key: cord-0862093-b4jms02f
authors: Karagodin, Ilya; Singulane, Cristiane Carvalho; Woodward, Gary M.; Xie, Mingxing; Tucay, Edwin S.; Tude Rodrigues, Ana C.; Vasquez-Ortiz, Zuilma Y.; Alizadehasl, Azin; Monaghan, Mark J.; Ordonez Salazar, Bayardo A.; Soulat-Dufour, Laurie; Mostafavi, Atoosa; Moreo, Antonella; Citro, Rodolfo; Narang, Akhil; Wu, Chun; Descamps, Tine; Addetia, Karima; Lang, Roberto M.; Asch, Federico M.
title: Echocardiographic Correlates of In-Hospital Death in Patients with Acute COVID-19 Infection: The World Alliance Societies of Echocardiography (WASE-COVID) Study
date: 2021-05-21
journal: J Am Soc Echocardiogr
DOI: 10.1016/j.echo.2021.05.010
sha: ec16c5c51285b5274075ee9d43d1b027a58d8a65
doc_id: 862093
cord_uid: b4jms02f

Background The novel severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) virus which has led to the global Coronavirus disease-2019 (COVID-19) pandemic is known to adversely affect the cardiovascular system through multiple mechanisms. In this international, multi-center study conducted by the World Alliance Societies of Echocardiography (WASE), we aim to determine the clinical and echocardiographic phenotype of acute cardiac disease in COVID-19 patients, to explore phenotypic differences in different geographic regions across the world, and to identify parameters associated with in-hospital mortality. Methods We studied 870 patients with acute COVID-19 infection from 13 medical centers in four world regions (Asia, Europe, United States, Latin America) who had undergone transthoracic echocardiograms (TTEs). Clinical and laboratory data were collected, including patient outcomes. Anonymized echocardiograms were analyzed with automated, machine learning-derived algorithms to calculate left ventricular (LV) volumes, ejection fraction (EF), and LV longitudinal strain (LS). Right-sided echocardiographic parameters that were measured included right ventricular (RV) LS, RV free wall strain (FWS), and RV basal diameter (RVBD). Multivariate regression analysis was performed to identify clinical and echocardiographic parameters associated with in-hospital mortality. Results Significant regional differences were noted in terms of patient co-morbidities, severity of illness, clinical biomarkers, and LV and RV echocardiographic metrics. Overall in-hospital mortality was 21.6%. Parameters associated with mortality in a multivariate analysis were age (OR 1.12 [1.05, 1.22], p = 0.003), previous lung disease (OR 7.32 [1.56, 42.2], p = 0.015), LVLS (OR 1.18 [1.05, 1.36], p = 0.012), lactic dehydrogenase (LDH) (OR 6.17 [1.74, 28.7], p = 0.009), and RVFWS (OR 1.14 [1.04, 1.26], p = 0.007). Conclusions LV dysfunction is noted in approximately 20% and RV dysfunction in approximately 30% of patients with acute COVID-19 illness and portend a poor prognosis. Age at presentation, previous lung disease, LDH, LVLS, and RVFWS were independently associated with in-hospital mortality. Regional differences in cardiac phenotype highlight the significant differences in patient acuity as well as echocardiographic utilization in different parts of the world.

The novel severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) virus which has led to the global Coronavirus disease-2019 (COVID- 19) pandemic is known to adversely affect the cardiovascular system through multiple mechanisms. In this international, multi-center study conducted by the World Alliance Societies of Echocardiography (WASE), we aim to determine the clinical and echocardiographic phenotype of acute cardiac disease in COVID-19 patients, to explore phenotypic differences in different geographic regions across the world, and to identify parameters associated with in-hospital mortality.

We studied 870 patients with acute COVID-19 infection from 13 medical centers in four world regions (Asia, Europe, United States, Latin America) who had undergone transthoracic echocardiograms (TTEs). Clinical and laboratory data were collected, including patient outcomes. Anonymized echocardiograms were analyzed with automated, machine learningderived algorithms to calculate left ventricular (LV) volumes, ejection fraction (EF), and LV longitudinal strain (LS). Right-sided echocardiographic parameters that were measured included right ventricular (RV) LS, RV free wall strain (FWS), and RV basal diameter (RVBD).

