Thinking Citizen Blog — Modeling the Corona Virus (II)
Today’s Topic: Modeling the Corona Virus (II): art not science, state by state variation, constantly evolving
So what’s the best corona virus model you have ever come across? Is any website tracking the relative performance of the most widely used models? If not, why not? Is this a tremendous opportunity for an enterprising undergraduate or graduate student in computer science, math, or statistics? Today, a continuation of last week’s note on the “Murray” model from the University of Washington. Experts — please chime in. Correct, elaborate, elucidate.
STATE BY STATE VARIATION: death forecasts revised down in many states but up in others
1. North Carolina (-80%), Pennsylvania (-75%), California (-70%), Texas (-65%) and Washington (-55%).
2. Up over 50% in New York, and 133% in New Jersey.
3.” More states are also predicted to experience their peak demand for hospital resources sooner. The Murray model now shows that Washington hit its apex on April 2–17 days earlier than forecast two weeks ago. California is expected to hit its peak on April 14–10 days earlier — though Gov. Gavin Newsom has projected the date at sometime in mid-May.”
WHY THE REVISIONS? “Models only as good as the assumptions that feed them.” (Fauci)
1. “The Murray team say their projections changed because their early modeling was based largely on data coming from China. “The time from implementation of social distancing to the peak of the epidemic in the Italy and Spain location is shorter than what was observed in Wuhan,” they note. This underlines the limitations of existing models.”
2. “As Anthony Fauci of the White House coronavirus task force noted last week, models are only as good as the assumptions fed into them. One problem is there isn’t consensus among public-health experts about how variables other than social distancing affect transmission and fatalities. Nor, by the way, do models account for how treatments could reduce deaths and hospital utilization.”
3. “The Murray March 26 study noted that its model accounts for population age structure but not the ‘many other factors that may influence the epidemic trajectory: the prevalence of chronic lung disease, the prevalence of multi-morbidity, population density, use of public transport, and other factors that may influence the immune response.’”
HUGE VARIATIONS EVEN AMONG POPULATIONS THAT WERE LOCKED DOWN AT THE SAME TIME
1. “New York’s per capita fatalities are about 25 times higher than California’s though both states shut down businesses at about the same time. New York state is about twice as dense as California, and New York City’s population is about 300 times more concentrated than the U.S. as a whole and 11 times more than Los Angeles County. More than half of people in New York City use public transportation where they have a higher risk of virus exposure.”
2. “Even within cities there is substantial variation among populations. In Chicago, the fatality rate for blacks is about six to seven times higher than for whites or Hispanics, which may be due to higher rates of cardiovascular disease and HIV. Life expectancy for black Americans is also about seven years lower than for Hispanics and three-and-a-half years lower than for whites.”
3. “In New York City, the coronavirus has hit the Bronx and West Queens much harder than Manhattan. But those neighborhoods also have large numbers of multigenerational households, which is thought to have stoked outbreaks in Italy and Spain. This could also help explain why per capita fatalities are 70% higher in Los Angeles County than in San Francisco.”