Am I Being Bullied on Social Media? An Ensemble Approach to Categorize Cyberbullying

People can easily reveal their aggressive remarks on social media platforms using the anonymity it provides. During the COVID-19 pandemic, the usage of social media has been increased several times according to surveys and people are vulnerable to cyber attacks now more than ever. Prevention of cyberbullying needs careful monitoring and identification. Most of the existing works on cyberbullying detection employed traditional machine learning classifiers with handcrafted fea-tures, and deep learning-based models have made their way in this domain very recently. Categorizing cyberbullying based on traits is a complex task and needs contextual consideration. In this work, we have proposed a new approach to detect cyberbullying on social media platforms using a neural ensemble method of transformer-based architectures with attention mechanism. Our proposed architecture is trained on one balanced and one imbalanced dataset and outperforms the given ML and DNN baselines by a significant margin in both cases. We achieved an average F1-score of 95.59% for five classes and 90.65% for six classes on the Fine-Grained Cyberbullying Dataset (FGCD), and 87.28% on Twitter parsed dataset. Our in-depth results provide great insights into the effectiveness of transformer-based models in cyberbullying detection and paves the way for future researches to combat this serious online issue. We have released our models and code.1 © 2021 IEEE.

Knowledge and Readiness of COVID-19 and MCO from the Perspective of Safety and Health Competent Persons and Training Providers in Malaysia

The effects and aftermath of the COVID-19 pandemic and ensuing Movement Control Order (MCO) will be felt for a long while, even after the MCO period ends. It is imperative to determine and measure the level of knowledge and readiness among Safety and Health Competent Persons (SHCPs) and Training Providers (TPs) in handling issues relating to COVID-19 during the MCO in Malaysia. A survey instrument was developed and validated by expert panels from academia and enforcement agency. A pilot study involving 40 respondents showed high reliability, with a Cronbach's alpha score of 0.95 and 0.96 for SHCP and TP. A fullblown research was carried out for two weeks from the 14th to 28th of April 2020. This cross-sectional study covers all states in Malaysia and has been distributed to approximately 3,000 respondents with a response rate of 10.6 % from respondents employed in various sectors. The construction sectors recorded the highest percentage of responses as compared to other sectors at 37 % followed by manufacturing sectors at 23 % and public services and statutory authorities at 18 %. The companies are a mix of small, medium-sized enterprises (47 %), and large establishments (53 %), which altogether have a wide range of turnover rates. This study highlights that SHCPs and TPs in Malaysia have moderate post-MCO awareness and readiness in terms of knowledge, skills, standard operating procedure, social distancing, and basic equipment to control and curb workplace contagion after the MCO is lifted. There are uncertainties regarding the TP's willingness to budget for and bear the daily costs of the decontamination procedure, personal protective equipment usage, space availability, and the COVID-19 screening process. The readiness of SHCPs and TPs have been measured and identified. SHCPs and TPs need more knowledge in post-MCO COVID-19 management, especially in terms of the capacity of the trained staff to adequately screen workers for COVID-19 symptoms after the MCO is lifted. SHCPs and TPs believe they are underprepared, but they are willing to learn more about pandemic preparedness if training options are provided. A plan of action to address these issues has been proposed to DOSH Malaysia for guidance on formulating the next step forward. © 2021, AIDIC Servizi S.r.l.

Household Materials for Homemade Masks: How Effective Are They?

Introduction: With the fear of uncertain behaviours and mechanisms of the coronavirus in the rapidly evolving COVID-19 pandemic, people are required to cover their mouth and nose to prevent the spread of the virus. This has become a challenge as most countries struggle with the dwindling stocks of face masks. However, the Centers for Disease Control and Prevention suggested the use of cloth masks as a substitute. In an effort to find suitable materials that serve this purpose and ensure sufficient protection afforded, the masks need to be tested. Methods: In this study, the penetration and breathing resistance of 13 easily obtained household materials potentially used by the public as face masks, were randomly selected and tested by using TSI Model 8130 Automated Filter Tester, while the thickness was measured by S-Cal EVO Proximity caliper. Results: The level of thickness ranged between 0.25 mm (satin) and 4.83 mm (diaper). Double-layer denim material showed the lowest percentage of penetration (Median, IQR: 27.50%, 27.05-28.80) while the greatest was single-layer lycra (Median, IQR: 90.60%, 80.80-92.10). Single-layer chiffon fabric showed the best breathing resistance (Median, IQR in mmH2O: 1.30, 0.90, 1.45) while the worst was diaper (Median, IQR in mmH2O: 87.20, 86.95, 87.25). Conclusion: Double-layer dried wet tissue, single-layer tissue paper and double-layer non-woven fabric can be considered to be used as face masks due to their efficiency based on penetration and breathing resistance factors. © 2021 UPM Press. All rights reserved.

