Difference between revisions of "Covid19"

From Knoesis wiki
Jump to: navigation, search
(COVID19: Pyschological Impact)
(COVID19: Pyschological Impact)
Line 1: Line 1:
==COVID19: Pyschological Impact==
+
==COVID-19: Public Health Study: Semantic Analysis of Social Media and New Big Data to understanding COVID-19's impact on mental health, addiction and gender-based violence==
  
 
''' Motivation: '''
 
''' Motivation: '''

Revision as of 01:31, 14 August 2020

COVID-19: Public Health Study: Semantic Analysis of Social Media and New Big Data to understanding COVID-19's impact on mental health, addiction and gender-based violence

Motivation:

Experts have warned about the potential rapid growth in several social and health consequences of COVID-19 on individuals and society, specifically Mental Health (Depression, Anxiety), Addiction (Substance-use), and Gender-based (or Domestic) Violence (GBV). We have been successfully utilizing social media measures for epidemiology and public health research, such as Drug abuse (leading to FDA warning), Mental health, harassment, and GBV. For COVID-19 we are exploring the following questions:

  1. Q1: How can we use social media to measure psychological and social impact in (near) real-time?
  2. Q2: Specifically, how does intervention in the form of state-level policy choices and implementations relate to mental health and addiction-related behaviors across different states? What evidence is there for adaptive/coping behavior?
  3. Q3: How do GenZ and Millennials express themselves in the outbreak, particularly in the context of Mental health?

Our approach for processing big social media data involves a series of state-of-the-art AI techniques utilizing human-curated knowledge bases, data mining, and semantic filtering procedures (see Technical Approach for details). We have collected >800 Million tweets from March 14 to April 10, 2020, and subsequently selecting 45 M Tweets with location (explicit location), 27 M Tweets of those with matches to entities in News, finally obtaining 15 M Tweets with exact matches to concepts in our human-curated Mental Health and Drug Abuse computationally accessible Knowledge Base (MHDA-Kb). Figure 1 illustrates the links between the words in tweets and their mental health interpretation in this knowledge base. The interpretation of social media data is assisted through ~700K COVID-related news articles (January 01 to March 29, 2020).

Examples tweets include that motivated our questions:

  1. "You believe I have any pleasure in this chaos? Jeez. I’ve been despairing for 2 months."
  2. "A feeling of hopelessness. Seems I am in a dark age. #coronavirus #COVID19".
  3. “self-isolated for two weeks and depression becoming unbearable. This coronavirus is worsening my anxiety a lot and I am terrified.”
  4. “side effects of hydroxychloroquine: "Mental/mood changes (such as confusion, personality changes, unusual thoughts/behavior, depression, feeling being watched, hallucinating"
  5. “The reason I’m on #Hxychloroquine sedative is that I actually have anxiety in my sleep. I don’t know why I’m having nocturnal anxiety attacks. It’s fucking awful. I’ll wake up and not even know where I am because I’m so scared or shook”

Social Quality Index (SQI): A Social Quality Index (SQI) is calculated from the aggregation of mental health and addiction components. Raw SQI takes into account tweet concepts abstracted through three different mental health lenses in the MHDA-Kb: Depression, Anxiety, and Drug Abuse Disorders. Raw SQI simply aggregates the relevant features with respect to each of these lenses in each message, and does not take into account preceding state conditions. Change in SQI is also potentially informative, particularly for comparisons between states. We transformed raw state SQI into a relative state ranking, to capture drifts between worsening and improving psychological conditions in social quality. SQI ranking is also used to examine the effect of external factors, such as school closure, business closure, unemployment, and lockdown (including the extension of lockdown).

News Coverage: [1]The Conversation


COVID19: Epidemiology Study with Exogenous Factors

[2]Study accepted at ACM KDD'20 AI for COVID Track

Collaborators: Nirmal Sivaraman, Dr. Sakthi Balan

Epidemiological models are the mathematical models that capture the dynamics of epidemics. The spread of the virus has two routes - exogenous and endogenous. The exogenous spread is from outside the population under study, and endogenous spread is within the population under study. Although some of the models consider the exogenous source of infection, they have not studied the interplay between exogenous and endogenous spreads. In this paper, we introduce a novel model - the Exo-SIR model that captures both the exogenous and endogenous spread of the virus. We analyze to find out the relationship between endogenous and exogenous infections during the Covid19 pandemic. First, we simulate the Exo-SIR model without assuming any contact network for the population. Second, simulate it by assuming that the contact network is a scale-free network. Third, we implemented the Exo-SIR model on a real dataset regarding Covid19. We found that endogenous infection is influenced by even a minimal rate of exogenous infection. Also, we found that in the presence of exogenous infection, the endogenous infection peak becomes higher, and the peak occurs earlier. This means that if we consider our response to a pandemic like Covid19, we should be prepared for an earlier and higher number of cases than the SIR model suggests if there are the exogenous source(s) of infection.