COVID policies may have averted 200,000 COVID deaths in U.S. But . . . that equals less than 1 day of life span extension per person. Has living this way for a year been worth it to you to add 1 day to your life?
Based on the comparison to Sweden and other countries that did not use a lockdown strategy to combat COVID, it may be that our lockdown strategy of issuing stay at home orders, mandating mask wearing, closing schools, closing churches, and closing restaurants and other businesses did not reduce COVID deaths at all compared to a strategy of simply informing people of their risk based on their age and health status and of the measures they could take to reduce their own risk and then letting them modify their behavior as they see fit, with no mandatory restrictions at all. But logically the lockdowns must have reduced infections and reduced COVID deaths somewhat. That was certainly the intent of the strategy. So how much might the U.S. strategy have reduced COVID deaths?
First, to estimate how many deaths our strategy might have prevented we need to know what percentage of the population has been infected and then what the infection fatality rate (IFR), or percentage of infected people who die, is. We know how many people have died of COVID, the numerator of that fraction, but we do not know as accurately how many people have been infected, the denominator.
To determine how many people have been infected, the best method is to conduct a seroprevalence survey, collecting blood from, ideally, a random sample of the population and assaying the blood or plasma for antibodies against the SARS-CoV-2 virus. The assumption is that if anyone has been infected, he or she will have raised antibodies to the virus and the antibodies will be detectable months or years later.
The best COVID seroprevalence study I am aware of in the U.S. involved collecting plasma from dialysis patients nationwide in early July and testing over 20,000 of them for COVID antibodies (1). The result was that 8.0% of the samples were positive for COVID antibodies, and after adjusting from the age, demographics, and zip codes of the patients to the age, demographics, and zip codes of the U.S. as a whole the researchers arrived at a conclusion that 9.3% of the U.S. population had been infected as of early July (reference 1). It takes about 10 days after infection to develop antibodies, so I consider that to be the infection rate as of July 1. It takes about 2 weeks after infection, on average, for people to die of COVID if they die from it. So to calculate the IFR I am using deaths as of July 15 to compare to the infections as of July 1. On July 15 we had 140,775 COVID deaths in the U.S., which if 9.3% of the population of 328 million was infected as of July 1 gives an IFR of 0.46%. That 0.46% IFR should be viewed as probably an upper estimate of the IFR in the spring wave of COVID because the data from (1) is more likely to underestimate the infected population than overestimate it because, first, we know many infected people never raise detectable antibodies (reference 2) and, second, one would expect dialysis patients, since they are rather sick, to be more careful to try to avoid COVID infection than most people and therefore to have a lower infection rate than the general population. Also, other researchers have calculated a 0.2% IFR in Los Angeles county (reference 3).
We had 140,775 COVID deaths as of July 15 and we have 267,472 on November 25, 2020. If the fatality rate per infected person has remained constant in that time at 0.46%, we now have 17.7% of the population having been infected as of November 25 (it is a 1.9-fold increase in deaths and if we multiply 9.3% x 1.9 the result is 17.7%.)
The percentage of the population infected is probably higher than that, because the fatality rate per infected person is clearly going down due to learning more how to treat patients and the virus mutating to a less lethal form. (The fatality rate for hospitalized persons in New York was 27% in April and is now 3%.) So the seroprevalence in the U.S. has to be at least 17.7% and is probably about 20% as of November 25 (Note A). The most our strategy could possibly have achieved is prevent the other 80 or 82.3% of the population from being infected. Obviously that is not the case, and if it were it would mean we will have no cases or deaths going forward. A fairly generous estimate would be that it has prevented and will prevent about 15% of the population from being infected, about three-quarters of percentage that has already been infected, and thus prevent 200,000 deaths (Note A). That does not fit the data because if our strategy had prevented that many cases and deaths, then our death rate should be about half that of Sweden’s instead of being more than Sweden’s. But let’s say it did.
If we are averting 200,000 COVID deaths, how many person-years of life did that save? We estimate elsewhere that the average life expectancy of the COVID dead had they not contracted COVID, considering both their age and their health condition, is 4 years. If our strategy has saved 200,000 lives from COVID death, then with an average life expectancy lost of 4 years, saving 200,000 lives would mean 800,0000 person-years of life. That sounds like a lot. But with a population of 328 million (dividing 800,000 person-years into 328 million persons) that works out to 0.0024 years of life extended on average or 0.9 days. And that is not 0.9 days in your 20s; that is 0.9 days at the end of your life in poor health.
Do you want to claim our strategy has prevented the infection of everyone in society who is not yet infected, and it will forever prevent those infections and we will have no more deaths from COVID after November 25, 2020? OK, that is not true but if it were then it has saved 5.5 days of life on average. Do you want to claim that the average life expectancy lost is 12 years rather than 4 years? OK, that is not true when you consider the pre-existing health status of people who die of COVID, but if it were true, then our strategy has saved 2.7 days of life expectancy on average. Do you want to claim both that our strategy prevented infection of everyone not yet infected as of November 25, 2020, and that the COVID dead were not in worse health than average for their age? Then our strategy has still only saved 22 days on average.