Hospitalized COVID-19 patients with active lymphoma have eight times higher risk of death than COVID-19 controls

BACKGROUND: Over 790,000 patients in the United States are currently living with or are in remission from lymphoma. It is established that lymphoma patients are at greater risk for both bacterial and viral infections. While there is limited research examining the risk of COVID-19 infection in patients with an active malignancy, even fewer studies have examined those with active lymphoma. This study aimed to examine the all-cause mortality of COVID-19 patients with active lymphoma compared to hospitalized COVID-19 control patients. METHODS: We performed a retrospective case-control and cohort study of adult inpatients diagnosed with COVID-19 infection in a tertiary, academic referral center in Richmond, Virginia. We analyzed the unadjusted and adjusted association of patients with active lymphoma diagnosis and all-cause hospital mortality. We performed multiple logistic regressions adjusting for age, gender, race, the month at presentation, which captures the health system's adaptation, and the remaining 30 individual diagnostic categories of the Elixhauser comorbidity index. We externally validated our findings using compiled data from 657 institutions across the United States on patients with lymphoma hospitalized for COVID-19. RESULTS: Among 628 inpatients with COVID-19, 1.1% (7) had active lymphoma. The unadjusted mortality of patients with lymphoma was 57.1% compared to 8.4% of the corresponding patients without lymphoma. The unadjusted OR for hospital death was 15.6 (95% CI 3.2 to 67, P=0.001). The adjusted OR of death in patients with lymphoma was 79.5 (95% 6.4 to 983, P= 0.001). The average adjusted mortality in patients with lymphoma was 65% compared with 8.4% among patients of equivalent age, gender, race, month of presentation and comorbidities. From aggregate data of COVID-19 patients across 657 US institutions, the average mortality for patients with lymphoma was 41.07% (95% CI 36.8 to 45.3) and for patients without lymphoma was 12.11% (95% CI 12.7 to 11.5). CONCLUSION: Our results show that, of those patients hospitalized for COVID-19 infection, the patients with active lymphoma have a nearly 8-fold increased risk of death compared to their non-lymphoma counterparts when adjusted for age, gender, race, month of presentation, and other comorbidities. External validation data demonstrated a greater than 3-fold increased risk of death in COVID-19 patients with active lymphoma compared to non-lymphoma patients. This research highlights the importance of mitigation strategies, such as social distancing and masking, to decrease the risk of COVID-19 infection in lymphoma patients and may have implications for prioritizing vaccines or therapies in the future. FIGURE:.

Hypothyroidism associated with triple mortality in patients hospitalized with COVID-19

BACKGROUND: Many conditions have been associated with severe COVID-19 disease. To date, the risk associated with pre-existing hypothyroidism remains unclear. Hypothyroidism affects the innate immune system. Patients with hypothyroidism have higher circulating inflammatory markers, which are associated with increased mortality in COVID-19. A prior study did not find a significant difference in the risk of hospitalization or death in patients with pre-existing hypothyroidism. This study aims to investigate a possible association between pre-existing hypothyroidism and death from COVID-19. METHODS: We performed a retrospective cohort study of adult inpatients diagnosed with SARS-CoV-2 infection in a tertiary, academic referral center in Richmond, Virginia. We analyzed the unadjusted and adjusted association of patients with a past medical history of hypothyroidism and all-cause hospital mortality. We performed adjusted logistic regressions adjusting for age, gender, race, the month at presentation (an adaptation of the health system), and the remaining 30 individual diagnostic categories of the Elixhauser comorbidity index. RESULTS: Fifty-three (8.2%) of the 649 COVID-19 inpatients had hypothyroidism. Patients with hypothyroidism were, on average, 15.3 years older (95% CI 10.3 to 20.4 years). The unadjusted mortality of patients with hypothyroidism was 22.6% compared with 7.4% in patients without hypothyroidism. The unadjusted mortality OR was 3.5 (95% CI 1.7 to 7.2, P=0.001). The adjusted OR for death was 3.6 (95% CI 1.4 to 9.3, P=0.007, abstract figure). The average adjusted mortality was 18.6% for patients with hypothyroidism compared with 7.8% in patients with equivalent age, gender, race, remaining comorbidities, and month of presentation. CONCLUSION: Our results suggest that pre-existing hypothyroidism is associated with a three-fold risk of death in patients hospitalized with COVID-19. There are conflicting reports in the literature on the association between hypothyroidism and severe COVID-19. Earlier descriptive studies did not report rates of thyroid disease in their cohorts. Further research is needed on the pathophysiology and effects of SARS-CoV-2 infection in hypothyroid individuals.

