In this first paper, the author attempted to estimate how many total cases there were in the US as of April 11 by using the death rate in South Korea, positive test rates in the individual states and death rates in those states. (SSRN Paper) He finds a highly variable level of rates across the US states, but that only about 1 in 5 infections has been identified. He is making a wrong assumption in thinking that South Korea had identified everyone with an infection and that therefore its death rate was an accurate reflection of ones that would occur everywhere. He also assumes there is some relationship between positive tests and death rates in a state, but that also is a questionable assumption because testing strategies, and consequently positive tests, varied widely by state. He tested by using asymptomatic rates from Iceland and China, which is really questionable. He comes up with a South Korea death rate of 2.3%, which is absurd. His total estimate of cases in the US is therefore likely inaccurate as well. Nice try, but shows how bad data and bad assumptions make a paper pretty worthless. That’s why we need peer review for papers from non-established researchers.
This paper addresses the issue of bias in understanding the epidemic by only looking at those cases which are identified by a positive test result. (Medrxiv Paper) The author notes, as other work has begun to, that the early course of this epidemic disproportionately features testing of and focus on serious illness while not detecting the much larger group of asymptomatic and mild cases, thereby inflating case fatality rates. They build an alternative model that reflects and adjusts for this sampling bias of more serious illness. It also splits the infected group more appropriately into mild and serious illness. Their testing of the model suggested that not accounting for testing sample bias does have a very significant effect on case fatality rates.
This paper is the next effort from a Stanford group led by renowned statistician John Ioannidis. (Medrxiv Paper) The group is continuing its work of parsing death rates. In this paper they examined deaths in several countries and 12 US states to compare deaths of those under and over age 65 and for people with and without underlying health conditions. Individuals under age 65 represented only 4.8% to 9.35% of all deaths in ten European countries, 13% in the UK and 7.8% to 23.9% in the United States. The older population had risks of death ranging from 14 to 84 times higher than the under 65 group. The absolute risk of dying from coronavirus was minimal for the younger group, equivalent in most cases to the risk of dying in a car accident. The proportion of people under age 65 with no pre-existing conditions who died was .6% to 2.6%. As the authors note, this extremely skewed distribution of serious illness and death has caused broad-based lockdowns that cannot now be justified by the real known risk to most of the population.
The Imperial College model effort, led by the now disgraced and ousted Neil Ferguson, predicted a wildly high number of deaths around the world from coronavirus, largely because they used bad, and known to be bad, Chinese data. People had been clamoring to see the actual computer code underlying the model. It was released and now one veteran software coder has pointed out a number of issues with the code itself, including its dated approach and the fact that it has a number of serious bugs. (Model Code) The modeling group apparently did not include an experienced software writer and did not test the software adequately. So in addition to using bad data, the model had bad code.
A doctor writes another column questioning the notion of flattening the curve by extensive lockdowns. (Column) His basic point is that flattening the curve isn’t saving lives at this point, it is just delaying the occurrence of death. He assumes that the virus can’t be contained or prevented from spreading. He assumes that contact tracing has less value than proponents suggest. He believes there is little danger of health resources being overwhelmed. He believes people do their own mitigation of spread efforts based on their perception of risk. He thinks we can identify hot spots quickly and react to them. He doubts that delaying infections changes the ultimate case fatality rate, which is likely very low in any event. He also notes that the longer we let the epidemic linger with extreme lockdowns, the more health harms we will experience from other causes.
Next up, another antibody study, this time from Switzerland. (Medrxiv Paper) The study, conducted in Geneva among a population sample, did weekly antibody surveys for 8 weeks of the same group of around 1300 people over the age of 5. The authors reported the first three weeks’ results. Overall, in the first week, 3.1% were positive, in the second, 6.1%, in the third 9.7%. The 5-19 age group had a positive rate of 6%, the 20-49 a rate of 8.5% and the 50 and older group, only 3.7%. There was wide uncertainty in all these results. As of April 30, Geneva had 5071 cases, or around 1% of the population. The authors estimate that there are ten undetected cases for every one reported by infection testing. It is remarkable how the studies are seeming to settle around this number. The test misses more true positives than true negatives. The authors believe rates will rise in future weeks. They found roughly similar rates in children and middle-age adults, suggesting that children do get infected at similar rates, but have milder illness. Switzerland had some lockdown measures in effect which likely dampened spread, especially among the elderly. The authors suggest this is a long way from population immunity, but that depends on the extent to which people are susceptible to being infected after exposure.
And here is a short column from a Swedish epidemiologist in which he explains that the extreme lockdowns don’t make any difference and cause some harms. (Lancet Column) He points out that Sweden’s death rate is higher than some countries but lower than others, and what matters is death rate at the end of the epidemic. He thinks there is little we can do to prevent spread of the epidemic and it is largely a silent phenomenon, almost all cases are asymptomatic or mild.
And finally, a column about statistics and numbers and how much we don’t know yet about the coronavirus and the epidemic by a well-known UK science writer. (Spectator Column)