Multivariate regression analysis was performed to identify clinical and echocardiographic parameters associated with in-hospital mortality.

Significant regional differences were noted in terms of patient co-morbidities, severity of illness, clinical biomarkers, and LV and RV echocardiographic metrics. Overall in-hospital mortality was 21.6%. Parameters associated with mortality in a multivariate analysis were age (OR 1.12 J o u r n a l P r e -p r o o f

The year 2020 has been marked by the global Coronavirus disease-2019 (COVID-19) pandemic, with over 108 million cases and 2.4 million deaths in 223 countries to date (1) . The novel virus known as severe acute respiratory syndrome-coronavirus-2 (SARS-CoV-2) has spread quickly across the globe and has inflicted historic levels of morbidity and mortality. While the respiratory system is the most directly affected, growing evidence suggests that COVID-related cardiovascular disease plays a significant role in disease severity and patient outcomes (2) (3) (4) .

Of note, the number of cases, deaths, and mortality rates vary considerably across various countries and regions. The reasons for such diversity have not yet been fully elucidated but could be socio-economic, genomic (5) , or multi-factorial in nature.

The International World Alliance Societies of Echocardiography (WASE) COVID-19 Study was designed to identify echocardiographic parameters that would be prognostic of clinical outcomes in patients with COVID-19 infection and to determine if COVID-19 heart disease presents differently in various geographic regions around the world. In this initial report of the WASE COVID-19 study we aim to describe the clinical characteristics and echocardiographic phenotype of acute cardiac disease in patients with COVID-19 infection and to explore their inhospital prognostic value.

J o u r n a l P r e -p r o o f

Adult patients (18 years old) with a diagnosis of SARS-CoV-2 infection (including a positive antigen or PCR test) during the first wave of the pandemic (January-September 2020) were considered for the study if a transthoracic echocardiogram (TTE) was performed during the initial COVID-19-related hospitalization. The data regarding hospital care and echocardiograms was collected retrospectively. Patients were enrolled at 13 medical centers in four world regions (Asia, Europe, United States, Latin America). The echocardiograms were ordered and acquired based on local clinical practices. Acceptable TTEs included both comprehensive and limited studies as long as at least the apical 4-chamber view was acquired. The study was approved by the local ethics or IRB committee.

Basic clinical information and DICOM cardiac ultrasound images were collected from the medical records, PACS systems or echo machines, deidentified, and transferred via a web-based system (Ultromics, Oxford, United Kingdom) for central analysis by a core group lead by the Principal Investigators.

Clinical information including demographic data, past medical history, vital signs, and biomarkers were collected by local investigators and stored in a secure web-based system (Castor EDC, United Kingdom). Blood pressure and heart rate were obtained at the time of the echocardiographic exam. Biomarkers were collected by each site whenever deemed clinically appropriate and all biomarkers collected within 72 hours of echocardiographic acquisition were included in the analysis. These biomarkers included troponin, lactic dehydrogenase (LDH), brain natriuretic peptide (BNP), D-Dimer, and C-Reactive Protein (CRP). To account for different biomarker assays used at each center, each biomarker (troponin, BNP, LDH, D-Dimer, and CRP) J o u r n a l P r e -p r o o f was classified as either normal, borderline abnormal (<2x upper limit of normal), or abnormal (>2x upper limit of normal), based on normal reference values for the particular center at which the patient was enrolled.

Image transfer was facilitated by a 2-step anonymization process to a cloud-based image analysis software. Left ventricular (LV) analyses were performed through commercially available artificial intelligence (AI) algorithms created by machine learning (Ultromics, United Kingdom), which generated an initial left ventricular ejection fraction (LVEF), end systolic and end diastolic volumes (LVESV, LVEDV) and longitudinal strain (LVLS); see additional details in Supplemental Material S1. As a means of quality control and validation, all LV measurements were repeated manually twice by independent British Society of Echocardiography accredited echocardiographer (human reads 1 and 2). The two human reads were performed independently by the same echocardiographer (to derive intra-observer variability) who was randomly selected from a pool of 7 independent operators, blinded to other echo reads and clinical data. All LV volumes and EFs were obtained by performing endocardial tracings and using the biplane method of disks (modified Simpson's Rule) (6) . Only cases with acceptable quality LV views were included, which was defined as lack of apical foreshortening with adequate visualization of all segments in the apical 4-chamber (4CH) view. LS was calculated as the average of all available segments from the 4CH and 2CH views, as a long axis view was not obtained in the vast majority of cases. Cut-offs for mild, moderately, and severely reduced LVEF as well as normal and abnormal LVLS were determined by the 2015 ASE/EACVI Guidelines for Cardiac Chamber Quantification (6) . Normal LVLS was defined as <-19%, mildly abnormal LVLS was defined as >-19% and <-15%, and severely abnormal LVLS was defined as >-15% (6, 7) . Only cases with acceptable quality RV views were included. Acceptable imaging quality was defined as presence of an RV-focused view with adequate visualization of the RV free wall.

