A multistage multimodal deep learning model for disease severity assessment and early warnings of high-risk patients of COVID-19

The outbreak of coronavirus disease 2019 (COVID-19) has caused massive infections and large death tolls worldwide. Despite many studies on the clinical characteristics and the treatment plans of COVID-19, they rarely conduct in-depth prognostic research on leveraging consecutive rounds of multimodal clinical examination and laboratory test data to facilitate clinical decision-making for the treatment of COVID-19. To address this issue, we propose a multistage multimodal deep learning (MMDL) model to (1) first assess the patient's current condition (i.e., the mild and severe symptoms), then (2) give early warnings to patients with mild symptoms who are at high risk to develop severe illness. In MMDL, we build a sequential stage-wise learning architecture whose design philosophy embodies the model's predicted outcome and does not only depend on the current situation but also the history. Concretely, we meticulously combine the latest round of multimodal clinical data and the decayed past information to make assessments and predictions. In each round (stage), we design a two-layer multimodal feature extractor to extract the latent feature representation across different modalities of clinical data, including patient demographics, clinical manifestation, and 11 modalities of laboratory test results. We conduct experiments on a clinical dataset consisting of 216 COVID-19 patients that have passed the ethical review of the medical ethics committee. Experimental results validate our assumption that sequential stage-wise learning outperforms single-stage learning, but history long ago has little influence on the learning outcome. Also, comparison tests show the advantage of multimodal learning. MMDL with multimodal inputs can beat any reduced model with single-modal inputs only. In addition, we have deployed the prototype of MMDL in a hospital for clinical comparison tests and to assist doctors in clinical diagnosis.

Are female-specific cancers long-term sequelae of COVID-19? Evidence from a large-scale genome-wide cross-trait analysis (preprint)

ABSTRACT Background Little is known regarding the long-term adverse effects of COVID-19 on female-specific cancers due to the restricted length of observational time, nor the shared genetic influences underlying these conditions. Methods Leveraging summary statistics from the hitherto largest genome-wide association studies conducted in each trait, we performed a comprehensive genome-wide cross-trait analysis to investigate the shared genetic architecture and the putative genetic associations between COVID-19 with three main female-specific cancers: breast cancer (BC), epithelial ovarian cancer (EOC), and endometrial cancer (EC). Three phenotypes were selected to represent COVID-19 susceptibility (SARS-CoV-2 infection) and severity (COVID-19 hospitalization, COVID-19 critical illness). Results For COVID-19 susceptibility, we found no evidence of a genetic correlation with any of the female-specific cancers. For COVID-19 severity, we identified a significant genome-wide genetic correlation with EC for both hospitalization ( r g =0.19, P =0.01) and critical illness ( r g =0.29, P =3.00×10 −4 ). Mendelian randomization demonstrated no valid association of COVID-19 with any cancer of interest, except for suggestive associations of genetically predicted hospitalization (OR IVW =1.09, 95%CI=1.01-1.18, P =0.04) and critical illness (OR IVW =1.06, 95%CI=1.00-1.11, P =0.04) with EC risk, none withstanding multiple correction. No reverse association was found. Cross-trait meta-analysis identified multiple pleiotropic SNPs between COVID-19 and female-specific cancers, including 20 for BC, 15 for EOC, and 5 for EC. Transcriptome-wide association studies revealed shared genes, mostly enriched in the hematologic, cardiovascular, and nervous systems. Conclusions Our genetic analysis highlights an intrinsic link underlying female-specific cancers and COVID-19 - while COVID-19 is not likely to elevate the immediate risk of the examined female-specific cancers, it appears to share mechanistic pathways with these conditions. These findings may provide implications for future therapeutic strategies and public health actions.

Regional coal power overcapacity assessment in China from 2020 to 2025

Abstact Since the beginning of 2020, the weak domestic power demand caused by the COVID-19 pandemic and the large-scale advancement of coal power construction projects may have further aggravated the coal power overcapacity in China. Given the new situation, this study collected the data on China’s installed electricity capacity and electricity demand during the 13th Five-Year Plan period. Moreover, a reasonable capacity evaluation model of coal power was established based on the energy and electric power balance to analyze China’s coal power overcapacity in 2019 and determine the reasonable capacity for 2025. The internal reasons of unreasonable energy structure and regional difference of energy structure are systematically discussed. Results show that the overcapacity in 2019 was approximately 170 GW, and the overcapacity situation in North, Northwest, and South China was particularly serious. The reasonable coal power capacity for 2025 under the basic situation is 950 GW, indicating that if all the coal power units under construction and planning are operated, the overcapacity in 2025 will be 300 GW. Sensitivity and comprehensive scenario analyses show that under different scenarios, the upper and lower limits of the reasonable coal power capacity in 2025 are 1083 and 794 GW, respectively. Finally, this paper proposes relevant policy recommendations to cope with China’s serious coal power overcapacity problem.

