and cardiovascular mortality in patients with right (RBBB) and left bundle branch block (LBBB) in a large electronic cohort of Brazilian patients.This observational retrospective study was developed with the database of digital ECGs from Telehealth Network of Minas Gerais, Brazil (TNMG). All ECGs performed from 2010 to 2017 in primary care patients over 16 years old were assessed. The electronic cohort was obtained by linking data from ECG exams (name, sex, date of birth, city of residence) and those from national mortality information system, using standard probabilistic linkage methods (FRIL: Fine-grained record linkage software, v.2.1.5, Atlanta, GA). Only the first ECG of each patient was considered. Clinical data were self-reported, and ECGs were interpreted manually by cardiologists and automatically by the Glasgow University Interpreter software. Hazard ratio (HR) for mortality was estimated using Cox regression.From a dataset of 1,773,689 patients, 1,558,421 primary care patients over 16 years old underwent a valid ECG recording during 2010 to 2017. We excluded 17,359 patients that didn't have a valid QRS measure from the Glasgow program and 11,091 patients from the control group that had QRS equal or above 120 ms and were not RBBB or LBBB. Therefore, 1,529,971 were included (median age 52 [Q1:38; Q3:65] years; 40.2% were male). In a mean follow-up of 3.7 years, the overall mortality rate was 3.34%. RBBB was more frequent (2.42%) than LBBB (1.32%). In multivariate analysis, adjusting for sex, age and comorbidities, both patients with RBBB (HR 1.32; CI 95% 1.27-1.37) and LBBB (HR 1.69; CI 95% 1.62-1.76) had higher risk of overall mortality. Women with RBBB had an increased risk of all-cause death compared to men (p 0.001). Cardiovascular mortality was higher in patients with LBBB (HR 1.77; CI 95% 1.55-2.01), but not for RBBB.Patients with RBBB and LBBB had higher risk of overall mortality. Women with RBBB had more risk of all-cause death than men. LBBB was associated with higher risk of cardiovascular mortality.

Digital electrocardiographs are now widely available and a large number of digital electrocardiograms (ECGs) have been recorded and stored. The present study describes the development and clinical applications of a large database of such digital ECGs, namely the CODE (Clinical Outcomes in Digital Electrocardiology) study. ECGs obtained by the Telehealth Network of Minas Gerais, Brazil, from 2010 to 17, were organized in a structured database. A hierarchical free-text machine learning algorithm recognized specific ECG diagnoses from cardiologist reports. The Glasgow ECG Analysis Program provided Minnesota Codes and automatic diagnostic statements. The presence of a specific ECG abnormality was considered when both automatic and medical diagnosis were concordant; cases of discordance were decided using heuristisc rules and manual review. The ECG database was linked to the national mortality information system using probabilistic linkage methods. From 2,470,424 ECGs, 1,773,689 patients were identified. After excluding the ECGs with technical problems and patients16 years-old, 1,558,415 patients were studied. High performance measures were obtained using an end-to-end deep neural network trained to detect 6 types of ECG abnormalities, with F1 scores80% and specificity99% in an independent test dataset. We also evaluated the risk of mortality associated with the presence of atrial fibrillation (AF), which showed that AF was a strong predictor of cardiovascular mortality and mortality for all causes, with increased risk in women. In conclusion, a large database that comprises all ECGs performed by a large telehealth network can be useful for further developments in the field of digital electrocardiography, clinical cardiology and cardiovascular epidemiology.

The electrocardiogram (ECG) is the most commonly used exam for the screening and evaluation of cardiovascular diseases. Here we propose that the age predicted by artificial intelligence (AI) from the raw ECG tracing (ECG-age) can be a measure of cardiovascular health and provide prognostic information. A deep convolutional neural network was trained to predict a patient’s age from the 12-lead ECG using data from patients that underwent an ECG from 2010 to 2017 - the CODE study cohort (n=1,558,415 patients). On the 15% hold-out CODE test split, patients with ECG-age more than 8 years greater than chronological age had a higher mortality rate (hazard ratio (HR) 1.79, p