Abnormal RVFWS was defined as >-20% (8) .

The primary outcome was defined as in-hospital death.

Continuous variables were expressed as mean (±SD) or median (interquartile range (IQR)) according to data distribution and compared using the Student t test or Wilcoxon ranks sum test, as appropriate. Categorical data, presented as number and percentages, were compared using χ² test. Binomial generalized linear models were used to evaluate the univariate and multivariate association between covariates (age, sex, ethnicity, region, baseline comorbidities, etc.), echocardiographic parameters and the outcome. Forward-stepwise binomial regression models (in which the choice of variables was carried out by an automatic procedure) were performed to evaluate the association between confounders, multiple variables and outcome and included the following univariate significant covariates: previous heart disease, previous lung disease, age, geographical region, LVEF, LVLS, RVFWS, LDH (log), BNP (log), CRP (log), shock, and admission to an intensive care unit (ICU). In this stepwise regression model, the choice of variables was carried out autonomously, by which the addition of each step-wise selected variable gave the most statistically significant improvement of the fit. This multivariate modelling method selects the best fitting model. All significant univariate variables were offered as input to the model initially, which then statistically applies an independent stepwise hierarchy J o u r n a l P r e -p r o o f to selecting the best fitting covariate from this input list. Results from regression models are reported as odds ratios (OR) and 95% confidence intervals.

As echocardiograms in the context of the COVID-19 pandemic were performed with operator safety in mind (9), not all traditional views were uniformly acquired. In order to determine accurate values and to build a homogeneous database, missing data for calculation of biplane LVEF and LVLS were determined using a multiple imputation model, following guidelines from the European Medicines Agency on confirmatory clinical trials (10) . Specifically, a Multiple Imputation by Chained Equations (MICE) method was used to derive the 2CH values for cases with a 4CH value but missing 2CH (n=89), in order to calculate biplane measures. The mean of the three LV reads (automated AI, human read 1 and human read 2) was taken as the final value.

AI analysis was used to derive inter-observer (human vs AI) variability. Reproducibility of LV measurements was tested on the entire cohort and in 40 random cases for the RV. Inter-and intra-observer variability were assessed using intra-class correlation coefficients (ICC), coefficients of variation (CoV, calculated as the absolute difference between pairs of repeated measurements in percent of their mean value), and mean of differences (see Supplemental Table   S2 ).

Over a nine-month period (January to September 2020), 870 patients were enrolled at 13 centers in nine countries. Baseline patient demographic and clinical characteristics are listed in Table 1 .

Median age was 60 years old and 56.1% were male. One hundred twenty-five patients (14.3%) were enrolled in the United States, 238 (27.4%) in Latin America, 160 (18.4%) in Europe, and 347 (39.9%) in Asia. The distribution of enrolled cases from each of the four global regions according to calendar month in 2020 is shown in Figure 1 , which depicts the rapid expansion of J o u r n a l P r e -p r o o f the pandemic during its first wave, starting in China, followed by Europe, the United States and Latin America in that chronological order.

By protocol design, all patients were hospitalized. At the time of echocardiogram acquisition 46.2% were admitted to an intensive care unit (ICU), 27.1% required mechanical ventilation, and 17.9% required hemodynamic support (inotropic drugs, vasopressors, intra-aortic balloon pump) ( Table 1) .

Troponin was tested in 301 patients and was elevated in 94% (borderline 22.6%, abnormal 71.4%) . CRP was recorded in 792 patients and was elevated in 87% (borderline 6.4%, abnormal 80.2%), while BNP was recorded in 359 patients and was elevated in 57% (borderline 12.8%, abnormal 44.6%). LDH was recorded in 524 patients and was elevated in 78% (borderline 48.7%, abnormal 29%) and D-Dimer was recorded in 614 patients and was elevated in 86% (borderline 16%, abnormal 70.2%).

