How to tap AI’s potential while avoiding its pitfalls

Banks’ climate pledges don’t add up to change, research finds

The impact of misleading headlines on Facebook

Credit: Martin Adams/Unsplash

Ideas Made to Matter

Health Care

COVID-19 cases are 12 times higher than reported


Consider the U.K.: population 66 million, 314,000 confirmed cases of COVID-19, and 44,000 deaths. Now consider Germany: 80 million people, 195,000 confirmed cases, 9,000 deaths (as of July 1).

Notable differences.

But what accounts for these disparities? Why have COVID-19 infection rates and fatalities differed so greatly between otherwise similar countries? The reasons are complicated, dependent on connections between a sweeping range of variables, from testing and individual behavior to hospital capacity and social policy. 

A new paper by MIT Sloan system dynamics professors and, with PhD candidate Tse Yang Lim, examines many of these relationships to give a global view of the pandemic. (The paper is available as a preprint and not yet been peer reviewed.)

Among the findings: Infections are 12 times higher and deaths 50% higher than previously reported, but herd immunity is “not a viable path out of the current pandemic,” Rahmandad said. And the value of testing erodes as the virus spreads, leaving measures like social distancing, better hygiene, and wearing masks the most useful way to contain the coronavirus.

The model the researchers created captures the feedbacks that either fuel or mitigate the spread of COVID-19. For instance, more testing increases the proportion of infections that are identified, and may motivate people to cut their contacts with others, driving down overall infections and serious cases. Hospitals would be more likely to have sufficient capacity, lowering the death rate.

But if the initial success of testing and distancing reduces caseload and deaths, people may not consider COVID-19 to be a serious risk, undermining distancing guidelines and mask use, driving infection rates up.  Hospitals might be overwhelmed, leading to more avoidable deaths, as happened in New York City in April, and in Houston in the summer. The interplay between these and other variables dynamically affects the size and shape of a given outbreak.

Using a simulation that accounted for all of these connections, the researchers collected data from 84 countries that, in total, account for more than 60% of the world’s population. They then calibrated the model with country-level data on testing, reported cases, reported deaths, excess overall mortality, and other information.

“Our model simulates the progression and spread of COVID-19, including how people interact, how many get sick, how many get tested, how many are hospitalized, how many die — and how people change their behavior in response to the risk they perceive,” Rahmandad said. “We then use a wide range of data to estimate the parameters of the model — say, what fraction of infections are asymptomatic, and how contagious the virus is —  to give the best match to the real world data.”

The model revealed several important insights. Most fundamentally, the magnitude of the epidemic is widely underreported. The researchers estimate that cases and deaths through June 18 are, respectively, 11.8 and 1.48 times higher than official reports across the 84 nations considered. (Recent data from the Centers for Disease Control and Prevention corroborates this finding.) Despite these elevated numbers, the authors note that no country is remotely close to establishing herd immunity.

The researchers were also able to estimate the Infection Fatality Rate — the probability of death after becoming infected — across nations and over time. That rate depends not only on the age and health of the population, but on the adequacy of health care and the effectiveness of protections for the most vulnerable, including the elderly, poor, and minorities, who often live in more crowded conditions, are less able to work from home, and are less likely to have access to good health care and hospitals. The researchers estimate IFR to be 0.68% on average, but find it varies substantially across nations: approximately 0.56% for Iceland, 0.64% for New Zealand, 0.99% for the United States, 1.59% for the UK, and 2.08% for Italy.

The model also illuminates how essential early and aggressive testing is in suppressing the virus and, ultimately, controlling the fatality rate. Because the contagion grows exponentially, even a few days’ delay can create tremendous differences in the size of the epidemic. If countries in mid-March had moved to test 0.1% of their total population per day — a rate higher than most countries at the time — total case count in the countries studied would have dropped from 88.5 million to 53 million, as of mid-June, and deaths would have fallen by nearly 200,000 — about one third of the estimated total.

However, the value of testing erodes as the epidemic grows, as the increasing number of infectious individuals overwhelms the ability to test, then trace and isolate those coming in contact with the infectious. And testing has little impact if people who come into contact with those testing positive don’t isolate themselves.

“The key driver of future trajectories is the extent of distancing, better hygiene, and mask use,” Rahmandad said. “Communities that bring down risky contacts quickly when the case counts rise are likely to see far fewer cases in the long run.”

In fact, if countries maintain high (rather than low) sensitivity to the public health risks of COVID-19 — avoiding premature return to normal activities and social interactions — cumulative cases in spring of 2021 could drop from 1.5 billion to 150 million; a full order of magnitude even without effective treatments or a vaccine.

The model provides quantitative insight into the potential infection and mortality rates of COVID-19 given different social and demographic contexts. It also makes clear current numbers grossly underestimate the spread of the virus. But perhaps most central, behavioral and social responses to COVID-19 are critical to reducing future cases. In a setting with more lenient policies, the virus will spread.

“While actual cases are far greater than official reports suggest, the majority of people remain susceptible. Waiting for herd immunity is not a viable path out of the current pandemic,” Rahmandad said. “Every community needs to keep the pandemic under control until a vaccine or treatment is widely available. A slow and half-hearted response only increases the human costs without offering much of an upside in terms of economic output.”

For more info Zach Church Editorial & Digital Media Director (617) 324-0804