Impacts of COVID-19 on Myanmar's chicken and egg sector, with implications for the sustainable development goals

CONTEXT Chicken and eggs make important contributions to the food and nutrition security of low-income households in Myanmar, making it important to understand the impacts of COVID-19 on Myanmar's poultry sector. OBJECTIVE First, we evaluate the responsiveness and resilience of different chicken and egg farming systems in Myanmar to the shock of COVID-19. Second, we evaluate implications of the performance of the chicken and egg sector during COVID-19 for the Sustainable Development Goals. METHODS We conducted six waves of telephone interviews from June to November 2020 with 269 chicken farms close to Yangon. We compared impacts in two types of production system - broilers and layers - using a survey of the same farms conducted in 2019 as a baseline. For each type of farm, we compared ‘integrated’ and ‘non-integrated’ farms, where integration involves combining production of chickens and fish. RESULTS AND CONCLUSIONS First, the COVID-19 pandemic severely impacted chicken and egg production. More than 30% of broiler farms and 10% of layer farms closed before June, 42% of long-term farmworkers were laid off, and indicators of business sentiment were much more pessimistic than in 2019. Second, the sector experienced a V-shaped recovery until September 2020 when a second wave of COVID-19 hit Myanmar. Third, the impacts of COVID-19 vary by production system. Broiler farms have a much shorter production cycle than layer farms and were able rapidly adjust operational status by closing or reopening, whereas very few layer farms reopened after closing. Fourth, integrated layer-fish farms proved more resilient to the shock of COVID-19 than layer farms, with 90% of layer-fish farms and 76% of layer farms remaining operational in November, but there was no difference in the performance of broiler-fish and broiler farms. Fifth, the slow supply response of layer farms has meant higher egg prices for consumers, likely affecting nutritional intakes and making it more difficult for Myanmar to achieve the second Sustainable Development Goal of ending hunger and malnutrition by 2030. SIGNIFICANCE The results contribute to understanding of the challenges faced by chicken farms in Myanmar during the COVID-19 pandemic, and the effectiveness of their adaptive responses. Results have implications for other countries in Asia where integrated livestock-fish farms are common, and other developing countries where the poultry sector is expanding rapidly.

EPAC regulates von Willebrand factor secretion from endothelial cells in a PI3K/eNOS-dependent manner during inflammation (preprint)

Coagulopathy is associated with both inflammation and infection, including infection with the novel SARS-CoV-2 (COVID-19). Endothelial cells (ECs) fine tune hemostasis via cAMP-mediated secretion of von Willebrand factor (vWF), which promote the process of clot formation. The e xchange p rotein directly a ctivated by c AMP (EPAC) is a ubiquitously expressed intracellular cAMP receptor that plays a key role in stabilizing ECs and suppressing inflammation. To assess whether EPAC could regulate vWF release during inflammation, we utilized our EPAC1 -null mouse model and revealed an increased secretion of vWF in endotoxemic mice in the absence of the EPAC1 gene. Pharmacological inhibition of EPAC1 in vitro mimicked the EPAC1 −/− phenotype. EPAC1 regulated TNFα-triggered vWF secretion from human umbilical vein endothelial cells (HUVECs) in a phosphoinositide 3-kinases (PI3K)/endothelial nitric oxide synthase (eNOS)-dependent manner. Furthermore, EPAC1 activation reduced inflammation-triggered vWF release, both in vivo and in vitro . Our data delineate a novel regulatory role of EPAC1 in vWF secretion and shed light on potential development of new strategies to controlling thrombosis during inflammation. Key Point PI3K/eNOS pathway-mediated, inflammation-triggered vWF secretion is the target of the pharmacological manipulation of the cAMP-EPAC system.

Clinical characteristics of COVID-19 and the model for predicting the occurrence of critically ill patients: a retrospective cohort study. (preprint)

Background: The present study aim to comprehensively report the epidemiological and clinical characteristics of the COVID-19 patients and to develop a multi-feature fusion model for predicting the critical ill probability. Methods: It was a retrospective cohort study that incorporating the laboratory-confirmed COVID-19 patients in the Chongqing Public Health Medical Center. The prediction model was constructed with least absolute shrinkage and selection operator (LASSO) logistic regression method and the model was further tested in the validation cohort. The performance was evaluated by the receiver operating curve (ROC), calibration curve and decision curve analysis (DCA). Results: A total of 217 patients were included in the study. During the treatment, 34 patients were admitted to intensive care unit (ICU) and no developed death. A model incorporating the demographic and clinical characteristics, imaging features and laboratory findings were constructed to predict the critical ill probability and it was proved to have good calibration, discrimination ability and clinic use. Conclusions: The prevalence of critical ill was relatively high and the model may help the clinicians to identify the patients with high risk for developing the critical ill, thus to conduct timely and targeted treatment to reduce the mortality rate.

An artificial intelligence system for predicting the deterioration of COVID-19 patients in the emergency department (preprint)

During the coronavirus disease 2019 (COVID-19) pandemic, rapid and accurate triage of patients at the emergency department is critical to inform decision-making. We propose a data-driven approach for automatic prediction of deterioration risk using a deep neural network that learns from chest X-ray images and a gradient boosting model that learns from routine clinical variables. Our AI prognosis system, trained using data from 3,661 patients, achieves an area under the receiver operating characteristic curve (AUC) of 0.786 (95% CI: 0.745-0.830) when predicting deterioration within 96 hours. The deep neural network extracts informative areas of chest X-ray images to assist clinicians in interpreting the predictions and performs comparably to two radiologists in a reader study. In order to verify performance in a real clinical setting, we silently deployed a preliminary version of the deep neural network at New York University Langone Health during the first wave of the pandemic, which produced accurate predictions in real-time. In summary, our findings demonstrate the potential of the proposed system for assisting front-line physicians in the triage of COVID-19 patients.