Electrocardiogram-based mortality prediction in patients with COVID-19 using machine learning.

BACKGROUND AND PURPOSE: The electrocardiogram (ECG) is frequently obtained in the work-up of COVID-19 patients. So far, no study has evaluated whether ECG-based machine learning models have added value to predict in-hospital mortality specifically in COVID-19 patients. METHODS: Using data from the CAPACITY-COVID registry, we studied 882 patients admitted with COVID-19 across seven hospitals in the Netherlands. Raw format 12-lead ECGs recorded within 72 h of admission were studied. With data from five hospitals (n = 634), three models were developed: (a) a logistic regression baseline model using age and sex, (b) a least absolute shrinkage and selection operator (LASSO) model using age, sex and human annotated ECG features, and (c) a pre-trained deep neural network (DNN) using age, sex and the raw ECG waveforms. Data from two hospitals (n = 248) was used for external validation. RESULTS: Performances for models a, b and c were comparable with an area under the receiver operating curve of 0.73 (95% confidence interval [CI] 0.65-0.79), 0.76 (95% CI 0.68-0.82) and 0.77 (95% CI 0.70-0.83) respectively. Predictors of mortality in the LASSO model were age, low QRS voltage, ST depression, premature atrial complexes, sex, increased ventricular rate, and right bundle branch block. CONCLUSION: This study shows that the ECG-based prediction models could be helpful for the initial risk stratification of patients diagnosed with COVID-19, and that several ECG abnormalities are associated with in-hospital all-cause mortality of COVID-19 patients. Moreover, this proof-of-principle study shows that the use of pre-trained DNNs for ECG analysis does not underperform compared with time-consuming manual annotation of ECG features.

Age stratified risk of cerebral venous sinus thrombosis after SARS-CoV-2 vaccination

Introduction: Cerebral Venous Sinus Thrombosis (CVST) as a part of the thrombosis and thrombocytopenia syndrome is a rare adverse drug reaction of SARS-CoV-2 vaccination. The estimated background rate of CVST in adults is around 1 case per million per month, and CVST with thrombocytopenia accounts for 8% of all CVST. We assessed the age-stratified risk of CVST with and without thrombocytopenia after SARS-CoV-2 vaccination. Methods: We estimated the absolute risk of any CVST, CVST with thrombocytopenia, and CVST without thrombocytopenia, within 28 days of first dose SARS-CoV-2 vaccination, using data from the European Medicines Agency's EudraVigilance database (until 13 June 2021). As a denominator, we used data on vaccine delivery from 31 European countries. For 22.8 million adults from 25 countries we estimated the absolute risk of CVST after the first dose of ChAdOx1 nCov-19 per age category. Results: The absolute risk of CVST within 28 days of first dose vaccination was 7.5 (95%CI 6.9- 8.3), 0.7 (95%CI 0.2-2.4), 0.6 (95%CI 0.5-0.7) and 0.6 (95%CI 0.3-1.1) per million of first doses of ChAdOx1 nCov-19, Ad26.COV2.S, BNT162b2 and mRNA-1273, respectively. The absolute risk of CVST with thrombocytopenia within 28 days of first dose vaccination was 4.4 (95%CI 3.9-4.9), 0.7 (95%CI 0.2-2.4), 0.0 (95%CI 0.0-0.1) and 0.0 (95%CI 0.0-0.2) per million of first doses of ChAdOx1 nCov-19, Ad26.COV2.S, BNT162b2 and mRNA-1273, respectively. In recipients of ChAdOx1 nCov-19, the risk of CVST, both with and without thrombocytopenia, was the highest in the 18-24 years age group (7.3 per million, 95%CI 2.8-18.8 and 3.7, 95%CI 1.0-13.3, respectively). The risk of CVST with thrombocytopenia was the lowest in ChAdOx1 nCov-19 recipients ≥70 years (0.2, 95%CI 0.0- 1.3). Age <60 compared to ≥60 was a predictor for CVST with thrombocytopenia (incidence rate ratio 5.79;95%CI 2.98-11.24, p<0.001). Discussion: The risk of CVST with thrombocytopenia within 28 days of first dose vaccination with ChAdOx1 nCov-19 was higher in younger age groups. The risk of CVST with thrombocytopenia was slightly increased in patients receiving Ad26.COV2.S, comparing with the estimated background risk. The risk of CVST with thrombocytopenia was not increased in recipients of mRNA vaccines for SARS-CoV-2.

