Predicting Emerging Themes in Rapidly Expanding COVID-19 Literature with Dynamic Word Embedding Networks and Machine Learning

Evidence from peer-reviewed literature is the cornerstone for designing responses to global threats such as COVID-19. The collection of knowledge and interpretation in publications needs to be distilled into evidence by leveraging natural language in ways beyond standard meta-analysis. Several studies have focused on mining evidence from text using natural language processing, and have focused on a handful of diseases. Here we show that new knowledge can be captured, tracked and predicted using the evolution of unsupervised word embeddings and machine learning. Our approach to decipher the flow of latent knowledge in time-varying networks of word-vectors captured thromboembolic complications as an emerging theme in more than 77,000 peer-reviewed publications and more than 11,000 WHO vetted preprints on COVID-19. Furthermore, machine learning based prediction of emerging links in the networks reveals autoimmune diseases, multisystem inflammatory syndrome and neurological complications as a dominant research theme in COVID-19 publications starting March 2021.

Explainable-Machine-Learning to discover drivers and to predict mental illness during COVID-19

ImportanceCOVID-19 pandemic has deeply affected the health, economic, and social fabric of nations. Identification of individual-level susceptibility factors may help people in identifying and managing their emotional, psychological, and social well-being. ObjectiveThis work is focused on learning a ranked list of factors that could indicate a predisposition to a mental disorder during the COVID pandemic. Data Sources and Study designIn this study, We have used a survey of 17764 adults in the USA at different age groups, genders, and socioeconomic statuses. MethodsThrough initial statistical analysis followed by Bayesian Network inference, we have identified key factors affecting Mental health during the COVID pandemic. Integrating Bayesian networks with classical machine learning approaches lead to effective modeling of the level of mental health. ResultsOverall, females are more stressed than males, and people of age-group 18-29 are more vulnerable to anxiety than other age groups. Using the Bayesian Network Model, we found that people with chronic medical condition of mental illness are more prone to mental disorders during the COVID age. The new realities of working from home, home-schooling, and lack of communication with family/friends/neighbors induces mental pressure. Financial assistance from social security helps in reducing mental stress during COVID generated economic crises. Finally, using supervised ML models, we predicted the most mentally vulnerable people with ~80% accuracy. QuestionWhat factors could affect mental health, and how could this level be predicted using the same factors during the of COVID-19? FindingIn this survey-data based study, multiple factors such as Social isolation, digital communication, working, and schooling from home, were identified as crucial factors of mental illness during Covid-19. Interestingly, behavioral changes such as wearing a mask and avoiding restaurants and public places were not found to be associated with mental health. MeaningRegular non-virtual communication with friends and family, healthy social life and social security are key factors and especially taking care of people with mental disease history is even more important.

Temperature and Humidity Do Not Influence Global COVID-19 Incidence as Inferred from Causal Models

The relationship between meteorological factors such as temperature and humidity with COVID-19 incidence is still unclear after 6 months of the beginning of the pandemic. Some literature confirms the association of temperature with disease transmission while some oppose the same. This work intends to determine whether there is a causal association between temperature, humidity and Covid-19 cases. Three different causal models were used to capture stochastic, chaotic and symbolic natured time-series data and to provide a robust & unbiased analysis by constructing networks of causal relationships between the variables. Granger-Causality method, Transfer Entropy method & Convergent Cross-Mapping (CCM) was done on data from regions with different temperatures and cases greater than 50,000 as of 13th May 2020. From the Granger-Causality test we found that in only Canada, the United Kingdom, temperature and daily new infections are causally linked. The same results were obtained from Convergent Cross Mapping for India. Again using Granger-Causality test, we found that in Russia only, relative humidity is causally linked to daily new cases. Thus, a Generalized Additive Model with a smoothing spline function was fitted for these countries to understand the directionality. Using the combined results of the said models, we were able to conclude that there is no evidence of a causal association between temperature, humidity and Covid-19 cases.

A Counterfactual Graphical Model Reveals Economic and Sociodemographic Variables as Key Determinants of Country-Wise COVID-19 Burden

