Commentary: What a lower death toll than first expected means

By Sandra McCoy and Pia MacDonald / The Washington Post

Covid-19 has killed more than 40,000 people in the United States so far; fewer than the 100,000 deaths originally projected. Hospital systems outside New York, New Jersey and Louisiana have mostly avoided being overwhelmed with coronavirus patients. Some states such as California may have managed to “flatten the curve” and slow the number of infections.

Now some people believe this suggests that the U.S. control strategy, with its disastrous economic consequences, was an extreme overreaction. Small but vocal protests have been organized in some states, calling for restrictions to be lifted immediately. President Trump has called to “liberate” states with businesses closed by social distancing orders. Seeking to ease economic strain, governors in Georgia, Tennessee, South Carolina and other places are considering easing restrictions and opening some, if not all, businesses in the coming days.

But public health professionals like us see the current mitigation strategy as proportionate and warranted. The lower-than-expected death tolls don’t demonstrate that the U.S. response to the coronavirus pandemic was a mistake; they show that it’s working.

Of course we should rigorously interrogate the rationale of an epidemic-control strategy that exacerbates societal inequity and has resulted in more than 95 percent of Americans being told to stay at home and 22 million newly unemployed workers since March 14. And it’s smart to consider alternative explanations for epidemic trends we observe and ask whether the control strategies are really responsible for those trends; this sort of question is, in fact, key to the scientific process.

Yet the paradox of public health is that when it works, nothing happens. Most people agree to be vaccinated without knowing if they will ever be exposed to an infectious disease like measles. Similarly, we may never know for sure how bad the Covid-19 pandemic would have been without aggressive control measures. But in confronting an easily transmissible new virus that is lethal to humans and has no known treatment or vaccine, our best option is to err on the side of caution.

In the exponential-growth phase of an outbreak, early intervention can seem like an overreaction, but it pays dividends in the future. Epidemiologist Britta Jewell explained last month that in the early stages of an epidemic, when cases were growing by as much as 30 percent every day, early action can significantly reduce the spread: “If you act today, you will have averted four times as many infections in the next month: roughly 2,400 averted infections, versus just 600 if you wait one week. That’s the power of averting just one infection.”

We can see preliminary evidence of the benefits of early and/or aggressive action in Taiwan and South Korea. Despite proximity and close ties to China, Taiwan took swift action weeks before its first covid-19 death, activating its Central Epidemic Command Center on Jan. 20, which began implementing border control measures coupled with rigorous case identification and quarantine using mobile phones for monitoring. Taiwan also scaled up production of personal protective equipment. So far, Taiwan has been able to avoid large-scale lockdowns and school closures and has reported only six deaths.

South Korea used similar border control and case-finding strategies, but its most notable approach was the early development and rapid deployment of testing via more than 500 clinics, including drive-throughs, as early as late January. South Korea’s epidemic peaked in late March and was quickly contained, resulting in only 238 deaths overall. In the United States, on the other hand, deaths from the virus are still doubling about every nine to 12 days even with aggressive control measures in place.

Our understanding of the pandemic and its trajectory is still on extremely shaky ground; which means our models are, too. The real number of cases and deaths the virus has caused is not clear. Testing is still dramatically unavailable, which impedes our understanding of the outbreak’s trajectory. Last week, the United States conducted about 160,000 tests per day, representing a small portion of the number of people who may actually be infected and a long way from the millions of tests per day we likely need to safely reopen the economy. Deaths are also likely undercounted, limited by the availability of testing and inconsistent procedures for attributing the cause of death, especially for deaths occurring outside a hospital.

Compounding this blurry picture are models based on extremely limited data. Models can help to compare the relative impact of control measures, but they are not crystal balls. Those claiming to predict the exact date of the peak of an epidemic or the number of deaths months into the future are alluring, but they make some epidemiologists cringe.

Covid-19 models are based on dozens of “best guesses” about test accuracy, immunity (if any), whether reinfection is possible, how much transmission comes from people without symptoms, and how people interact in communities, to name a few factors about which we still know very little. Consequently, models are very sensitive to rapidly changing information: One widely cited model initially predicted 240,000 total deaths in the United States by August, but in response to new information, revised that on March 30 to 82,141 deaths and then last week to 68,841 deaths. That doesn’t necessarily mean the models were incorrect; it often means the underlying data have changed.

