Machine learning based on clinical characteristics and chest CT quantitative measurements for prediction of adverse clinical outcomes in hospitalized patients with COVID-19.

OBJECTIVES: To develop and validate a machine learning model for the prediction of adverse outcomes in hospitalized patients with COVID-19. METHODS: We included 424 patients with non-severe COVID-19 on admission from January 17, 2020, to February 17, 2020, in the primary cohort of this retrospective multicenter study. The extent of lung involvement was quantified on chest CT images by a deep learning-based framework. The composite endpoint was the occurrence of severe or critical COVID-19 or death during hospitalization. The optimal machine learning classifier and feature subset were selected for model construction. The performance was further tested in an external validation cohort consisting of 98 patients. RESULTS: There was no significant difference in the prevalence of adverse outcomes (8.7% vs. 8.2%, p = 0.858) between the primary and validation cohorts. The machine learning method extreme gradient boosting (XGBoost) and optimal feature subset including lactic dehydrogenase (LDH), presence of comorbidity, CT lesion ratio (lesion%), and hypersensitive cardiac troponin I (hs-cTnI) were selected for model construction. The XGBoost classifier based on the optimal feature subset performed well for the prediction of developing adverse outcomes in the primary and validation cohorts, with AUCs of 0.959 (95% confidence interval [CI]: 0.936-0.976) and 0.953 (95% CI: 0.891-0.986), respectively. Furthermore, the XGBoost classifier also showed clinical usefulness. CONCLUSIONS: We presented a machine learning model that could be effectively used as a predictor of adverse outcomes in hospitalized patients with COVID-19, opening up the possibility for patient stratification and treatment allocation. KEY POINTS: • Developing an individually prognostic model for COVID-19 has the potential to allow efficient allocation of medical resources. • We proposed a deep learning-based framework for accurate lung involvement quantification on chest CT images. • Machine learning based on clinical and CT variables can facilitate the prediction of adverse outcomes of COVID-19.

Performance of nucleocapsid and spike-based SARS-CoV-2 serologic assays.

There is an urgent need for an accurate antibody test for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). We have developed 3 ELISA methods, trimer spike IgA, trimer spike IgG, and nucleocapsid IgG, for detecting anti-SARS-CoV-2 antibodies. We evaluated their performance along with four commercial ELISAs, EDI™ Novel Coronavirus COVID-19 ELISA IgG and IgM, Euroimmun Anti-SARS-CoV-2 ELISA IgG and IgA, and one lateral flow assay, DPP® COVID-19 IgM/IgG System (Chembio). Both sensitivity and specificity were evaluated and the probable causes of false-positive reactions were determined. The assays were evaluated using 300 pre-epidemic samples and 100 PCR-confirmed COVID-19 samples. The sensitivities and specificities of the assays were as follows: 90%/100% (in-house trimer spike IgA), 90%/99.3% (in-house trimer spike IgG), 89%/98.3% (in-house nucleocapsid IgG), 73.7%/100% (EDI nucleocapsid IgM), 84.5%/95.1% (EDI nucleocapsid IgG), 95%/93.7% (Euroimmun S1 IgA), 82.8%/99.7% (Euroimmun S1 IgG), 82.0%/91.7% (Chembio nucleocapsid IgM), 92%/93.3% (Chembio nucleocapsid IgG). The presumed causes of false positive results from pre-epidemic samples in commercial and in-house assays were mixed. In some cases, assays lacked reproducibility. In other cases, reactivity was abrogated by competitive inhibition (spiking the sample with the same antigen that was used for coating ELISAs prior to performing the assay), suggesting positive reaction could be attributed to the presence of antibodies against these antigens. In other cases, reactivity was consistently detected but not abrogated by the spiking, suggesting positive reaction was not attributed to the presence of antibodies against these antigens. Overall, there was wide variability in assay performance using our samples, with in-house tests exhibiting the highest combined sensitivity and specificity. The causes of "false positivity" in pre-epidemic samples may be due to plasma antibodies apparently reacting with the corresponding antigen, or spurious reactivity may be directed against non-specific components in the assay system. Identification of these targets will be essential to improving assay performance.

Clinical Factors Associated with Progression and Prolonged Viral Shedding in COVID-19 Patients: A Multicenter Study.