Echocardiograms were obtained a median of 3 days after admission (IQ 1-9). Due to safety concerns related to transmission of the COVID-19 virus, enrolling centers followed different echocardiographic acquisition protocols and therefore completeness of the echocardiographic studies was variable: 10 of the 13 centers performed limited exams as their primary COVID inpatient practice and 3 out of the 13 centers performed comprehensive exams. The median number of video loops acquired per exam was 32 (Q1-Q3: 19-42). The 4CH view was evaluable in 722 exams (83% of enrolled patients), while both the 2CH and 4CH views were evaluable in 633 of them (75% of enrolled patients). The remaining 89 cases had 4Ch without 2Ch views, for which the 2Ch value was estimated by multiple imputation, following guidelines from the European Medicines Agency, as described in the methods section. Only cases in which a 4CH J o u r n a l P r e -p r o o f view was available for quantification were used for LV analysis. Baseline characteristics and outcomes of these 722 patients were similar to the entire cohort and those excluded from LV analysis (Supplemental Table S1 ). Of the available cases, the percentage of echocardiograms with good image quality by region were 88.2% for Asia, 86.6% for Europe, 86.7% for USA, and 69.3% for Latin America. Echocardiograms included a dedicated RV-focused view of sufficient quality for strain analysis in 589 patients (68% of enrolled patients). The mean monthly LVEF and RVFWS measurements remained relatively constant throughout the calendar year (Supplemental Figure S1 ).

The mean LVEF was 60.2% (±12.3), LVEDV was 107.9 (±45.1) and LVESV was 44.8 mL (±33.7) ( Table 2) . Eighty-three percent of the patients had LVEF >50%, 11% had mild LV dysfunction (EF 40-50%), 5% moderate dysfunction (LVEF 30-40%), and 3% had severe LV dysfunction (LVEF <30%). The mean LVLS was -18.7% (±5.3) and mean RVFWS was -22.8% (±6.1). Severely abnormal LVLS (>-15%) and abnormal RVFWS (>-20%) was present in 22% and 29% of patients, respectively. The mean RVBD was 40 mm (±25) and 33% had a dilated RV (RVBD>42 mm). The overall distribution of patients according to ventricular function (LVLS and RVFWS) and region can be seen in Supplemental Figure S2 . Among J o u r n a l P r e -p r o o f

Inter and intra-observer variability analysis, as defined by intra-class correlation coefficients (ICC), coefficient of variance (CoV) and mean of differences are presented in Supplemental   Tables S2 and S3 . Reproducibility was excellent for intraobserver LV variables and good for interobserver LVEF and LVLS. Both intra and interobserver reproducibility of RVFWS and RVBD were excellent, while very good for RV LS.

Comparison of baseline clinical, treatment and echocardiographic characteristics suggest significant regional differences in patients with acute COVID-19 infection throughout the first wave of the pandemic. Compared to those enrolled in the US, Latin America, and Europe, patients enrolled in Asia had less comorbid conditions (heart and lung disease, Figure 2) , a better biomarker profile (LDH, BNP and D-Dimer) (Figure 3) , and required less hemodynamic/LV support and mechanical ventilation (Supplemental Figure S3) . Importantly, regional differences were also observed in terms of LVEF, LVLS, RVFWS, and RVBD, with best values observed in Asia, followed in worsening order by Europe, Latin America, and the USA. LV and RV measurements by region are shown in Figure 4 .

Overall in-hospital mortality was 21.6 % (188 patients). In-hospital mortality in each region is shown in Figure 5 . As assessed by univariate analysis (Table 3) In the multivariate analysis, the independent clinical parameters were age, LDH and previous lung disease. Furthermore, LVLS and RVFWS demonstrated independent associations with inhospital mortality (Table 3) . Since LVLS and RVFWS appeared to be co-correlates, these variables were modelled separately.