Background Ventricular pre-excitation is characterized by the presence of atrioventricular accessory pathways, predisposing to arrhythmias. Although it is well established that risk stratification in symptomatic patients should be invasive, there is a lack of evidence of the benefit in asymptomatic. Objective Evaluate ventricular pre-excitation in the electrocardiogram (ECG) as a risk factor for overall mortality in patients of Telehealth Network of Minas Gerais (TNMG), Brazil. Methods This observational study was developed with the database of digital ECGs (2010-2017) from TNMG. The electronic cohort was obtained by linking data from ECG exams and those from the national mortality information system. Only the first ECG was considered. Clinical data were self-reported, and ECGs were interpreted manually by cardiologists and automatically by the Glasgow University Interpreter software. Hazard ratio (HR) for mortality was estimated using weighted Cox regression. Results Nearly 1 665 667 patients were included (median age: 50 [Q1: 34; Q3: 63] years; 41.4% were male). In a mean follow-up of 3.7 years, the overall mortality rate was 3.1%. The prevalence of ventricular pre-excitation was 0.07%. In multivariate analysis, adjusting for sex and age, ventricular pre-excitation was not associated with an increased risk of mortality (HR: 1.41; 95% confidence interval [CI]: 0.56-3.57; p = .47) when compared to the whole sample or to patients with normal ECG (HR: 1.41; 95% CI: 0.53-4.36; p = .43). In a subanalysis on accessory pathway location, there was no evidence of a higher risk of death related to any location. Conclusion Ventricular pre-excitation was not associated with an increased risk of mortality in a primary care cohort.

The electrocardiogram (ECG) is the most commonly used exam for the screening and evaluation of cardiovascular diseases. Here, the authors propose that the age predicted by artificial intelligence from the raw ECG tracing can be a measure of cardiovascular health and provide prognostic information., The electrocardiogram (ECG) is the most commonly used exam for the evaluation of cardiovascular diseases. Here we propose that the age predicted by artificial intelligence (AI) from the raw ECG (ECG-age) can be a measure of cardiovascular health. A deep neural network is trained to predict a patient's age from the 12-lead ECG in the CODE study cohort (n = 1,558,415 patients). On a 15% hold-out split, patients with ECG-age more than 8 years greater than the chronological age have a higher mortality rate (hazard ratio (HR) 1.79, p < 0.001), whereas those with ECG-age more than 8 years smaller, have a lower mortality rate (HR 0.78, p < 0.001). Similar results are obtained in the external cohorts ELSA-Brasil (n = 14,236) and SaMi-Trop (n = 1,631). Moreover, even for apparent normal ECGs, the predicted ECG-age gap from the chronological age remains a statistically significant risk predictor. These results show that the AI-enabled analysis of the ECG can add prognostic information.

We propose a new method to generate explanations for end-to-end classification models. The explanations consist of meaningful features to the user, namely contextual features. We instantiate our approach in the scenario of automated electrocardiogram (ECG) diagnosis and analyze the explanations generated in terms of interpretability and robustness. The proposed method uses a noise-insertion strategy to quantify the impact of intervals and segments of the ECG signals on the automated classification outcome. These intervals and segments and their impact on the diagnosis are common place to cardiologists, and their usage in explanations enables a better understanding of the outcomes and also the identification of sources of mistakes. The proposed method is particularly effective and useful for modern deep learning models that take raw data as input. We demonstrate our method by explaining diagnoses generated by a deep convolutional neural network.

In this work, we present a method to explain “end-to-end” electrocardiogram (ECG) signal classifiers, where the explanations were built along with seniors cardiologist to provide meaningful features to the final users. Our method focuses exclusively on automated ECG diagnosis and analyzes the explanation in terms of clinical accuracy for interpretability and robustness. The proposed method uses a noise-insertion strategy to quantify the impact of intervals and segments of the ECG signals on the automated classification outcome. An ECG segmentation method was applied to ECG tracings, to obtain: (1) Intervals, Segments and Axis; (2) Rate, and (3) Rhythm. Noise was added to the signal to disturb the ECG features in a realistic way. The method was tested using Monte Carlo simulation and the feature impact is estimated by the change in the model prediction averaged over 499 executions and a feature is defined as important if its mean value changes the result of the classifier. We demonstrate our method by explaining diagnoses generated by a deep convolutional neural network. The proposed method is particularly effective and useful for modern deep learning models that take raw data as input.

The role of automatic electrocardiogram (ECG) analysis in clinical practice is limited by the accuracy of existing models. Deep Neural Networks (DNNs) are models composed of stacked transformations that learn tasks by examples. This technology has recently achieved striking success in a variety of task and there are great expectations on how it might improve clinical practice. Here we present a DNN model trained in a dataset with more than 2 million labeled exams analyzed by the Telehealth Network of Minas Gerais and collected under the scope of the CODE (Clinical Outcomes in Digital Electrocardiology) study. The DNN outperform cardiology resident medical doctors in recognizing 6 types of abnormalities in 12-lead ECG recordings, with F1 scores above 80% and specificity over 99%. These results indicate ECG analysis based on DNNs, previously studied in a single-lead setup, generalizes well to 12-lead exams, taking the technology closer to the standard clinical practice., A preliminary version of this work titled: "Automatic Diagnosis of Short-Duration 12-Lead ECG using a Deep Convolutional Network " was presented in the Machine Learning for Health Workshop at NeurIPS 2018 and was made available under a different identifier: arXiv:1811.12194. The current version subsumes all previous versions

