Estimating the ultimate toll of the novel coronavirus is useful for health care planning, understanding the virus’s economic impact, and for our personal sense of wellbeing. The problem is, as stated by Tom Frieden, former director of the Centers for Disease Control and Prevention, “Anyone who says they know where this is going doesn’t know enough about the virus: We just don’t know.”
Uncertainty about the current disease has ample precedent. At the outset of the H1N1 swine flu outbreak in 2009, President Obama’s Council of Advisors on Science and Technology predicted that the virus would result in 900,000 to 1.8 million hospitalizations and 30,000 to 90,000 deaths. Actual figures a year later were found to be 274,000 hospitalizations and 12,000 deaths.
It’s not as if epidemiologists don’t have sophisticated techniques for predicting consequences of such a disease. Mathematical disease models rely on variables such as r0, representing the contagiousness of a disease, and case-fatality ratio (CFR), the lethality of the disease. But, with the possible exception of east Asian nations such as Korea and Japan, which were prepared from earlier experience with SARS, most nations don’t have a good idea of the most important figure that determines variables for the sophisticated models — the actual incidence of the disease, which can be found only with wholesale testing given how many coronavirus cases have no or few symptoms.
In many fields, practitioners rely on heuristics, logical rules of thumb, as alternatives or checks for more sophisticated models when, for whatever reason, they cannot be relied upon. With this approach, known facts can be interpreted with judgement or intuition applied to a narrowed scope of unresolved issues, hopefully facilitating an answer that, however imprecise, at least provides directional guidance. Of course, where sophisticated models are viable, they are likely preferable. But the heuristic approach may nevertheless offer value.
With coronavirus, as noted, the most important variable, incidence of the disease, is not known, so consider what is available. Generally, human deaths are closely and accurately tracked, with Iran, which has been digging many more graves than announced deaths, being an exception. As with much data from China, questions have been raised about the accuracy of their statistics. But as Catherine Troisi, an epidemiologist at the University of Texas Health Science Center at Houston, states in Fortune “[I]f they’re covering something up, they’re not doing a good job of it.” Perhaps some deaths were missed early in the epidemic’s outbreak, but trends should not be affected.
Another statistic known with good reliability is population. The New York Times published estimates by Nebraska University public health researcher Dr. James Lawler in which 99 percent of projected fatalities occurred among those over 60, so estimates of a region’s population over age 65 should be a reliable statistic proportional to coronavirus lethality.
Examples with longer coronavirus history allow application of these knowable statistics to the U.S., China and its Hubei province, Italy and Korea, which all experienced earlier outbreaks.
Eleven percent of China’s population is 65 years old or older. Hubei, where the city of Wuhan is located, has a population of 58.5 million, so 6.4 million should be 65 plus. As of March 17, the province had experienced 3,111 deaths, according to the excellent Johns Hopkins Coronavirus Resource Center. The resulting proportion of deaths to people 65 or older in Hubei has thus been 0.048 percent; but, unfortunately, while daily deaths have fallen to low single-digits from over 100, it can’t be assumed there won’t be still more fatalities. Active coronavirus cases (total cases less recoveries and deaths) in Hubei peaked in mid-February, approximately two months after the disease was first documented.
If it’s assumed that Hubei deaths won’t more than double from their current level, the resulting 0.096 percent death/elderly ratio applied to 53 million Americans 65 and over equates to 52,000 U.S. deaths. The overwhelming majority of China’s coronavirus cases are in Hubei, so comparison with the whole country produces much lower estimates. China’s overall 65 plus population is 25 times Hubei’s so, comparing the U.S. to China as a whole, projected deaths fall to about 2,000.
Hard though it may be to believe, it is possible to have a worse outbreak than in Hubei, where the emerging pandemic was initially covered up. Sadly, that is the case for Italy. From the first coronavirus death on February 21, Italian deaths rose to 2,503 by March 17. It took Hubei over two months to reach such a level.
Unlike Hubei and China, which are clearly on the downslope, Italy has a rising death rate, so it’s difficult to project how far Italy is through its contagion cycle. At a comparable stage of the contagion, Italian deaths are double those of Hubei, so if we double the Hubei-based estimate above, it produces a projected 104,000 U.S. deaths.
South Korea is the other country with enough history to compare with the U.S. Korea’s first death occurred on February 20, nine days earlier than for the U.S. At a comparable stage to March 17 for the U.S. with 97 deaths, Korea had 60 fatalities. The U.S. death rate was 62 percent higher. Korea’s 65-plus population is 7.2 million for a 0.001 percent ratio of deaths to elderly. At a comparable stage of spread to Hubei, Korea had just 2.4 percent of Hubei’s deaths. Korea’s death rate, while low, is still heading up. It may be conservative to estimate that Korea is just 5 percent through its total contagion cycle, so multiplying the country’s current death/elderly ratio by 20 and adding 60 percent for the U.S. produces projected deaths of 19,000.
More important than the specific numbers in this column is a frame of reference to put the coronavirus outbreak into an understandable context. It isn’t precise to make subjective judgements about whether the U.S. will be better or worse off than Hubei, Italy or Korea or how far those countries are through their infection cycles. But for most people, including experts, it can be more intuitive to grasp than guessing at r0 and CFR in a complete absence of data.
This column’s projections can reasonably be questioned. They are much lower than most estimates using epidemiology models, which characteristically range from 200,000 to 1.7 million. If the figure ultimately comes in well below those estimates, it may be due to precautions taken as a result of those estimates. This column focuses on deaths because that is the most reliable data. But demands upon the health care system for hospitalization should be proportional.
While the health care system must prepare for the worst case, there is a reasonable scenario for a better outcome — that deaths won’t be much worse than a severe flu and that the corner can be turned within the two months it took in Hubei. Let us hope — and act.
Douglas Carr is a financial markets and macroeconomics researcher. He has been a think tank fellow, professor, executive and investment banker.