Background: During a pandemic, it is important for clinicians to stratify patients and decide who receives limited medical resources. Machine learning models have been proposed to accurately predict COVID-19 disease severity. Previous studies have typically tested only one machine learning algorithm and limited performance evaluation to area under the curve analysis. To obtain the best results possible, it may be important to test different machine learning algorithms to find the best prediction model. Objective: In this study, we aimed to use automated machine learning (autoML) to train various machine learning algorithms. We selected the model that best predicted patients’ chances of surviving a SARS-CoV-2 infection. In addition, we identified which variables (ie, vital signs, biomarkers, comorbidities, etc) were the most influential in generating an accurate model. Methods: Data were retrospectively collected from all patients who tested positive for COVID-19 at our institution between March 1 and July 3, 2020. We collected 48 variables from each patient within 36 hours before or after the index time (ie, real-time polymerase chain reaction positivity). Patients were followed for 30 days or until death. Patients’ data were used to build 20 machine learning models with various algorithms via autoML. The performance of machine learning models was measured by analyzing the area under the precision-recall curve (AUPCR). Subsequently, we established model interpretability via Shapley additive explanation and partial dependence plots to identify and rank variables that drove model predictions. Afterward, we conducted dimensionality reduction to extract the 10 most influential variables. AutoML models were retrained by only using these 10 variables, and the output models were evaluated against the model that used 48 variables. Results: Data from 4313 patients were used to develop the models. The best model that was generated by using autoML and 48 variables was the stacked ensemble model (AUPRC=0.807). The two best independent models were the gradient boost machine and extreme gradient boost models, which had an AUPRC of 0.803 and 0.793, respectively. The deep learning model (AUPRC=0.73) was substantially inferior to the other models. The 10 most influential variables for generating high-performing models were systolic and diastolic blood pressure, age, pulse oximetry level, blood urea nitrogen level, lactate dehydrogenase level, D-dimer level, troponin level, respiratory rate, and Charlson comorbidity score. After the autoML models were retrained with these 10 variables, the stacked ensemble model still had the best performance (AUPRC=0.791). Conclusions: We used autoML to develop high-performing models that predicted the survival of patients with COVID-19. In addition, we identified important variables that correlated with mortality. This is proof of concept that autoML is an efficient, effective, and informative method for generating machine learning–based clinical decision support tools.

Fibrosis is a characteristic of Duchenne muscular dystrophy (DMD), yet the cellular and molecular mechanisms responsible for DMD fibrosis are poorly understood. Utilizing the Collagen1a1‐GFP transgene to identify cells producing Collagen‐I matrix in wild‐type mice exposed to toxic injury or those mutated at the dystrophin gene locus (mdx) as a model of DMD, we studied mechanisms of skeletal muscle injury/repair and fibrosis. PDGFRα is restricted to Sca1+, CD45− mesenchymal progenitors. Fate‐mapping experiments using inducible CreER/LoxP somatic recombination indicate that these progenitors expand in injury or DMD to become PDGFRα+, Col1a1‐GFP+ matrix‐forming fibroblasts, whereas muscle fibres do not become fibroblasts but are an important source of the PDGFRα ligand, PDGF‐AA. While in toxin injury/repair of muscle PDGFRα, signalling is transiently up‐regulated during the regenerative phase in the DMD model and in human DMD it is chronically overactivated. Conditional expression of the constitutively active PDGFRα D842V mutation in Collagen‐I+ fibroblasts, during injury/repair, hindered the repair phase and instead promoted fibrosis. In DMD, treatment of mdx mice with crenolanib, a highly selective PDGFRα/β tyrosine kinase inhibitor, reduced fibrosis, improved muscle strength, and was associated with decreased activity of Src, a downstream effector of PDGFRα signalling. These observations are consistent with a model in which PDGFRα activation of mesenchymal progenitors normally regulates repair of the injured muscle, but in DMD persistent and excessive activation of this pathway directly drives fibrosis and hinders repair. The PDGFRα pathway is a potential new target for treatment of progressive DMD. © 2016 The Authors. The Journal of Pathology published by John Wiley & Sons Ltd on behalf of Pathological Society of Great Britain and Ireland.