Aims: Atrial fibrillation (AF) is a public health problem and its prevalence is increasing worldwide. Electronic cohorts, with large electrocardiogram (ECG) databases linked to mortality data, can be useful in determining prognostic value of ECG abnormalities. Our aim is to evaluate the risk of mortality in patients with AF from Brazil. Methods: This observational retrospective study of primary care patients was developed with the digital ECG database from the Telehealth Network of Minas Gerais, Brazil. ECGs performed from 2010 to 2017 were interpreted by cardiologists and the University of Glasgow automated analysis software. An electronic cohort was obtained linking data from ECG exams and those from a national mortality information system, using standard probabilistic linkage methods. We considered only the first ECG of each patient. Patients under 16 years were excluded. Hazard ratios (HR) for mortality were adjusted for demographic and self-reported clinical factors and estimated with Cox regression. Results: From a dataset of 1,773,689 patients, 1,558,421 were included, mean age 51.6 years; 40.2% male. There were 3.34% deaths from all causes in 3.68 years of median follow up. The prevalence of AF was 1.33%. AF was an independent risk factor for all-cause mortality (HR 2.10, 95%CI 2.03–2.17) and cardiovascular mortality (HR 2.06, 95%CI 1.86–2.29). Females with AF had a higher risk of overall and cardiovascular mortality compared with males (p < 0.001). Conclusions: AF was a strong predictor of cardiovascular and all-cause mortality in a primary care population, with increased risk in women. Condensed abstract To assess risk of mortality in AF patients, an electronic cohort was obtained linking data from ECG exams of Brazilian primary care patients and a national mortality information system. From 1,558,421 patients, AF (prevalence 1.33%) carried a higher risk of overall and cardiovascular mortality, with increased risk in women. What’s New This is the first study with a large Brazilian electronic cohort to evaluate the risk of mortality linked to AF in primary care patients. AF patients from a Brazilian primary care population had a higher risk of death for all causes (HR 2.10, 95%CI 2.03–2.17) and cardiovascular mortality (HR 2.06, 95%CI 1.86–2.29). Female patients with AF had an increased risk of overall and cardiovascular mortality compared with male patients (p < 0.001).

Aging affects the electrocardiogram (ECG) with higher incidence of abnormalities in older patients. A recent study suggests that an ECG-age can be predicted by artificial intelligence (AI) and can be used as a measure of cardiovascular health. Since the prognosis impact of this measure is unknown

Traditionally, in health surveillance, high risk zones are identified based only on the residence address or the working place of diseased individuals. This provides little information about the places where people are infected, the truly important information for disease control. The recent availability of spatial data generated by geotagged social media posts offers a unique opportunity: by identifying and following diseased individuals, we obtain a collection of sequential geo-located events, each sequence being issued by a social media user. The sequence of map positions implicitly provides an estimation of the users’ social trajectories as they drift on the map. The existing data mining techniques for spatial cluster detection fail to address this new setting as they require a single location to each individual under analysis. In this paper we present two stochastic models with their associated algorithms to mine this new type of data. The Visit Model finds the most likely zones that a diseased person visits, while the Infection Model finds the most likely zones where a person gets infected while visiting. We demonstrate the applicability and effectiveness of our proposed models by applying them to more than 100 million geotagged tweets from Brazil in 2015. In particular, we target the identification of infection hot spots associated with dengue, a mosquito-transmitted disease that affects millions of people in Brazil annually, and billions worldwide. We applied our algorithms to data from 11 large cities in Brazil and found infection hot spots, showing the usefulness of our methods for disease surveillance.

Every day text-based social media channels are flooded with millions of messages that comprise the most diverse topics. These channels are being used as a rich data source for monitoring different real world events such as natural disasters and disease outbreaks, to name a few. However, depending on the event being investigated, this monitoring may be severely affected by data scarcity and noise, allowing just coarse grain analysis in terms of time and space, which lack the specificity necessary for supporting actions at the local level. In this context, we present a methodology to handle data scarcity and noise while monitoring real world events using social media data in a fine grain. We apply our methodology to dengue-related data from Brazil, and show how it could improve significantly the performance of event monitoring at a local scale almost doubling the correlation observed in some cases.