Electrocardiogram-based mortality prediction in patients with COVID-19 using machine learning

Background The electrocardiogram (ECG) is an easy to assess, widely available and inexpensive tool that is frequently used during the work-up of hospitalized COVID-19 patients. So far, no study has been conducted to evaluate if ECG-based machine learning models are able to predict allcause in-hospital mortality in COVID-19 patients. Purpose With this study, we aim to evaluate the value of using the ECG to predict in-hospital all-cause mortality of COVID-19 patients by analyzing the ECG at hospital admission, comparing a logistic regression based approach and a DNN based approach. Secondly, we aim to identify specific ECG features associated with mortality in patients diagnosed with COVID-19. Methods and results We studied 882 patients admitted with COVID-19 across seven hospitals in the Netherlands. Raw-format 12-lead ECGs recorded after admission (<72 hours) were collected, manually assessed, and annotated using pre-defined ECG features. Using data from five out of seven centers (n = 634), two mortality prediction models were developed: (a) a logistic regression model using manually annotated ECG features, and (b) a pre-trained deep neural network (DNN) using the raw ECG waveforms. Data from two other centers (n = 248) were used for external validation. Performance of both prediction models was similar, with a mean area under the receiver operating curve of 0.69 [95%CI 0.55- 0.82] for the logistic regression model and 0.71 [95%CI 0.59-0.81] for the DNN in the external validation cohort. After adjustment for age and sex, ventricular rate (OR 1.13 [95% CI 1.01-1.27] per 10 ms increase), right bundle branch block (3.26 [95% CI 1.15-9.50]), ST-depression (2.78 [95% CI 1.03-7.70]) and low QRS voltages (3.09 [95% CI 1.02-9.38]) remained as significant predictors for mortality. Conclusion: This study shows that ECG-based prediction models at admission may be a valuable addition to the initial risk stratification in admitted COVID-19 patients. The DNN model showed similar performance to the logistic regression that needs time-consuming manual annotation. Several ECG features associated with mortality were identified.

Clinical features and prognostic factors in Covid-19: A prospective cohort study.

BACKGROUND: Mortality rates are high among hospitalized patients with COVID-19, especially in those intubated on the ICU. Insight in pathways associated with unfavourable outcome may lead to new treatment strategies. METHODS: We performed a prospective cohort study of patients with COVID-19 admitted to general ward or ICU who underwent serial blood sampling. To provide insight in the pathways involved in disease progression, associations were estimated between outcome risk and serial measurements of 64 biomarkers in potential important pathways of COVID-19 infection (inflammation, tissue damage, complement system, coagulation and fibrinolysis) using joint models combining Cox regression and linear mixed-effects models. For patients admitted to the general ward, the primary outcome was admission to the ICU or mortality (unfavourable outcome). For patients admitted to the ICU, the primary outcome was 12-week mortality. FINDINGS: A total of 219 patients were included: 136 (62%) on the ward and 119 patients (54%) on the ICU; 36 patients (26%) were included in both cohorts because they were transferred from general ward to ICU. On the general ward, 54 of 136 patients (40%) had an unfavourable outcome and 31 (23%) patients died. On the ICU, 54 out of 119 patients (45%) died. Unfavourable outcome on the general ward was associated with changes in concentrations of IL-6, IL-8, IL-10, soluble Receptor for Advanced Glycation End Products (sRAGE), vascular cell adhesion molecule 1 (VCAM-1) and Pentraxin-3. Death on the ICU was associated with changes in IL-6, IL-8, IL-10, sRAGE, VCAM-1, Pentraxin-3, urokinase-type plasminogen activator receptor, IL-1-receptor antagonist, CD14, procalcitonin, tumor necrosis factor alfa, tissue factor, complement component 5a, Growth arrest-specific 6, angiopoietin 2, and lactoferrin. Pathway analysis showed that unfavourable outcome on the ward was mainly driven by chemotaxis and interleukin production, whereas death on ICU was associated with a variety of pathways including chemotaxis, cell-cell adhesion, innate host response mechanisms, including the complement system, viral life cycle regulation, angiogenesis, wound healing and response to corticosteroids. INTERPRETATION: Clinical deterioration in patients with severe COVID-19 involves multiple pathways, including chemotaxis and interleukin production, but also endothelial dysfunction, the complement system, and immunothrombosis. Prognostic markers showed considerable overlap between general ward and ICU patients, but we identified distinct differences between groups that should be considered in the development and timing of interventional therapies in COVID-19. FUNDING: Amsterdam UMC, Amsterdam UMC Corona Fund, and Dr. C.J. Vaillant Fonds.