Receiver-operating characteristic (ROC) curves were used to generate the optimal cut-off values for continuous variables (age, LDH, LVLS, and RVFWS) for associations with mortality (Supplemental Figure S5) . The optimal cut-off values were 64 years for age (AUC=0.71), 389 U/L for LDH (AUC=0.74), -16.7% for LVLS (AUC=0.59), and -20.2% for RVFWS (AUC=0.65). Using a composite ROC model with all these optimal cut-offs, we identified the best performing combination for association with mortality, which had 80.0% sensitivity, 66.5% specificity and an AUC=0.80 (Supplemental Figure S6 ).

In this report from the global WASE COVID-19 study, we have shown that LV dysfunction is observed in approximately 20% and RV dysfunction in approximately 30% of patients with acute SARS-CoV-2 infection, and that age at presentation, previous lung disease, LDH, LVLS and RVFWS are independently associated with in-hospital mortality. This is the first large, international study including 13 medical centers spanning nine countries and four world regions, designed to characterize cardiac phenotype in patients with acute COVID-19 infection. Overall, these observations can be explained by regional differences in baseline co-morbidities and patient acuity or may, in fact, reflect a population bias stemming from changes in echocardiographic practices as a result of growing concern for viral spread and more ubiquitous implementation of safe practices. A few months into the global COVID-19 pandemic, providers were urged by ASE/ACC recommendations to defer all elective examinations and to perform problem-focused, limited echocardiograms whenever possible, and only for urgent or emergent indications (9). The regional differences observed in the WASE COVID-19 study may partially reflect the different echocardiographic practices in various parts of the world, with American providers performing TTEs primarily on the most critically ill patients (as recommended by the ASE/ACC) as compared to a more broader patient population in other regions of the world.

Another possible explanation is that given the earlier onset of the pandemic, patients recruited from Asia may have demonstrated "burned out" COVID-19 disease on echocardiography that was not as severe compared to other regions, who may eventually show similarly improved cardiac phenotypes. Through future studies and continued follow-up of our patient cohort, we will aim to determine whether cardiac involvement has changed over time since the initial months of the pandemic.

Early studies out of Wuhan, China demonstrated a link between myocardial injury (as determined by troponin elevation) and increased risk of mechanical ventilation and mortality (11, 12) . Similarly, a subsequent study from New York City demonstrated that a significant proportion of patients admitted with COVID-19 infection had cardiac involvement (36%) and J o u r n a l P r e -p r o o f that troponin elevation was associated with higher risk of death (4). In our population, troponin was not associated with death; while other biomarkers were significant in the univariate analysis (LDH, BNP, D-dimer and CRP), though LDH was the only independent parameter associated with mortality.

The lack of association with mortality for Troponin in our study may have several explanations.

On one hand there may be a component of selection bias, as it was tested only in 35% of the patients and was elevated in nearly all of them. However, the elevated troponin contrasts with the overall good LV and RV function of the population. In patients with sepsis, such as those with severe COVID-19 infection, elevations in troponin may not necessarily be the result of direct necrosis of cardiomyocytes, but may be the result of increased cell permeability leading to the release of troponin degradation products through the cell membrane in non-necrotic cardiomyocytes (13, 14) . While several studies have concluded that elevated troponin is predictive of mortality in sepsis, other studies did not find it to be an independent predictor of mortality (13, (15) (16) (17) . Given the severity of COVID-19 infection of our patient population, it is possible that while elevated troponin served as a marker of severe illness, it was in many cases not reflective of direct myocardial damage or injury, as suggested by a weak correlation between troponin and LVLS, and was therefore by itself not associated with mortality.

More recently, reports from the American Heart Association COVID-19 Cardiovascular Disease registry described that cardiovascular co-morbidities including coronary artery disease, hypertension, and diabetes dramatically increased the risk of in-hospital mortality in patients with COVID-19 infection, and that the risk of death was particularly high for older, non-white males (3, 18) . Although these results are consistent with our findings, the WASE COVID-19 J o u r n a l P r e -p r o o f study is unique in that we included multiple cardiac markers during the acute phase of the infection to describe clinical and echocardiographic parameters associated with short-term outcomes. In addition, the cardiac phenotypic data was obtained by independent, centralized analysis using advanced echocardiographic techniques including automated, AI-driven technologies. A recent international survey conducted by the European Association of Cardiovascular Imaging (EACVI) in April 2020 on 1216 patients with presumed or confirmed COVID-19 infection showed a slightly higher proportion of left and right-sided involvement (39 and 33%, respectively) than observed in our study. (19) This survey, however, was conducted during a short two-week time period during the height of the pandemic, while the WASE COVID-19 study was conducted over a longer period spanning a larger portion of the pandemic, only included patients with laboratory confirmed SARS-CoV-2 infection, and echocardiograms were analyzed centrally by a core group. Of note, the percentage of left-sided involvement seen in the COVID-19 patients in our study was similar to that observed in a recent study by Churchill et al (20) though slightly less than that observed by Sud et al (21) .