Public health professionals use several strategies to slow a growing epidemic. We can reduce the number of people who are susceptible, often through use of a vaccine. We can reduce how long a person is contagious, typically through treatment that hastens recovery. Neither option is currently available. Another path is to reduce the likelihood of transmission if someone is exposed to the coronavirus; through personal protective equipment (such as N95 masks) and infection control (such as emphasizing hygiene).

Both strategies are being used in health-care settings, and some have been recommended for use by the general public. But the last and widest-reaching option is to limit how often people contact each other, which can slow disease spread by making it harder for infected people to pass the virus to anyone else. This is why social distancing is currently our only population-level strategy to flatten the curve.

We have clues that social distancing works. Comparing San Francisco, Seattle, New York City and New Orleans, we have an interesting picture of how the timing of reduced movement may have influenced epidemic trajectory. We analyzed Google Community Mobility Reports to identify the day that movement in each city first fell by an average of 50 percent or more across four categories (retail, grocery, transit and workplaces).

On the day movement fell by 50 percent in San Francisco, the city reported its first death. By the time movement fell that much in Seattle, New York or New Orleans, they had already reported more than 50 deaths each – potentially indicating a later start on widespread social distancing. Mortality in San Francisco has remained low so far (20 deaths, about 2 per 100,000), while the death rate in New York is about 114 per 100,000 people and in New Orleans, 88 per 100,000. Seattle’s trajectory presents more of a mystery; despite crossing the 50 percent movement threshold after already reporting numerous deaths, it has managed to keep the death rate to about 17 per 100,000, potentially because many of the early deaths were tied to a single nursing home.

Taken together, these data hint that very early and effective social distancing can slow Covid-19. This idea is bolstered by evidence from the 1918 influenza pandemic; communities that intervened earlier then had a lower mortality rate than those intervening later.

Ultimately, arguments about the response being worse than the disease are forcing a false dichotomy between an unrealistic option that no one ever considered seriously — letting the virus spread unchecked — and the current strategy. In an alternative universe where we let the virus tear through the country unabated, more than a million Americans would die and millions more would be infected, a catastrophic burden for society and the health-care system.

Obviously, this was never an ethical or realistic option. Celebrating the fact that it didn’t happen while criticizing the actual response ignores the value of the time gained from suppression strategies that can be used for further public health intervention planning and the development of vaccines and therapeutics.

We don’t know yet whether a middle-ground approach like the one taken by Sweden will prove effective; there, most schools and restaurants remain open, and while the elderly have been told to self-isolate, the country is not under lockdown. Some have pointed to Sweden as a success story that the United States should emulate. But Sweden’s death rate is now one of the highest in Europe, at 175 deaths per 1 million (higher than the U.S. death rate of 136 per 1 million). It’s also higher than other Nordic countries that took early and aggressive action; more than twice the rate of Denmark (64 deaths per 1 million) and more than five times that of Norway (28 deaths per 1 million). Some Swedes are angered that the strategy has exacted a heavy toll on the elderly, as a third of fatalities have been in people living in care homes.

Whether the benefits of containment outweigh the costs is a critical and evolving question. Data from epidemics in the 20th century reveal that pandemics are bad for economies, but economic growth rebounds; and it often rebounds faster in places that take more forceful action to stop disease. In 1918, cities that took action earlier and more aggressively reduced mortality also had higher economic activity in the year after the pandemic than cities that delayed. Cities that reacted 10 days earlier had 5 percent higher manufacturing employment afterward than cities that reacted later, and an additional 46 days of control measures increased manufacturing by 6.5 percent after the pandemic.

Initial cost-benefit analyses of the Covid-19 pandemic conclude that the benefits of containment probably outweigh the costs, with some economists advocating for a longer shutdown. Most Americans agree: More than 65 percent say it won’t be safe to gather in groups of 10 or more people until June or later.

The coming months will continue to be disruptive and intensive, but decisions must be based on data. Instead of arguing about past missteps or lamenting unrealistic alternatives, we should accept than an aggressive strategy is necessary and may be cost-effective. Then we can focus our energy on damage control; both for our health and our economic well-being.

Sandra McCoy is an associate professor in the division of epidemiology and biostatistics at the University of California Berkeley School of Public Health. She is trained as an infectious disease epidemiologist, and her research focuses on identifying innovative strategies to create demand for sexual and reproductive health services. Pia MacDonald is a senior epidemiologist at RTI International and adjunct associate professor at the University of North Carolina Gillings School of Global Public Health. Her research and work span infectious disease epidemiology and surveillance, global health security, public health preparedness, and outbreak investigation and response.