COVID-19’s toll reveals need for fast detection, response
US set to begin development of outbreak warning systems
Shaken by the death toll and economic devastation wrought by COVID-19, the U.S. is poised to begin developing early warning systems to detect disease outbreaks, make forecasts and slow their spread to ensure they do not become pandemics.
President Joe Biden’s COVID-19 plan includes establishing a National Center for Epidemic Forecasting and Outbreak Analytics.
Just two weeks ago, Cleveland Clinic and IBM announced plans to team up on a massive Discovery Accelerator that will build better models of diseases and speed the process of advancing promising treatments from the lab to the patient.
IBM will power the project by contributing an advanced quantum computer, which can read the genetic
scripts of viruses and patients a thousand times faster than previous computers.
Asked about the president’s plan to establish a new forecasting center, Lara Jehi, a neurologist at Cleveland Clinic’s Lerner College of Medicine, said, “He’s describing what we’re building, platforms that align research . ... The root problem is how slow it currently is to get a discovery in the lab and turn it into a treatment. Right now, that interval is about 17 years.”
The exception was Operation Warp Speed, the program launched by former President Donald Trump that produced extremely effective vaccines less than a year after China announced the first COVID-19 cases. The Warp Speed program, however, was hampered in December and January by poor planning and distribution of the vaccines.
Jehi stressed that the Discovery Accelerator and President Biden’s proposed epidemic forecasting center should be able to co-exist and tackle problems without duplicating each other’s efforts.
“If we have learned anything from the pandemic, it is that it takes big teams looking at problems from so many different angles because you really don’t know which one will work,” she said. “I think the days of people working separately are over.”
Rasmus Bech Hansen, founder and chief executive officer of the London-based science data-gathering company Airfinity, put it this way:
“Nobody can say they are an expert in COVID-19. They are an expert in one little area. To really understand COVID-19, you have to be somewhat of a policy expert and know all of the social effects. You need to be an epidemiology expert and understand how pandemics work. You need to be an expert in vaccines. You need to be an expert in virology. It is really important that you be an expert in genomics and testing.”
Although some scientists have compared the forecasting of outbreaks to the forecasting of weather, others say making predictions about infectious diseases is a far more complex task.
“I don’t know if weather forecasting is a useful analogy,” said Ricardo Castillo-Neyra, who co-directs the Zoonotic Disease Research Lab, a collaboration between the University of Pennsylvania where he works and a university in Lima, Peru. “Predictions are very good for weather forecasts. It’s not that simple for infectious diseases. The number of variables is much greater.”
Start with the sheer number of potential pathogens. A 2011 editorial in Nature said that if all the viruses on earth “were laid end to end they would stretch for 100 million light years. Furthermore, there are 100 million times as many bacteria in the ocean as there are stars in the known universe.”
Disease forecasters must then take into account the movements of people and disease-carrying insects such as mosquitoes, the effects of climate change and the nature of the pathogens themselves.
Forecasts must also account for social and cultural factors. Rabies from dogs, for example, is almost unheard of in the U.S., where most people keep their pets inside at night. However, dog rabies is far more common in parts of Latin America and Africa, where dogs are allowed to roam freely during the day and are kept outside for security at night.
An even more basic problem was brought into sharp relief by the new coronavirus that causes COVID-19.
“I don’t know of a field that has ever predicted something that had never existed before,” said Jeffrey Shaman, a professor of environmental health sciences at Columbia University’s Mailman School of Public Health.
Although thousands of researchers study bats, mosquitoes and numerous animal species that spread or act as habitats for viruses and bacteria, it is extremely difficult, if not impossible, to guess which pathogens will jump to humans, as the new coronavirus did.
Nonetheless, it may be possible to detect new diseases in humans much faster than ever before, according to Hansen at Airfinity. “I think the key thing the world is missing is better genomic sequencing.”