But the reality is that all the sacrifices we have made have added less than 1 day per person of life expectancy. I calculate 0.9 days even with a generous estimate of how many lives have been saved that is close to a worst case scenario of what might have happened if we had ignored COVID. Does anyone think it has been worth living the way we have lived for these 10 months, and probably living this way for a year, to add 1 day to our lives? I certainly don’t. I wouldn’t even be willing to give up handshakes and hugs for a year to add one day to my life, let alone all the other sacrifices we have made.
Now, some will say, “Well, you are not making these sacrifices for yourself; you are making them for others, for the elderly who are at risk of death.” I would respond, your life is just as valuable as anyone else’s. You are just as worthy of love and consideration as anyone else. The Bible says “Love your neighbor as yourself,” not “instead of yourself.” The proper way to analyze this ethically is to ask, “Would you be willing to make these sacrifices for your own benefit, to extend your own life?”
Second, some will say, “Well, you are hoping to add 1 year, or 4 years, or occasionally 20 years to someone’s life by preventing a COVID death.” OK, then the question to ask is not “Would you make these sacrifices to add less than 1 day to your life?,” but “Would you make these sacrifices for a 1/400 chance of adding 1 year to your life or a 1/1,600 chance of adding 4 years to your life.” It is the same calculation and phrasing it that way with probabilities just confuses the issue. The better way to phrase the question is simply: Would you be willing to make these sacrifices, to live the way we have lived for one year, to add 0.9 days to your life?
Would you? Do you think it has been worth it to live this way for a year to add less than a day to your life?
Anand, S. et al. 2020. Prevalence of SARS-CoV-2 antibodies in a large nationwide sample of patients on dialysis in the USA: a cross-sectional study. The Lancet 396: 1335-1344. Published: September 25, 2020 DOI: https://doi.org/10.1016/S0140-6736(20)32009-2.
Burgess, S. et al. Are we underestimating seroprevalence of SARS-CoV-2? BMJ 2020; 370 doi: https://doi.org/10.1136/bmj.m3364 (Published 03 September 2020) Cite this as: BMJ 2020;370:m3364
Sood, N et al. Seroprevalance of SARS-CoV-2-specific antibodies among adults in Los Angeles County, California in April 10-11, 2020. JAMA 2020; 323:2425-27.
I assume here that the infection fatality rate (deaths per person infected) has stayed constant.
Deaths due to COVID increased from 140,775 on July 15 to 267,472 on November 25, a 1.9-fold increase;
if infected persons increased by the same proportion, it is 9.3% x 1.9 = 17.7% as of November 25.
If the infection fatality rate (IFR) or deaths per infected person has gone down since July 15, which almost all experts believe is true, then the percentage of the population infected has increased more than the number of deaths and is even more than 17.7% as of Nov. 25, 2020. I would estimate it is actually about 20% as of November 25.
The evidence that the IFR is lower in the fall/winter wave than it was in the first wave last spring is that the fatality rates per hospitalized person and per diagnosed infected person are both going down. In New York city in the spring 27% of hospitalized COVID patients died; now only 3% do. For Minnesota I calculate from the state health department numbers 36.3% of hospitalized COVID patients before July 15 died, and since then it has been 21.9%. Also 3.9% of laboratory-confirmed cases in Minnesota died before July 15 versus 0.93% since then.
So about 20% of the U.S. population had been infected by November 25. New York state was at 33.6% of the population infected as of early July, according to the same study I have been relying on (1), and New York state has had only 5.7% as many deaths between July 15 and Nov. 25 as had happened by July 15. If the number of infected persons has stayed proportional to deaths, then New York state is now at 35.5% infected as of November 25, 2020. So their infections are dramatically slowing once they hit 30% of the population infected, suggesting that is the range where herd immunity begins, and that the nation is unlikely to exceed 40% of the population infected in 2021, even if we did not have a vaccine.
200,000 additional deaths corresponds to an additional 13.2% of the population infected (assuming the same infection fatality rate as prior to July 15), which added to the 17.3% infected as of November 25 would be 31% infected in an alternate world where we took Sweden’s approach of no mandatory closings and restrictions. It seems unlikely the whole U.S. could have gone above 31% infected when New York is plateauing at less than 40% infected.
Another way of looking at it is that the University of Washington COVID model as of now (Dec. 1, 2020) projects a best case scenario of 406,000 COVID deaths in the U.S. total on March 1, 2021 (in the absence of a vaccine). Their realistic projection is 471,000 deaths on March 1, 2021. My estimate of 200,000 additional COVID deaths in the U.S. if we had done nothing or followed Sweden’s model, added to the 406,000 best case projection for March 1, 2021, with our current policies, means instead we would have had 606,000 COVID dead on March 1, 2021. At the case fatality rate of 0.46% that we had in the first wave, that would mean at least 40% of the total U.S. population infected. When New York state is still not at 40% infected and looks like it may never get much past that point, it seems very unlikely the U.S. as a whole would have gotten to 40% infected in the absence of the restrictions we have imposed.
So I think the estimate of 200,000 additional COVID dead if we had done nothing is probably an overestimate. It is pretty close to a worst case possibility.