Antiphospholipid syndrome (APS) is an autoimmune disease that manifests as arterial, venous, or microcirculatory thromboses as well as obstetric complications.¹ The diagnosis of APS requires the presence of IgG or IgM anti-β2-glycoprotein-1 or anticardiolipin antibodies by enzyme-linked immunosorbent assay or lupus anticoagulant (LA) assays that must persist for more than 12 weeks. Coagulopathy in patients with coronavirus disease 2019 (COVID-19) is a common complication that jeopardizes the clinical course and is associated with poorer outcomes.^2,3 This COVID-19 coagulopathy presents mainly as a prothrombotic state, and there is evidence that anticoagulation may reduce mortality rates.⁴ The partial thromboplastin time (PTT) has been found to be prolonged in many patients with COVID-19 and may indicate the presence of LA.⁵ Most patients with COVID-19 have elevated levels of C-reactive protein (CRP), and CRP is known to interfere with LA PTT-based tests, such as the hexagonal phase phospholipid neutralization assay STACLOT-LA.⁶

Convalescent plasma with severe acute respiratory disease coronavirus 2 (SARS-CoV-2) antibodies (CCP) may hold promise as a treatment for coronavirus disease 2019 (COVID-19). We compared the mortality and clinical outcome of patients with COVID-19 who received 200 mL of CCP with a spike protein IgG titer ≥ 1:2430 (median 1:47,385) within 72 hours of admission with propensity score–matched controls cared for at a medical center in the Bronx, between April 13 and May 4, 2020. Matching criteria for controls were age, sex, body mass index, race, ethnicity, comorbidities, week of admission, oxygen requirement, D-dimer, lymphocyte counts, corticosteroid use, and anticoagulation use. There was no difference in mortality or oxygenation between CCP recipients and controls at day 28. When stratified by age, compared with matched controls, CCP recipients less than 65 years had 4-fold lower risk of mortality and 4-fold lower risk of deterioration in oxygenation or mortality at day 28. For CCP recipients, pretransfusion spike protein IgG, IgM, and IgA titers were associated with mortality at day 28 in univariate analyses. No adverse effects of CCP were observed. Our results suggest CCP may be beneficial for hospitalized patients less than 65 years, but data from controlled trials are needed to validate this finding and establish the effect of aging on CCP efficacy.

The causes of coagulopathy associated with coronavirus disease 2019 (COVID-19) are poorly understood. We aimed to investigate the relationship between von Willebrand factor (VWF) biomarkers, intravascular hemolysis, coagulation, and organ damage in COVID-19 patients and study their association with disease severity and mortality. We conducted a retrospective study of 181 hospitalized COVID-19 patients randomly selected with balanced distribution of survivors and nonsurvivors. Patients who died had significantly lower ADAMTS13 (a disintegrin and metalloproteinase with a thrombospondin type 1 motif, member 13) activity, significantly elevated lactate dehydrogenase levels, significantly increased shistocyte/RBC fragment counts, and significantly elevated VWF antigen and activity levels compared with patients discharged alive. These biomarkers correlate with markedly elevated D-dimers. Additionally, only 30% of patients who had an ADAMTS13 activity level of less than 43% on admission survived, yet 60% of patients survived who had an ADAMTS13 activity level of greater than 43% on admission. In conclusion, COVID-19 may present with low ADAMTS13 activity in a subset of hospitalized patients. Presence of schistocytes/RBC fragment and elevated D-dimer on admission may warrant a work-up for ADAMTS13 activity and VWF antigen and activity levels. These findings indicate the need for future investigation to study the relationship between endothelial and coagulation activation and the efficacy of treatments aimed at prevention and/or amelioration of microangiopathy in COVID-19.