Hansen suggested that countries could have begun analyzing the genetic scripts of their populations as soon as China announced the appearance of COVID-19. Random sequencing of people might have picked up the disease’s signature long before health officials were reporting cases and hospitals were becoming overwhelmed.
Sequencing was once enormously expensive and time-consuming. (The Human Genome Project, completed in 2003, took 13 years and cost $3 billion). Today, though, genomes can be sequenced for under $1,000, and a Chinese firm recently claimed to have reduced the cost to $100.
That raises the possibility of what Hansen called “preventative testing,” random sequencing of portions of the population in order to pick up illnesses even before a hospital can connect patients with similar symptoms to a new disease.
“I think about it a little like what we have with earthquakes,” Hansen said. “There are these listening systems that monitor the ground and underground to determine when an earthquake is coming.”
While the new forecasting centers will undoubtedly make use of the latest genetic technology, it may be years before they employ random sequencing of the population.
Instead, centers would start off by seeking the first signs of an outbreak in the mass of data flowing from hospitals and governments around the globe.
The first warning could be as simple as a handful of unusual pneumonia or flu cases that don’t fit the known diseases of an area.
“What you want to see is if this is something that’s above the baseline,” said Nasia Safdar, medical director of infection control at UW Health. “A new virus never seen before that’s potentially contagious.”
Global cooperation, communication
To improve the speed and accuracy of forecasting, centers will need to establish extensive lines of communication. The Centers for Disease Control and Prevention and World Health Organization have systems in place, but experts say much more is needed.
“It’s an enormous undertaking,” Safdar said. “If there is a country that is battling an outbreak, but doesn’t have the resources, there has to be some sort of international cooperation to share that kind of information.”
Health experts have stressed the importance of monitoring so-called hot spots for new diseases, including areas of Southeast Asia and Africa.
“There’s another issue that nobody talks about,” said Castillo-Neyra at the University of Pennsylvania.
“In the Amazon, poultry companies are building factories. That’s dangerous because you have wild animals in contact with poultry. Many villages have no sewage system. Any latrine they have just goes to the river. You have thousands of animals in very small spaces and they produce so much manure it’s very hard to contain it. Also, when you have thousands of animals in small spaces they are going to have pathogens.”
The next pandemic could start on a large farm. Or it could begin in a crowded, impoverished city such as Nairobi, Kenya, where people and animals live in close proximity.
At Airfinity, Hansen said disease forecasting centers should develop algorithms that scour social media sites such as Facebook for clusters of terms coming from one geographical area. Phrases such as “persistent cough,” “trouble breathing” or “nasty flu,” all coming from the same part of the world, may be the first tipoff of an outbreak.
Once a disease has been detected, computer models can be designed to predict where and how it might spread. Ideally, governments can then translate those predictions into clear messages about mask-wearing, social-distancing and other actions people can take to protect themselves and others.
At Columbia, Shaman has developed rolling models that track the flu season and adjust week to week as new flu numbers become available. But predicting the shifts of a known, recurring disease such as influenza is not nearly as difficult as making predictions about a novel disease such as COVID-19.
“It’s because it’s happening all the time that it’s been predictable,” Shaman said of flu. “You have a system that’s stationary where the rules of the game are not changing so much.”
Health leaders send similar messages about the flu each year and people tend to respond in the same way.
COVID-19 predictions were complicated by the decisions of individual states to either shut down or reopen, employ mask mandates or lift them. In addition, some people have changed their behaviors as the pandemic has worn on, giving up on social distancing and mask wearing.
Shaman stressed that the president’s proposed epidemic forecasting center would be best off using a variety of models.
“What we’ve seen over and over again is that an ensemble of models performs better than a single model,” he said. “But you don’t just throw all of the models together and average them. The aim would be to establish some kind of accountability. You see which models perform better and you give those models more weight.”
At Cleveland Clinic, Jehi said the IBM quantum computer will speed up the critical work of responding to a new disease.
“We cannot waste a year developing tests to diagnose the next pathogen,” she said.