TOP
The novel corona virus-affected world has been inundated with medical testing and related published data like never before. The volume and nature of testing has evolved significantly in the last few weeks. Antibody tests, which reveal whether a person tested has been infected with SARS-CoV-2, have been promoted as key to reopening the economy and restoring a degree of normalcy. The hope is that people who have been infected and consequently developed antibodies may be immune to reinfections and can return to “normal” life. What does this mean for assessing the state of the pandemic at large?
On Monday, April 20, we heard officials in Los Angeles County suggest that roughly 4.1% of the county’s adult population has already had the coronavirus, which translates to around 442,000 people.1 That’s a much higher number than the county’s confirmed cases, which totaled 13,816 as of Tuesday, April 21 at 9:00 a.m.
Only a few days later we heard from Governor Cuomo of New York that antibody tests had been performed on 3,000 people selected randomly across approximately 20 grocery stores and without considering any evidence of symptoms.2 The test results suggest around 14% of New Yorkers across the state and 21% of New York City residents have been infected with the virus. This translates to approximately 2.7 million infected people across the state as compared to 263,000 reported cases and a case fatality ratio of a little over half a percent.
Random sampling of the population yields the true magnitude of the COVID-19 outbreak. Since the count of confirmed cases overlooks some of the milder or asymptomatic cases, the observed case fatality ratio is exaggerated. For example, New York has a case fatality ratio of nearly 6% for reported cases and just 0.6% for the number of cases implied by the antibody survey.
A similar test was conducted in the small German community of Gangelt with a population of 12,500.3 This was the first cluster of confirmed COVID-19 cases in Germany starting a chain of infections after a carnival session in mid-February. Preliminary results, from a yet unpublished survey and based on a small sample size, would suggest an existing level of immunity of 14% in that community (antiSARS-CoV2 IgG4 positive with a specificity of >99%) at the beginning of April. About 2% of the surveyed people had a current SARS-CoV-2 infection. In total, the infection rate for the sampled population is 15%, which results in a case fatality ratio of just under 0.4%.
This is all good news for Life insurers as case fatality ratios are coming into line with early expectations of less than 1%, but what about for P&C insurers whose topline is inextricably linked to the reopening of the economy, unemployment levels and GDP growth?
In medical diagnosis, test sensitivity is the ability of a test to correctly identify those with the condition, whereas test specificity is the ability of the test to correctly identify those without the condition.
During a recent White House press briefing, the U.S. Special Representative for Global Health Diplomacy, Dr. Deborah Birx, cautioned that “these tests are not 100% sensitive or specific.”5 She explained further that if 1% of the population of country X (of, say, 1000 people) is infected with the virus but the test is 99% specific and 100% sensitive, 10 infected and 10 uninfected people would test positive. So only 50% of those that tested positive would have a true positive result while the remaining 50% would be false positives. This margin of error can be reduced by testing the first responders or healthcare workers since their exposure to the virus has been “the greatest”. In the case of the New York “shoppers” survey, this was clearly not the case since those workers were likely at work and not grocery shopping.
We agree with the caution given by Dr. Birx (which is grounded in Bayes’ theorem) that focuses on the more practical question that needs to be asked for reopening the economy - and that is of prime interest for P&C insurers - namely, what is the likelihood that someone has had the infection given their test results are positive?
On February 4, 2020, the U.S. Department of Health and Human Services Secretary issued emergency use authorization for diagnosis of SARSCoV-2, which allowed tests to be offered based on manufacturer-reported data without formal FDA clearance. In response, dozens of companies began to market laboratory-based immunoassays and point-of-care tests.
There are two reference standards required for antibodies to SARS-CoV-2: specimens truly positive and specimens truly negative. Positive is easier although not perfect. This is achieved by drawing serum from people whose detectable virus should have antibodies 14 days after the infection. The negative reference standard is more difficult. What defines someone who is certain not to have exposure to SARS-COV-2? One possibility is to use stored specimens from before 2019.
A study performed by the UC San Francisco Medical School and University of California, Berkeley examined 12 serologic tests. The results are preliminary pending publication. Among these tests, sensitivity was 81.8-100% and specificity ranged from 84.3-100%. High rates of positive results were not reached until at least two weeks into clinical illness.
The study reinforces the need for validation using standardized sample sets with: 1) Known positives from individuals with a range of clinical presentations at multiple time points after onset of symptoms, 2) Pre-COVID-19 outbreak samples for specificity, and 3) Samples from individuals with other viral and inflammatory illnesses as cross-reactivity controls. Neither the California, New York, nor Gangelt surveys published the methodology for true serologic status. It is worth cautioning here that antibodies can cross-react with other coronaviruses. Thus, a positive test can occur in people who could not have had COVID-19.
The antibody test performed by the New York State Department of Health - Wadsworth Center cited specificity between 93-100% without reference to any validation.6 If we were to assume that post validation, the test had a specificity rate around 90% with sensitivity of around 80%, the estimated proportion of the population infected in New York would be closer to approximately 6% which would translate to approximately 1.2 million individuals.7 The good news is that even with 90% specificity, the case fatality ratio in New York would be around 1.5% which is significantly lower than the current ratio of approximately 5% that we are observing across the country.
Herd immunity describes the condition of protection from an infectious pathogen because enough of the population is immune to limit the spread of disease. Are these estimates of true prevalence enough for herd immunity?
Individuals can gain immunity to a virus in two ways: by contracting the virus and recovering from it because of the development of antibodies, or by getting a vaccine. Either may induce antibodies that protect against future infection.
The benchmark for herd immunity depends on contagiousness of the pathogen, which is measured by a value called R0. R0 indicates the average number of individuals who will get the disease from one infected patient.
The World Health Organization estimated the SARS-CoV-2 R0 between 2.5 and 3. Some scientists suggest a rate of transmission significantly greater than that. The Mathematical Modeling of Infectious Diseases Unit at the Institut Pasteur in Paris calculated a rate of 3.3 for France8 and U.S. researchers calculated a median R0 of 5.7.9
Johns Hopkins University’s Bloomberg School of Public Health predicts that we will need to reach at least 70% immunity to achieve herd immunity for the novel coronavirus (assuming R0 of 3 and using the formula 1-(1/R0) for natural equilibrium).10
They also suggest the first confirmed case in the U.S. was on January 27, 2020. Assuming an antibody testing specificity rate of 90%, and an estimated 6% of the population in New York being infected in a matter of approximately 2.5 months, achieving at least 70% immunity in the state of New York could take over two years!
A point to note here is the development of vaccines. Initial clinical tests are currently taking place. If we instituted large vaccination programs as soon as possible, we could reach our goal sooner.
In conclusion, as these forecast models are evolving, such predictions are still best described as “guesstimates” and would be largely influenced by social distancing, hospital infrastructure resiliency, and the development of vaccines.
Special thanks to our colleagues Dr. Sandra Mitic and Andres Webersinke for their contributions to this blog.
where:
