From George Avery, PhD. MPA
Dr. Avery has a PhD in Health Services Research from the University of Minnesota School of Public Health, and has conducted significant research in the area of public health emergency preparedness, including five journal articles and two book chapters on the topic. He has served on several CDC advisory boards, including a panel on preparedness and emergency response centers, and consulted for the Defense Department on Medical Civic Action program doctrine. He has edited a special issue of the research journal Bioterrorism and Biodefense and served as a reviewer for the Journal of Homeland Security and Emergency Management as well as Disaster Medicine and Public Health. He is a health services researcher with a medical analytics firm in the Midwest, and has formerly been a professor with the public health program at Purdue and worked from 1990-2000 with the Arkansas Department of Health’s Division of Public Health Laboratories.
We are seeing a panic reaction towards the newly emerged SARS-COVID-2 [Wuhan] epidemic, marked by panic buying of items including the much-joked about toilet paper, drastic action by political figures that often impinges on basic civil rights, and potentially devastating lasting economic impact. Much of this has been fueled by naïve and sensationalist reporting of fatality rates, such as a March 10, 2020 report by the Bloomberg news service that implies that 3.4-3.5% of infected individuals die (https://www.bloomberg.com/news/articles ... rus-update
). This has caused comparisons to the 1919 Influenza A:H1N1 pandemic and its 2.5% case fatality rate, which would qualify as a level 5 event on the CDC’s Pandemic Severity Index (PSI) and has led to a panicked overreaction worldwide. This case fatality rate, however, to a trained epidemiologist is obviously a significant overestimation of the actual fatality rate from the disease.
Ascertainment bias is a systematic error in statistical estimation of a population parameter resulting from errors in measurement - usually, in undermeasurement of a parameter. In this case, we are underestimating the actual number of cases in the population, which is the denominator in the calculation of the estimated case fatality rate. We are accurately estimating deaths, but to get the case fatality rate, we divide deaths by our estimate of the number of cases. Because that it too low due to measurement error, the estimate of the case fatality rate is too high.
For example, for a hypothetical disease if we have three deaths and observed ten cases, then the case fatality rate is 30% (3/10=0.3 or 30%). If, however, there were actually 300 cases, and only 10 were observed and reported, ascertainment bias has led us to underestimate the cases and overestimate the case fatality rate, which is actually 1% (3/300=0.01 or 1%).
In this case, in the absence of population-based screening to more actually estimate the total number of cases, we are only counting cases who are sick enough to seek health care -- almost all disease reports are made by healthcare professionals. We are missing people who have no more than a cold or who are infected but show no symptoms, individuals who almost certainly make up the overwhelming majority of actual cases. Thus, as in my hypothetical example, we are overestimating the case fatality rate for the disease.
There is, however, data available on SARS-COVID-2 [Wuhan] that allows us to get a better grasp on the actual case fatality rates for the virus.
One case is that of the cruise ship Diamond Princess, which achieved some notoriety from the well-publicized outbreak among its 3711 passengers and crew in January and February of 2006. Held aboard in constricted quarters, the population was subject to 3068 polymerase chain reaction (pcr) tests, which identified 634 individuals (17%) as infected, with over half of these infections (328 ) producing no symptoms. Seven infected passengers died, all of them over the age of 70. Adjusting the data for age, researchers at the London Institute of Tropical Medicine have estimated a fatality rate per infection (IFR) for the epidemic in China of 0.5% (95% CI: 0.2-1.2%) during the same period. This is far below the earlier estimates of 3.4% or greater that were promoting panic over the epidemic. See Russell et al, Estimating the infection and case fatality ratio for COVID-19 using age-adjusted data from the outbreak on the Diamond Princess cruise ship, MedRXIV 2020 at https://www.medrxiv.org/content/10.1101 ... 2.full.pdf
South Korea has also implemented far wider population-based screening than the US, expanding their screening past suspected cases to voluntary population screening in geographies frequented by identified cases. As of March 15, as Stanford University economist Richard Epstein has noted, they performed over 235,000 tests and identified 8, 162 infections with 75 deaths (CFR=0.91%). Again, only about 10% of the deaths were in the population under the age of 60. See https://www.hoover.org/research/coronav ... t-pandemic
. While their population screening efforts were far better than that of the United States, this was still not a broad-based screening effort (such as was used on the Diamond Princess), being biased because while it looked at a broader population, it still was enriched with cases by looking only at a segment of the population with a higher risk. Still, the case fatality rate is significantly below the 3.4% rate that caused the public panic.
What we are likely seeing, in my estimation, is an epidemic with a real case fatality rate between 0.2 and 0.5%, which is similar to the 1957 Asian Influenza A:H2N2 or 1968 Hong Kong Influenza A:H3N2 pandemics, which were also essentially virgin field respiratory epidemics. These pandemics rate, not as PSI5 events, but as PSI2 events on the CDC scale. They are certainly atypical and more severe than a PSI1 event (such as a routine seasonal flu epidemic), but not a shattering event like the 1919 influenza A:H1N1 pandemic. These earlier pandemics essentially tripled the number of deaths due to influenza experienced annually, and were posed little long-term economic or other damage to the population despite being handled without the extreme measures that are currently being adopted or proposed by political figures. Like those pandemic events, SARS-COVID-2 [Wuhan] has its most significant impact on elderly or otherwise compromised individuals, with few fatalities observed in the population under the age of 60. From what we have observed, half of those infected show no symptoms, 40% show mild symptoms such as a cold, and only about 2% advance to serious or critical illness. What is needed now is for politicians and the population to pause, take a deep breath, and address the epidemic with rational measures, such as social distancing of the older population, ring screening around identified cases, quarantine of identified infected individuals, and adequate hospital triage systems to protect other patients and health care staff rom infection in order to preserve our ability to treat the most severe cases. This is a strategy identified by myself and colleagues at Purdue in 2007 to ensure adequate capacity to deal with another true influenza pandemic, and it applies to this one as well.