ImportanceInsights into the country-wise differences in COVID-19 burden can impact the policies being developed to control disease spread. ObjectivePresent study evaluated the possible socio-economic and health related factors (and their temporal consistency) determining the disease burden of COVID-19. DesignA retrospective analysis for identifying associations of COVID-19 burden. SettingData on COVID-19 statistics (number of cases, tests and deaths per million) was extracted from the website https://www.worldometers.info/coronavirus/ on 10th April and 12th May. Variables obtained to estimate the possible determinants for COVID-19 burden included economic-gross domestic product; socio-demographic-Sustainable Development Goals, SDGs indicators related to health systems, percentage Chinese diaspora; and COVID-19 trajectory-date of first case in each country, days between first reported case and 10th April, days between 100th and 1000th case, and government response stringency index (GRSI). Main outcomes and MeasuresCOVID-19 burden was modeled using economic and socio-demographic determinants. Consistency of inferences for two time points at three levels of increasing statistical rigor using (i) Spearman correlations, (ii) Bayesian probabilistic graphical model, and (iii) counterfactual impact was evaluated. ResultsCountries economy (reflected by GDP), mainly through the testing rates, was the major and temporally consistent determinant of COVID-19 burden in the model. Reproduction number of COVID-19 was lower where mortality due to water, sanitation, and hygiene (WaSH) was higher, thus strengthening the hygiene hypothesis. There was no association between vaccination status or tuberculosis incidence and COVID burden, refuting the claims over BCG vaccination as a possible factor against COVID-19 trajectory. Conclusion and RelevanceCountries economy, through testing power, was the major determinant of COVID-19 burden. There was weak evidence for hygiene hypothesis as a protective factor against COVID-19.

Less Wrong COVID-19 Projections With Interactive Assumptions

COVID-19 pandemic is an enigma with uncertainty caused by biological and health systems factors. Although many models have been developed all around the world, transparent models that allow interacting with the assumptions will become more important as we test various strategies for lockdown, testing and social interventions and enable effective policy decisions. In this paper we developed a suite of models to guide development of policies under different scenarios when the lockdown opens. These had been deployed to create an interactive dashboard called COVision which includes the Agent based Models (ABM) and classical compartmental models i.e. Susceptible-Infected-Recovered (SIR) and Susceptible-Exposed-Infected-Recovered (SEIR) approaches. Our tool allows simulation of scenarios by changing strength of lockdown, basic reproduction number(R0), asymptomatic spread, testing rate, contact rate (Beta), recovery rate (Gamma), incubation period and starting number of cases. We optimized ABMs and classical compartmental models to fit the actual data, both of which performed well in terms of R-squared, root mean squared error (RMSE) and mean absolute percentage error (MAPE). Out of the three models in our suite, ABM was able to capture the data better than SIR and SEIR and achieved an RSQ of 92.3% for India and 89% for Maharashtra for the next 30 days. We also computed R0 using SIR and SEIR models which were found to be decreasing over the different periods of lockdown indicating the effectiveness of policies and interventions. Finally, we formulated ICU bed requirements using our best models. Our evaluation suggests that ABM models were able to capture the dynamic nature of the epidemic for a longer duration of time while classical SIR and SEIR models performed inefficiently for longer terms. The visual interactivity and ability to simulate outcomes under different parameters will allow the policymakers to make informed decisions for estimating the strength of lockdown to be implemented and testing rates. Further, our models were able to highlight the differences at state level for the parameters such as R0 and contact rates and hence can be applied for state specific decision making. An interactive dashboard http://covision.tavlab.iiitd.edu.in have been hosted as a web-server for the war level monitoring of the covid19 pandemic in India in public domain

CovidNLP: A Web Application for Distilling Systemic Implications of COVID-19 Pandemic with Natural Language Processing

The flood of conflicting COVID-19 research has revealed that COVID-19 continues to be an enigma. Although more than 14,000 research articles on COVID-19 have been published with the disease taking a pandemic proportion, clinicians and researchers are struggling to distill knowledge for furthering clinical management and research. In this study, we address this gap for a targeted user group, i.e. clinicians, researchers, and policymakers by applying natural language processing to develop a CovidNLP dashboard in order to speed up knowledge discovery. The WHO has created a repository of about more than 5000 peer-reviewed and curated research articles on varied aspects including epidemiology, clinical features, diagnosis, treatment, social factors, and economics. We summarised all the articles in the WHO Database through an extractive summarizer followed by an exploration of the feature space using word embeddings which were then used to visualize the summarized associations of COVID-19 as found in the text. Clinicians, researchers, and policymakers will not only discover the direct effects of COVID-19 but also the systematic implications such as the anticipated rise in TB and cancer mortality due to the non-availability of drugs during the export lockdown as highlighted by our models. These demonstrate the utility of mining massive literature with natural language processing for rapid distillation and knowledge updates. This can help the users understand, synthesize, and take pre-emptive action with the available peer-reviewed evidence on COVID-19. Our models will be continuously updated with new literature and we have made our resource CovidNLP publicly available in a user-friendly fashion at http://covidnlp.tavlab.iiitd.edu.in/. Data Availability StatementAll the data used in this study are publicly available from the WHO Covid-19 Global Literature on coronavirus disease maintained at https://search.bvsalud.org/global-literature-on-novel-coronavirus-2019-ncov/. Our analysis and the interactive resource CovidNLP is publicly available in a user friendly fashion at http://covidnlp.tavlab.iiitd.edu.in