Multiple mechanisms have been proposed for myocardial injury following COVID-19 infection, including direct injury caused by binding of the SARS-CoV-2 spike protein to the ACE-2 receptor in the myocardium leading to a myocarditis-like syndrome, as well as indirect injury caused by an inflammatory storm induced by the body's immune response (22, 23) . Given the known link between COVID-19 and cardiovascular disease, identification of the underlying cardiac abnormalities in patients with COVID-19 infection and myocardial injury using available imaging modalities is critically important. In a small series of patients with cardiac magnetic resonance imaging, a significant proportion of cardiac involvement in patients recovered from COVID-19 infection was noted, as evidenced by myocardial edema in 54% of patients and late J o u r n a l P r e -p r o o f gadolinium enhancement in 31% of patients (22) . In a recent study of 305 patients from New York City and Milan who had undergone transthoracic echocardiographic (TTE) examinations, Giustino et al showed that patients with myocardial injury had an increased prevalence of major echocardiographic abnormalities and that following multivariable adjustment, myocardial injury with TTE abnormalities was associated with a higher risk of death compared to myocardial injury without TTE abnormalities (24) . In the WASE COVID-19 study, we have performed central, blinded analyses of echocardiograms that included LVEF and strain of both ventricles and have shown that both left-sided and right-sided markers of function (LVLS and RVFWS) are independently associated with mortality, while an elevated troponin level was not. Using a composite ROC model, we showed that the sensitivity of the model for association with mortality is improved when adding strain-based echocardiographic parameters to clinical parameters. Importantly, LVEF in our study failed to show independent association with mortality in a model that included strain variables (which differed from recent findings published by Faridi et al (25) ), while LVLS and RVFWS did, further highlighting the importance of advanced echocardiographic assessment of COVID-19 patients to obtain a more complete risk assessment. Importantly, beyond its association with mortality, echocardiography is an important diagnostic tool in COVID-19 that can help point the provider towards a particular diagnosis, for example ischemic heart disease in the setting of regional wall motion abnormalities or sepsis in the setting of hyperdynamic LV function.

Because the lungs are the main target organ of SARS-CoV-2 and given the large prevalence of acute respiratory distress syndrome (ARDS) in critically ill patients with COVID-19 infection, the right ventricle (RV) is thought to be particularly susceptible to dysfunction following COVID-19 infection (26) . Previous studies have demonstrated RV failure as a sequelae of acute J o u r n a l P r e -p r o o f lung injury and ARDS (27) , as the RV is easily affected by changes in pulmonary vascular resistance (28) . Investigators from Israel showed that in 100 patients diagnosed with COVID-19 infection who had undergone TTE, RV dilatation and dysfunction was the most common echocardiographic finding, and that worsening RV function was the most common finding in patients with clinical deterioration (29) . A group of investigators from Wuhan, China found that RVLS was the most predictive of mortality in patients with COVID-19 infection compared to other conventional RV parameters (26) . In view of the above, and because the lungs are the main target organ of the SARS-CoV-2 virus, our finding that RVFWS is independently associated with mortality in COVID-19 patients worldwide makes sense from a pathophysiologic standpoint and confirms the findings from Israel and China.

Importantly, our study's findings reflect the cardiac manifestations of COVID-19 infection during the early months (first wave) of the global pandemic, starting in January 2020. Since that time, we have seen rapid improvement in early diagnosis, therapeutics, and management strategies, resulting in improved mortality rates across health systems, even when adjusted for demographic and clinical factors (30) . In the WASE-COVID 19 study, we did see an improvement in overall in-hospital mortality from July to September (Supplemental Figure S4) , which could be attributed both to seasonal factors as well as to improved therapeutics as noted above.

Finally, the WASE COVID-19 study was unique in its use of artificial intelligence (AI) and machine learning to calculate the main left-sided parameters (LVEF and LVLS) in this study.