Comparison of influenza and COVID-19 disease burden among black patients at a large urban care center

RATIONALE: Studies have demonstrated racial disparities in COVID-19 outcomes, with black Americans having higher rates of infection, hospitalization, and death. Similarly, CDC data has shown higher Influenza-related mortality in the black American population. While COVID-19 is a deadlier viral respiratory illness, a comparison between influenza and COVID-19 can provide insight into racial disparities and clarify if there is excess disease burden of COVID-19 on black American communities compared with another viral pneumonia. METHODS: We performed a four-year retrospective cohort study (2016-2020) of adult inpatients tested with SARS-CoV-2 or Influenza (A or B) infection in a tertiary, academic referral center in Richmond, Virginia. We compared the unadjusted and adjusted positivity rate, and mortality between black and non-black patients. We performed multiple logistic regression to adjust for age and gender and applied the models to estimate and compare the predicted adjusted mortality. RESULTS: The proportion of black patients admitted for Influenza from 2016-2020 was significantly greater than the proportion of black patients admitted with COVID-19, 66.6% vs. 57.5% (p <0.01). The unadjusted mortality for Influenza + patients was 1.6% (31). The unadjusted mortality for SARSCoV- 2 + patients were 5.6% (125). Black patients had lower unadjusted OR for death for influenza (0.6 95% CI 0.6-0.64, p<0.001) and OR of death for SARS-CoV-2 (0.84, 95% 0.73-0.96, p=0.01). The findings persisted after adjusting for age and gender in influenza patients (OR 0.68 95% CI 0.64-0.74, p<0.001) but not in SARSCoV- 2 patients (OR 0.91 95% CI 0.8-1.05, p=0.2). CONCLUSION: In our predominantly black American cohort, we found no significant association between race and in-hospital adjusted mortality related to COVID-19. Our findings are contrary to larger cohorts and CDC data which shows increased mortality in the black American population. The higher proportion of black patients with Influenza than COVID-19 also indicates that in our population there is not an excess burden from COVID-19 compared to previous Influenza data, although for both COVID-19 and Influenza black patients are overrepresented compared to demographics of VCU's catchment area. The reason for these findings is not clear. Our cohort was composed predominantly of black Americans as is Richmond, VA. It is possible that in this setting the provision of community health or outreach regarding COVID-19 disease prevention to black communities was more effective, reducing excess COVID-19 disease burden. Further research to identify how structural racism and social determinants of health affect vulnerable communities and factors that mitigate these effects is necessary.

Insurance disparities are associated with mortality from COVID-19 in central virginia

RATIONALE: The Center for Disease Control has reported that racial and ethnic groups are more susceptible to COVID-19 with worse outcomes. Inequalities in social determinants of health, particularly income, access to healthcare, and housing, may play a role in these differences. Health insurance has been shown to affect health outcomes for acute and chronic conditions. Uninsured and Medicaid recipients face worse outcomes for conditions like pneumonia, myocardial infarction, and lung cancer. Health insurance has also been described as a surrogate marker for social determinants of health. We set out to investigate these parameters in COVID-19 hospitalized patients in central Virginia. METHODS: We performed a retrospective cohort study of adult inpatients diagnosed with SARS-CoV-2 infection in a tertiary, academic referral center in central Virginia. We analyzed unadjusted and adjusted patient demographics like age, gender, race, ethnicity, insurance primary payer, and Elixhauser comorbidities with hospital all-cause mortality. We calculated adjusted and unadjusted mortality rates, odds ratios, and confidence intervals. We constructed a geospatial analysis of the adjusted mortality by zip code. RESULTS: Black patients constituted 56.1% of the cohort (276 patients out of 492). Hispanic patients 17.7%. Average age was 55.7 (SD 17) years. Majority of patients had Medicare (35.8%), followed by Medicaid (19.3%), no insurance specified or uninsured 15.7%, private insurance 14.8%, state/ prison insurance 11.8% and military insurance 2.6%. When adjusted for comorbidities, age, gender, race and ethnicity was not associated with mortality. Private insurance and unspecified insurance status were associated with both unadjusted lower mortality OR 0.09 (95% CI 0.014 - 0.63, p=0.01) and 0.04 (95% CI 0.004 - 0.63, p=0.01) and adjusted OR 0.07 (95% CI 0.01- 0.90, p =0.04) and 0.04 (95% CI 0.002 - 0.73, p=0.03) respectively. These findings persisted after removing the prison population (p=0.02). CONCLUSION: Despite nationwide trends indicating worse outcomes for specific racial groups affected by COVID-19, in our diverse cohort race was not associated with a significant difference in mortality. Private insurance was associated with lower mortality versus public insurance. This finding persisted when adjusting for confounders and removing the prison population.In central Virginia, health insurance status is a predictor of COVID-19 outcomes and may serve as a surrogate for health disparities. Uninsured/public insured individuals may have lower socioeconomic status or reside in medically underserved areas, limiting access to care. Further investigation is needed to elucidate the largest risk factors and design interventions to curtail the impact of COVID19 on this population.