Citations

Abstract Background New-onset atrial fibrillation (AF) is common in patients who are hospitalized with acute illness and is associated with overall poor outcomes. Current clinical and demographic predictors of new-onset AF have limitations. Electrocardiogram (ECG) waveforms, combined with these factors as part of multimodal approach could help improve prediction of new-onset AF. Purpose The purpose of our study was to develop a multimodal, deep-learning system that combines ECG waveforms with clinical data for prediction of new-onset AF in hospitalized patients. Methods We developed and compared two separate 1-dimensional models with a fusion model for prediction of new-onset AF. (a): one-dimensional convolutional neural network (1D-CNN) that learns patterns from the raw ECG waveforms. (b): a Multi-Layer Perceptron (MLP), unsupervised machine learning model which utilizes data from structured variables such as: age, heart rate, routinely obtained laboratory data and medications. The two modules each produce latent feature vectors, which are then placed end-to-end and fed into a final Multi- MLP head that outputs the probability of New-onset AF during hospitalization. We then compared this fused MLP model with the two simpler unimodal models based solely on ECG waveforms and clinical data respectively. Results On the held-out test set data, the feature-level fused model achieved an Area Under the Curve (AUC) of 0.81. This significantly outperformed both the ECG-alone model (0.77) and the clinical-alone model (0.74). By adjusting the threshold, we can fine-tune the tool at a threshold of 0.5, it flags 83% of patients who later develop AF (207 of 248) with a precision of 10%, a setting that could be useful for wider screening. By adjusting the threshold of 0.98 for a high-precision alert configuration, it has the ability to predict 46% of future occurrence of AF cases (113 of 248) while reducing false alarms by 81% (from 1,797 to 328), increasing precision to 26% and achieving an overall accuracy of 89%, a setting useful when resources for follow-up are limited. Conclusions Combining ECG signals with clinical data significantly enhances the model’s ability to detect new-onset AF in hospitalized patients as compared to a uni-dimensional CNN model. The dynamic nature of ML model also allows for adjustment of its decision threshold which can aid in broader screening and improving adverse outcomes related to occurrence of AF.ROC Curve for Fusion model

Background The purpose of this study was to assess the clinical benefits of guideline‐directed medical therapy (GDMT) for heart failure (HF) in adults with congenital heart disease and HF with reduced ejection fraction. Methods HF with reduced ejection fraction was defined as stage B/C HF and systemic left ventricular ejection fraction <50%. GDMT score was assessed at baseline encounter (baseline GDMT score) and 1‐year follow‐up. GDMT uptitration was calculated as defined as ∆GDMT score from baseline. Cox regression was used to assess the relationship between HF therapy (baseline and ∆GDMT score) and outcomes (HF hospitalization and mortality). Results Of 778 patients (age 44±16, 480 [62%] males), baseline GDMT score was 2 (1, 3) and ∆GDMT score was 0.62 (95% CI, 0.55–0.69). Of 778, only 258 (33%) had GDMT uptitration (∆GDMT score >0). Higher use of GDMT at baseline (adjusted HR per 1 point increase in GDMT score 0.81, 95% CI, 0.74–0.88, P <0.001), and increased use of GDMT over time (adjusted HR for 1 point increase in ∆GDMT score HR 0.70, 95% CI, 0.61–0.79, P <0.001) were associated with lower risk of HF hospitalization. Higher baseline GDMT score (adjusted HR 0.85, 95% CI, 0.76–0.94, P <0.001), and ∆ GDMT score (adjusted HR 0.82, 95% CI, 0.71–0.94, P <0.000) were also associated with lower mortality. Conclusions GDMT use at baseline and GDMT uptitration were associated with improved outcomes and suggest a dose‐dependent relationship between use of GDMT and risk of adverse outcomes. Furthermore, 67% of patients did not receive GDMT uptitration, suggesting suboptimal therapy, and opportunities for improvement.