We also conducted a robust reproducibility analysis with several rounds of manual measurements to ensure that the AI-based results were both reliable and reproducible. To our knowledge, this is the largest multi-center, echocardiographic-based study on patients with COVID-19 infection that has employed AI and machine learning for determination of both LV and RV function.

The limitations of our study include a relatively small number of patients compared to the overall scale of the global COVID-19 pandemic. In spite of this limitation, our patient cohort included patients from nine different countries representing most continents, including some of the most affected regions worldwide. Additionally, overall image quality of echocardiograms was less than optimal compared to non-pandemic standards, reflecting widespread safety concerns during the early months of the pandemic as well as lack of a standardized, universal acquisition We also acknowledge that while LVLS and RVFWS were independently associated with mortality, their individual sensitivity is limited, therefore affecting the ROC analysis (AUC 0.59 and 0.65, respectively, as seen in Supplemental Figure S5) , and thus should be used in conjunction with other clinical prognostic factors such as age and LDH. Finally, it is possible that the clinical indications for echocardiograms resulted in selection bias, with our study population reflecting mostly patients in the severe spectrum of the disease. Nevertheless, this study provides valuable insights into the global utilization of echocardiography during the COVID-19 pandemic and demonstrates that in cases where echocardiography is clinically indicated, it provides crucial prognostic information.

This study confirms the profound effect of the SARS-CoV-2 virus on the cardiovascular system and highlights the significant regional differences in both cardiac phenotype and echocardiographic utilization. LVLS, RVFWS, age, LDH, and previous lung disease were independently associated with in-hospital mortality, while LVEF was not.

Tables. Table 1 J o u r n a l P r e -p r o o f

Instituto Nacional de Ciencias Medicas y Nutricion (INCMNSZ)

King's College Hospital

Coronavirus disease (COVID-19) pandemic

Myocardial Injury in COVID-19 Patients: The Beginning or the End?

Impact of Cardiovascular Disease on Outcomes Among Hospitalized COVID-19 Patients: Results From >14,000 Patients Across the United States. Virtual AHA Scientific Sessions

Prevalence and Impact of Myocardial Injury in Patients Hospitalized With COVID-19 Infection

SARS-CoV-2 genomic variations associated with mortality rate of COVID-19

Recommendations for cardiac chamber quantification by echocardiography in adults: an update from the American Society of Echocardiography and the European Association of Cardiovascular Imaging

Research to Practice: Assessment of Left Ventricular Global Longitudinal Strain for Surveillance of Cancer Chemotherapeutic-Related Cardiac Dysfunction

Strain Analysis of the Right Ventricle Using Two-dimensional Echocardiography

ASE Statement on Protection of Patients and Echocardiography Service Providers During the 2019 Novel Coronavirus Outbreak: Endorsed by the American College of Cardiology

Characteristics and clinical significance of myocardial injury in patients with severe coronavirus disease 2019

Association of Cardiac Injury With Mortality in Hospitalized Patients With COVID-19 in Wuhan, China

Circulating high sensitivity troponin T in severe sepsis and septic shock: distribution, associated factors, and relation to outcome

Increased troponin in patients with sepsis and septic shock: myocardial necrosis or reversible myocardial depression?

Cardiac troponin I does not independently predict mortality in critically ill patients with severe sepsis

Elevated cardiac troponin levels in critically ill patients: prevalence, incidence, and outcomes

The prognostic value of atrial and brain natriuretic peptides, troponin I and C-reactive protein in patients with sepsis

Racial and Ethnic Differences in Presentation and Outcomes for Patients Hospitalized with COVID-19: Findings from the American Heart Association's COVID-19 Cardiovascular Disease Registry

Global evaluation of echocardiography in patients with COVID-19

Echocardiographic Features of COVID-19 Illness and Association with Cardiac Biomarkers

Echocardiographic Findings in Patients with COVID-19 with Significant Myocardial Injury

Cardiac Involvement in Patients Recovered From COVID-2019 Identified Using Magnetic Resonance Imaging

Clinical features of patients infected with 2019 novel coronavirus in Wuhan, China

Characterization of Myocardial Injury in Patients With COVID-19

Left Ventricular Systolic Function and Inpatient Mortality in Patients Hospitalized with Coronavirus Disease 2019 (COVID-19)

Prognostic Value of Right Ventricular Longitudinal Strain in Patients With COVID-19

Incidence and prognostic value of right ventricular failure in acute respiratory distress syndrome

Right ventricular failure in acute lung injury and acute respiratory distress syndrome

Spectrum of Cardiac Manifestations in COVID-19: A Systematic Echocardiographic Study

Risk-Adjusted Mortality Rates

Coronavirus disease (COVID-19) pandemic

Myocardial Injury in COVID-19 Patients: The Beginning or the End?