Coronavirus disease 2019 (COVID-19) is a global pandemic associated with a high mortality. Our study aimed to determine the clinical risk factors associated with disease progression and prolonged viral shedding in patients with COVID-19. Consecutive 564 hospitalized patients with confirmed COVID-19 between January 17, 2020 and February 28, 2020 were included in this multicenter, retrospective study. The effects of clinical factors on disease progression and prolonged viral shedding were analyzed using logistic regression and Cox regression analyses. 69 patients (12.2%) developed severe or critical pneumonia, with a higher incidence in the elderly and in individuals with underlying comorbidities, fever, dyspnea, and laboratory and imaging abnormalities at admission. Multivariate logistic regression analysis indicated that older age (odds ratio [OR], 1.04; 95% confidence interval [CI], 1.02-1.06), hypertension without receiving angiotensinogen converting enzyme inhibitors or angiotensin receptor blockers (ACEI/ARB) therapy (OR, 2.29; 95% CI, 1.14-4.59), and chronic obstructive pulmonary disease (OR, 7.55; 95% CI, 2.44-23.39) were independent risk factors for progression to severe or critical pneumonia. Hypertensive patients without receiving ACEI/ARB therapy showed higher lactate dehydrogenase levels and computed tomography (CT) lung scores at about 3 days after admission than those on ACEI/ARB therapy. Multivariate Cox regression analysis revealed that male gender (hazard ratio [HR], 1.22; 95% CI, 1.02-1.46), receiving lopinavir/ritonavir treatment within 7 days from illness onset (HR, 0.75; 95% CI, 0.63-0.90), and receiving systemic glucocorticoid therapy (HR, 1.79; 95% CI, 1.46-2.21) were independent factors associated with prolonged viral shedding. Our findings presented several potential clinical factors associated with developing severe or critical pneumonia and prolonged viral shedding, which may provide a rationale for clinicians in medical resource allocation and early intervention.

A retrospective study of the initial chest CT imaging findings in 50 COVID-19 patients stratified by gender and age.

OBJECTIVE: To retrospectively analyze and stratify the initial clinical features and chest CT imaging findings of patients with COVID-19 by gender and age. METHODS: Data of 50 COVID-19 patients were collected in two hospitals. The clinical manifestations, laboratory examination and chest CT imaging features were analyzed, and a stratification analysis was performed according to gender and age [younger group: <50 years old, elderly group ≥50 years old]. RESULTS: Most patients had a history of epidemic exposure within 2 weeks (96%). The main clinical complaints are fever (54%) and cough (46%). In chest CT images, ground-glass opacity (GGO) is the most common feature (37/38, 97%) in abnormal CT findings, with the remaining 12 patients (12/50, 24%) presenting normal CT images. Other concomitant abnormalities include dilatation of vessels in lesion (76%), interlobular thickening (47%), adjacent pleural thickening (37%), focal consolidation (26%), nodules (16%) and honeycomb pattern (13%). The lesions were distributed in the periphery (50%) or mixed (50%). Subgroup analysis showed that there was no difference in the gender distribution of all the clinical and imaging features. Laboratory findings, interlobular thickening, honeycomb pattern and nodules demonstrated remarkable difference between younger group and elderly group. The average CT score for pulmonary involvement degree was 5.0±4.7. Correlation analysis revealed that CT score was significantly correlated with age, body temperature and days from illness onset (p < 0.05). CONCLUSIONS: COVID-19 has various clinical and imaging appearances. However, it has certain characteristics that can be stratified. CT plays an important role in disease diagnosis and early intervention.

Specificity and Performance of Nucleocapsid and Spike-based SARS-CoV-2 Serologic Assays.

There is an urgent need for an accurate antibody test for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). In this paper, we have developed 3 ELISA methods, trimer spike IgA, trimer spike IgG, and nucleocapsid IgG, for detecting anti-SARS-CoV-2 antibodies. We evaluated their performance in comparison with four commercial ELISAs, EDI Novel Coronavirus COVID-19 ELISA IgG and IgM, Euroimmun Anti-SARS-CoV-2 ELISA IgG and IgA, and one lateral flow assay, DPP COVID-19 IgM/IgG System (Chembio). Both sensitivity and specificity were evaluated and the causes of false-positive reactions were determined. The assays were compared using 300 pre-epidemic samples and 100 PCR-confirmed COVID-19 samples. The sensitivities and specificities of the assays were as follows: 90%/100% (in-house trimer spike IgA), 90%/99.3% (in-house trimer spike IgG), 89%/98.3% (in-house nucleocapsid IgG), 73.7%/100% (EDI nucleocapsid IgM), 84.5%/95.1% (EDI nucleocapsid IgG), 95%/93.7% (Euroimmun S1 IgA), 82.8%/99.7% (Euroimmun S1 IgG), 82.0%/91.7% (Chembio nucleocapsid IgM), 92%/93.3% (Chembio nucleocapsid IgG). The presumed causes of positive signals from pre-epidemic samples in commercial and in-house assays were mixed. In some cases, positivity varied with assay repetition. In other cases, reactivity was abrogated by competitive inhibition (spiking the sample with analyte prior to performing the assay). In other cases, reactivity was consistently detected but not abrogated by analyte spiking. Overall, there was wide variability in assay performance using our samples, with in-house tests exhibiting the highest combined sensitivity and specificity. The causes of false positivity in pre-epidemic samples may be due to plasma antibodies apparently reacting with the analyte, or spurious reactivity may be directed against non-specific components in the assay system. Identification of these targets will be essential to improving assay performance.