A Head Full of Coronavirus Research, Part 3

This afternoon’s rash of papers includes a couple focused on economic aspects of the reactions to the coronavirus epidemic and others on the usual issues related to antibodies and prevalence.

The first examines whether an extreme lockdown is really the best approach.   (Lockdown Analysis)  These papers done by economists use far better methods and more sophisticated analyses than those done by epidemiologists or other researchers, in general.  This is because economists in recent decades have been employed by business to understand consumer behavior in detail and build complex models to do so.  This paper does an extensive analysis and finds that targeted lockdowns not only are economically less damaging, but result in lower loss of life.

This paper used a survey of 10000 people to ascertain economic impacts of the lockdown.   (NBER Paper)   50% of respondents said they had endured an income or wealth loss as a result of the shutdowns.  The average income loss so far has been $5923.  The average wealth loss is $33,482.  There has been a 31 percentage point drop in consumer spending.  Respondents expected unemployment to be 13 percentage points higher, and for unemployment to not recover for 3 to 5 years.

Another paper delivers a message that would seem obvious at this point, but that apparently most Governors and other politicians haven’t been able to grasp.  (NBER Paper)  The epidemic affects the population in different ways in terms of infection rates and seriousness of illness, especially by age, so maybe we could target our mitigation of spread measures and minimize economic damage.  At least the authors recognized that the population was heterogenous, but they used data from the disavowed Imperial College study on epidemic parameters, which makes some of their numbers suspect.

Next up is a paper on the nature of antibodies in a small study of hospitalized patients and asymptomatic carriers.  (Medrxiv Paper)  The patients each had several tests over the course of treatment.  All hospitalized patients had antibodies to two portions of the virus by 13 days after symptom development.  Women and patients with severe disease tended to have stronger levels of antibody.  There did not appear to be cross-reactivity with seasonal coronavirus strains.  Looking at 25 asymptomatic patients, lower levels of antibodies were found and none in a number of patients.

Another study looked at prevalence of infection in Luxembourg.  (Medrxiv Paper)  A representative sample of the population, consisting of over 1800 adults between 18 and 79, was selected for both infection and antibody testing.  The study is intended to consist of repeated testing for up to a year and these were initial results.  A small number of study members had had positive infection results before the study began and all these people had positive antibody tests.  Several other study members who had not previously tested positive, also had antibodies.  A very small portion, .3%, of the population was estimated to have been infected.

Dr. Ionnadis is back with another paper, this one summarizing the results of antibody studies and calculating infection fatality rates from that analysis.  (Medrxiv Paper)  To be included, studies had to consist of at least 500 people.  The infection fatality rate was determined by taking deaths as of a certain point and dividing by the estimate of true infection rate.  12 studies were included with adjusted antibody rates of .31% to 33%.  Infection fatality rates were estimated by the author as between .02% to .4%.  Early estimates based on Chinese data gave a CFR of 3.4%, but this data almost certainly omitted huge numbers of cases and a large number of deaths.  Notwithstanding what should have been obvious concerns about data quality, a number of models predicted massive deaths, and those predictions have led to extensive job losses and other damage.  In addition to the estimates, the accompanying commentary notes that extreme skewing in age of deaths would make use of separate CFRs prudent.