ABSTRACT Background Ventricular tachycardia (VT) is a life‐threatening arrhythmia that requires both relatively rapid and accurate detection in intensive care units (ICUs). Continuous monitoring systems play a crucial role in detecting them. However, previous studies have reported that nearly 9 out of 10 arrhythmia alarms in ICUs tend to be false positives, which usually transpire to a well‐documented phenomenon called “alarm fatigue” that leads to desensitization, delayed responses, and increased cognitive burden on healthcare providers. Methods We developed a deep learning based, one‐dimensional convolutional neural network (1D‐CNN) to classify VT alarms using multiple raw waveform inputs, including two electrocardiogram (ECG) leads, photoplethysmogram (PPG) and arterial blood pressure (ABP) signals. The model was trained using the publicly available VTaC Arrhythmia Benchmark Dataset. We used the 10‐second waveform segments that preceded each VT alarm, pre‐processed and then used them to train the machine learning model to correctly classify the VT alarm. Results On the test set, the model achieved an area under the receiver operating characteristic curve of 0.901, overall accuracy of 83.22%, F1‐score of 73.3%, sensitivity of 77.53%, specificity of 85.63%, and positive predictive value of 69.57%. The model successfully detected over three‐quarters of them while significantly reducing false positive rates for the detection of VT. Conclusions This study demonstrates that a deep learning based 1D‐CNN model using short segments of raw waveform data can achieve robust performance in distinguishing true and false VT alarms.

Abstract Aims Echo-derived left atrial (LA) stiffness index improves prognostication in patients with heart failure preserved ejection fraction but has not been studied in adults with repaired coarctation of aorta (COA). We hypothesized that, among adults with COA, those with high LA stiffness index would have worse disease severity indices [peak VO2, NT-proBNP, and right ventricular to pulmonary artery (RV-PA) coupling], heart failure hospitalization, and all-cause mortality. Methods and results LA stiffness index was estimated using the baseline echocardiogram as the quotient of septal E/e′ and LA reservoir strain (E/e′/LARS). Using the median septal LA stiffness index as the cut-off, patients were divided into high LA stiffness index (LA stiffness index &gt;0.28) vs. low LA stiffness index groups (LA suggests index ≤0.28). The study comprised of 729 patients [age 36 ± 16 years; males 437 (60%)]. LA stiffness index correlated with RV-PA coupling (r = −0.56, P &lt; 0.001), NT-proBNP (r = 0.62, P &lt; 0.001), and predicted peak VO2 (r = −0.51, P &lt; 0.001). High LA stiffness index was associated with heart failure hospitalization (adjusted HR 3.17, 95% confidence interval (CI) 1.84–7.36, P = 0.007) and all-cause mortality (adjusted HR 2.81, 95% CI 1.28–12.32, P = 0.005). LA stiffness index had superior prognostic performance for predicting 5-year risk of heart failure hospitalization (AUC 0.803, 95% CI 0.756–0.849, P &lt; 0.001) and all-cause mortality (AUC 0.786, 95% CI 0.742–08.30, P &lt; 0.001) compared with conventional echocardiographic indices of left heart diastolic function. Conclusion LA stiffness index improves prognostication in adults with COA and can potentially be used to guide timing of intervention or monitor response to therapy.

Ventricular tachycardia (VT) is a life-threatening arrhythmia that requires both relatively rapid and accurate detection in intensive care units (ICUs). Continuous monitoring systems play a crucial role in detecting them. However, previous studies have reported that nearly 9 out of 10 arrhythmia alarms in ICUs tend to be false positives, which usually transpire to a well-documented phenomenon called “alarm fatigue” that leads to desensitization, delayed responses and increased cognitive burden on healthcare providers. We developed a deep learning based, one-dimensional convolutional neural network (1D-CNN) to classify VT alarms using multiple raw waveform inputs, including two electrocardiogram (ECG) leads, photoplethysmogram (PPG) and arterial blood pressure (ABP) signals. The model was trained using the publicly available VTaC Arrhythmia Benchmark Dataset. Ten-second waveform segments preceding each alarm were pre-processed and used to train the model to correctly classify the VT alarm. On the test set, the model achieved an area under the receiver operating characteristic curve (AUC-ROC) of 0.901, overall accuracy of 83.22%, F1-score of 73.3%, sensitivity of 77.53%, specificity of 85.63%, and positive predictive value (PPV) of 69.57%. The model successfully detected over three-quarters of them while significantly reducing false positive rates for the detection of VT. This study demonstrates that a deep learning based 1D-CNN model using short segments of raw waveform data can achieve robust performance in distinguishing true and false VT alarms.

