A note of a virtual conference on AI & CoViD-19 - Section 1
About one year ago, a friend persuades me to visit HAI in a morning after my office hour of YC on a university campus.
At that moment, I started to write a book about “Augmented Intelligence” with a human-centered design, but could not gather enough primary material. Unfortunately, I could not find anyone to talk.
It is interesting to know that when I continue to write a book, their information come to me with the same topic. Seriously, I have to complete a book after a long delay.
Full Agenda: HAI Stanford
Full video: Youtube
Session I: Landscape and Framing - Started at 9:57
Challenges Responding to CoViD-19: Perspective from a Physician and Policy Maker.
- It is difficult to set an alarm for an outbreak. There is a small group of people. The previous example was Ebola.
- Lack of information about the outbreak in China.
- First action with containment strategy with travel ban.
- Lost an opportunity because of a problem with testing of CDC.
- A summary of the US situation: California, Washington and New York vs. Florida and Texas.
How to use big data and AI
- Priority about surrotic testing (blood test, chopstick), e.g. how many people in the community have potentially been infected by CoViD-19 and recovered to make decisions to open up parts of the community.
- Data with respective to hemisphere.
- Thermometer to America.
- Leadership to put the community to work together.
- Take what we are learning now to help the weaker public healthcare system like Africa.
- How to unroll shelter-in-place: Need health authority and government with science guidance.
- How to connect testing and shelter in place.
- What could policy makers do to help people get accurate information?
An Academic Medical Center’s Data Science Response to a Pandemic.
- How many patients, ICU bed, PPEs?
- Source of information: Regional (health departments, census age breakdown) and Institutional (Lab test report, ADT feed, presenting symptoms)
- What we learn: Growth rate calculation, disease burden estimation.
- Modeling effort: Test positive rate, admission rate, ICU admission rate.
- Three different models:
- Hospital bed and resources use.
- Project county hospitalization.
- Estimate case trend and policy impact.
Clinical care decision
- Who to be tested, a better screen, co-infection needs more aggressive care.
- Clinical insight
There are two different kinds of EMR
- Epidemic simulation (SEIR)
- Impact of policy interventions.
- 10-12 diverse inputs, which are all guesses at the moment.
- Example: Epidemic curve - An example of Open AI
- Take away: Despite everyone best’s effort, we don’t have the best accurate input.
- Simple calculators that tell us in the next few days:
- Few inputs: cases, hospitalization, bed capacity.
- Hard to get reliable counts of these simple inputs.
Suggestion: Use hospitalization data:
- Local region for Healthcare system capacity.
- There is a 12-14 days lag between intervention and an apex. Notice: day-to-day variation can be misled. A growth rate of 15% could lead to 5x-6x times higher than it is when you intervened.
- Example: CoViD-19 tool provided http://surf.stanford.edu/covid-19-tools/
Issues in Responsible Reporting of CoViD-19
- Personal story about a speaker. Focus on containment.
- Tough decisions about social impact of news with example of Italy. How to keep a balance between informing and alarming.
- Public health crisis: misinformation, disinformation, spread of false news.
- Example of some rumours by the Ebola epidemic and Twitter account. Done by an internet agency.
- A free press, but journalists are not well-supported.
- News organizations have no plan for covering epidemics/ pandemics. Lack of health/ science dedicated to writing about this topic.
- Solutions about information toolkits.
Global best practices in Controlling the CoViD-19 Pandemic.
Four responses: South Korea, China, Hong Kong and Singapore.
- Singapore: Strict containment with strong strict travel restriction, quarantine, temperature screening, rapid tests. 2000 tests/ day. Did not have a lockdown, shelter in place. Unified gov response and transparent communication with daily whats-app messaging, a comic book about quarantine, developed an app called trace contacting.
- South Korea: Rigours testing. Rapid testing 100k/ day, 600 test centers, 300k at peak. High-surveillance, quarantine and communication. Lack of strict lockdown.
- China: A bold and aggressive approach. Expand their infrastructure very rapidly: 40,000K to Wuhan and 15K epidemiologists to do contact tracing. Strategy: high tech.
- Hongkong: Closure of schools, businesses, amusement; wristbands for tracking those in quarantine.
- Extensive preparation with the 2003 SARS-CoV epidemic.
- Swift and decisive physical distancing.
- Rapid testing.
- Frequent coordination between public and community.
- Example of Italy: Containment is possible if you test early with a drop 3% -> 0.25% infected.
- Example of Iran.
- The US: Lack of testing. State by states health policy.
- India: Massive lockdown.
- Taiwan: Big Data Analytics.
- Best practices and the future:
- Human and planetary health: Environment, human and animal health.
- Surveillance to detect emerging zoonotic disease in animals and humans.
- Evidence-based intervention to build community resilience to prevent, adapt and respond to pandemics.
Tagged #summary, #note, #book.