Impact of Cardiovascular Disease on Outcomes Among Hospitalized COVID-19 Patients: Results From >14,000 Patients Across the United States. Virtual AHA Scientific Sessions

Prevalence and Impact of Myocardial Injury in Patients Hospitalized With COVID-19 Infection

SARS-CoV-2 genomic variations associated with mortality rate of COVID-19

Recommendations for cardiac chamber quantification by echocardiography in adults: an update from the American Society of Echocardiography and the European Association of Cardiovascular Imaging

Research to Practice: Assessment of Left Ventricular Global Longitudinal Strain for Surveillance of Cancer Chemotherapeutic-Related Cardiac Dysfunction

Strain Analysis of the Right Ventricle Using Two-dimensional Echocardiography

ASE Statement on Protection of Patients and Echocardiography Service Providers During the 2019 Novel Coronavirus Outbreak: Endorsed by the American College of Cardiology

Characteristics and clinical significance of myocardial injury in patients with severe coronavirus disease 2019

Association of Cardiac Injury With Mortality in Hospitalized Patients With COVID-19 in Wuhan, China

Circulating high sensitivity troponin T in severe sepsis and septic shock: distribution, associated factors, and relation to outcome

Increased troponin in patients with sepsis and septic shock: myocardial necrosis or reversible myocardial depression?

Cardiac troponin I does not independently predict mortality in critically ill patients with severe sepsis

Elevated cardiac troponin levels in critically ill patients: prevalence, incidence, and outcomes

The prognostic value of atrial and brain natriuretic peptides, troponin I and C-reactive protein in patients with sepsis

Racial and Ethnic Differences in Presentation and Outcomes for Patients Hospitalized with COVID-19: Findings from the American Heart Association's COVID-19 Cardiovascular Disease Registry

Global evaluation of echocardiography in patients with COVID-19

Echocardiographic Features of COVID-19 Illness and Association with Cardiac Biomarkers

Echocardiographic Findings in Patients with COVID-19 with Significant Myocardial Injury

Cardiac Involvement in Patients Recovered From COVID-2019 Identified Using Magnetic Resonance Imaging

Clinical features of patients infected with 2019 novel coronavirus in Wuhan, China

Characterization of Myocardial Injury in Patients With COVID-19

Left Ventricular Systolic Function and Inpatient Mortality in Patients Hospitalized with Coronavirus Disease 2019 (COVID-19)

Prognostic Value of Right Ventricular Longitudinal Strain in Patients With COVID-19

Incidence and prognostic value of right ventricular failure in acute respiratory distress syndrome

Right ventricular failure in acute lung injury and acute respiratory distress syndrome

Spectrum of Cardiac Manifestations in COVID-19: A Systematic Echocardiographic Study

Trends in COVID-19 Risk-Adjusted Mortality Rates

BNP (B), and D-Dimer (C) biomarkers. In all cases, the highest values were observed in the United States, followed by Europe, Latin America, and then Asia, with ranges depicted in the plots above. Central dots and lines represent means±SD. NS, not significant

Violin plots depicting regional differences in left ventricular ejection fraction (LVEF; A), left ventricular longitudinal strain (LVLS; B), right ventricular free wall strain (C) and right ventricular basal diameters (D)

Katie Ions, Nancy Spagou, Alex Hudson, Jake Kenworthy, Lokken Wong, Angela Mumith, Will Hawkes, Rizwan Sarwar from Ultromics; Marcus Schreckenberg, Niklas Hitschrich, Sabrina Zoernack from TOMTEC; Andrea Van Hoever from the American Society of Echocardiography; Patricia Marques, Gaynor Jones, Seamus Walker, Gurpreet Dhadday, Aiko Nepomuceno, Jherick Tay, Claire Mitchell, Sarah Hastings, Emma-Jayne Robins.