Background Patients with cardiohepatic syndrome are at higher risk of adverse outcomes in the cardiac intensive care unit. We hypothesized that cardiohepatic syndrome phenotypes would exhibit differences in their clinical and transthoracic echocardiogram hemodynamic profiles, portending a higher risk of mortality. Methods We included unique CICU patients with an admission diagnosis of heart failure from 2007 to 2018 with available data for 1 or more admission liver function tests. Echocardiographic variables were extracted from patients who had a transthoracic echocardiogram within 1 day of admission. We assigned patients to 1 of 4 mutually exclusive cardiohepatic syndrome phenotype groups: normal, hepatocellular, cholestasis, and combined. Results We included 3992 CICU patients; 2483 had an available transthoracic echocardiogram. Patients with cholestasis had more right ventricular systolic dysfunction and congestion, whereas those with hepatocellular injury had more left ventricular systolic dysfunction and poor forward flow; hemodynamics and biventricular function were worst in the combined group. In‐hospital mortality occurred in 14.6% and was higher for patients with cholestasis or hepatocellular injury, especially those with greater extent/severity and worse biventricular function or poor hemodynamics. After adjustment, only the group with hepatocellular injury (adjusted odds ratio [OR], 1.35 [95% CI, 1.04–1.75], P =0.03) and the combined group (adjusted OR, 1.75 [95% CI, 1.30–2.34], P <0.001) remained at higher risk. One‐year mortality was highest in the combined group (adjusted hazard ratio [HR], 1.29 [95% CI, 1.10–1.51], P =0.002). Conclusions Cardiohepatic syndrome phenotypes are associated with distinct echocardiographic profiles. Patients with the combined cardiohepatic syndrome phenotype were at highest risk of mortality, particularly if this was combined with poor cardiac function.

Abstract Aims Systemic ventricular (SV) strain imaging improves early detection of systolic dysfunction in patients with congenital heart disease and biventricular circulation. However, its role for risk stratification in the Fontan population has not been systematically studied. We hypothesized that among Fontan patients with normal SV ejection fraction (Echo_EF) at baseline, those with reduced SV longitudinal strain (Echo_LS) would have a higher risk of progressive SV systolic dysfunction, greater neurohormonal activation, and cardiovascular events (heart failure hospitalization/death/transplant) during follow-up. Methods and results We studied adults with Fontan palliation and normal SV systolic function (Echo_EF &gt;50%) at baseline who underwent follow-up echocardiogram &gt;36 months after the baseline echocardiogram. Echo_LS was assessed from the apical 4-chamber window using speckle tracking echocardiography. Patients were divided into reduced vs. normal Echo_LS groups, using −18% as the cut-off point. Echo_LS, Echo_EF, and NT-proBNP were assessed at baseline and follow-up encounters. Of 272 patients (age, 28 ± 8 years), 98 (36%) and 174 (64%) had reduced vs. normal Echo_LS. The reduced_Echo_LS group had progressive SV systolic dysfunction [worse temporal decline in Echo_EF (relative_Δ_Echo_EF −12.8% vs. −2.1%, P &lt; 0.001)], and higher prevalence of new-onset SV systolic dysfunction [30% (29/98) vs. 8% (13/174), P &lt; 0.001]. The reduced Echo_LS group also had greater neurohormonal activation [relative Δ_ NT-proBNP 36 (95%CI, 27–43) % vs. 17 (95%CI, 11–23) %, P &lt; 0.001], and higher risk of cardiovascular events (adjusted hazard ratio 1.84, P = 0.009). Conclusion Echo_LS improves risk stratification in Fontan patients with normal Echo_EF and may provide an opportunity for early interventions to improve outcomes.