Close

These Technologists Are Trying to Make COVID-19 Risk Assessment More of a Science

The microCOVID Project crunches your location and desired activity into a real-world risk estimate

8 min read
Illustration showing charts, covid icons and callouts.
iStockphoto

During the course of the pandemic, public-health authorities have tried the best they could to provide guidelines about what is or isn’t safe. That’s a tough job, for many reasons, not the least of which is that scientific understanding of the COVID-19 virus and its transmission is constantly evolving, as are conditions on the ground. What’s more, the notion that activities can be categorized as either definitely safe or definitely dangerous is, obviously, an oversimplification. How then do you decide what’s okay to do?

Fortunately, a group of technically minded volunteers that calls itself the microCOVID Project offers a framework within which to make science-based decisions about appropriate daily behavior in the light of the risk each poses for becoming infected with SARS-CoV-2. The group’s online tools take into account enough detail to make those assessments meaningful while not being overwhelming in terms of the amount of input you have to provide.

One of the contributors to the micoCOVID project is Benjamin Shaya, who was willing to discuss the origin and evolution of this project with Spectrum.

IEEE Spectrum: Could you take a moment to outline what the microCOVID calculator is?

Shaya: The microCOVID calculator is a single-page web tool where you can enter details about an activity or job or prolonged encounter with a person—such as living with them or dating them—and get an estimate of how likely you are to get COVID from that interaction.

How many users does the microCOVID calculator have?

We get about a 2,000 page views a day from about 1,400 unique visitors. But usage spiked to about double that number around Thanksgiving.

And people would use this calculator in the framework of some sort of risk budget, is that right? 

Yes. The microCOVID calculator gives you a number—the number of microCOVIDs, where one microCOVID is one-millionth of a chance of getting COVID. To make a decision, you frame the results in terms of your budget: how many microCOVIDs you are willing to accumulate per week, per year, per lifetime, similar to the way that you would budget with dollars.

Tell me about yourself.

I'm a software engineer by trade and an electrical engineer by training. Now I work for Google, which is where I spent most of my career.

Can you also tell me about your collaborators in the microCOVID project?

The original team is a group six people who lived together in a house called “Ibasho” in San Francisco’s mission district. It’s a reference to Ada Palmer’s sci-fi novel Too Like the Lightning [Tor Books, 2016]. In that book, people live with the people who they choose to be their family, not their biological family.

Even before shelter-in-place started, this group was on top of news coming out of Wuhan. Some of them are good friends with people who are work in biosecurity who were sounding the alarm as early as January and February of 2020 that something was going to happen. If this hits the United States, they realized, we're going to lock down. 

Then in March of 2020 they locked down, and they thought, “Well, this sucks. We want to see our friends. We want to see our partners.”

For the next four months, they were basically negotiating everything that folks could do, and they came up with this pod that included myself and a few other people, which allowed us to see a small number of people who didn't live in the house. It came with a bunch of rules, for example, “Don't go on more than this many walks with this many people” and “Report if you're feeling a fever.”

You make all these rules, and then everyone asks for an exception: Can I see a dentist? Can I get a haircut? Can the housemate who does not have a partner go on a date with a new person?

In what way were these people qualified to determine the appropriate rules to keep them all safe?

By any standard qualification, they weren’t. But to give you a little context about this group, one of them is a medical doctor, a primary-care provider. One of them does research on AI and has a master's in neuroscience. So she has the chops to go in and read papers and cut through what's nonsense. 

Were those initial rules then codified and sent around or put on a white board or something?

Yeah, they were put into a Google Doc. Because they were so early on this, their Google Doc got copied and spread all over the community. I've looked into joining group houses, and the people there said, “Yes, here’s our agreement.” It was a copy of the Ibasho’s Google Doc. 

But you and they worked on more than just a set of rules. The framework you created gives everyone a certain risk budget to spend. How did that kind of thinking about a risk budget arise?

That started with Catherine Olsson, an AI-safety researcher who does a lot of thinking around existential risks to humanity. Perhaps you've heard of micromorts, a system devised in 1980 at General Motors Research Laboratories for considering how risky something is in terms of how likely is it to kill you. Such schemes are helpful, because most people are phenomenally bad at estimating risks.

We can tell you with fairly good confidence how likely are you to get COVID from this or that interaction. What we can’t tell you is how much risk should you tolerate.

A lot of us were fairly numerically minded, and the response of the community—and even that of some people within the house—didn't feel proportional to the risk of COVID. Some were feeling, “I must avoid COVID at all costs,” which is intuitively wrong, in the same way that avoiding at all costs anything that involves some chance of death isn't consistent with a lifestyle that involves driving a car—or even with living, because living involves a chance of death.

So at this point someone said, “We need to come up with some quantitative tools to evaluate COVID-19 risk,” is that it?

Catherine and Josh, who is a software engineer at a FinTech company, went into deep-research mode for two months. They borrowed heavily from the aerosol work by a chemistry professor at the University of Colorado, Boulder, Jose-Luis Jimenez. A lot of the model for aerosols came from that.

I'm really glad that you bring up like this risk tolerance because it's one of the most contentious numbers in the whole project. We can tell you with fairly good confidence how likely are you to get COVID from this or that interaction. What we can’t tell you is how much risk should you tolerate. But ultimately, members of this household decided to use a 1 percent chance of getting COVID per year as the tolerance that they were going to hold one another to.

For an average person between 16 and 49 years old, that matches the risk of long-term health consequences from COVID, as we understood the risks in July of 2020, to the risk of having long-term health impacts from driving the average amount an American drives in a year. 

Is this about the time that you are other programmers got involved?

The group of six was tempted to just kind of leave it there and use the system for themselves. But they soon realized that if they want to use this to hang out with other people, those other people needed, at least, to buy into the idea. To see people outside their group of six, they needed both to be safe and to convince other people to hang out with them.

So they started writing up a whitepaper. At this point, Sarah, the doctor in the group said, “Hey, I want to show this to my patients.” That started the ball rolling in getting this work out into their immediate community. The Ibashoans are all really community minded people, and I think hearing about their friends' suffering and isolation was really painful to them.

Sarah's partner started the website, and I happened to be staying at Ibosho for a week—the week before they were intending to go live. Everyone was heads down on their laptops, either writing code, finishing off some research, or editing the whitepaper.

I thought to myself, “Well, I'm here. I might as well help out.” I'd learned React last summer, so I got involved. Just before launch, we started moving everything from an ad-hoc Facebook-messenger based system into GitHub and Slack. We took on a few more people who started once we launched, people posting GitHub issues or volunteering to do some work. And we started adding those people to the Slack and GitHub team.

When did this become a public tool for Sarah’s patients to use? 

We had some friends of the house beta testing, but it was the end of August that we officially went live.

Can you tell me about the evolution of the project since then?

Some of original six declared their part done after the original launch. But Catherine stayed on as kind of director of research until March. There was research on air purifiers that was coming out in the fall. We had to scramble in January to catch up with B.1.1.7 variant, which was making headlines. And I think we were one of the first organizations to quantify how effective we thought vaccines would be at reducing transmission, also taking data about case reductions and then combining that with models that other researchers had produced for relative risk of transmission from asymptomatic versus symptomatic carriers. That came out in February. And then we followed up when Johnson & Johnson released their data and researchers published a big study on the efficacy of vaccines, based on people in Israel, which let us refine things further.

But I see you haven’t yet integrated results from an even bigger study on vaccine efficacy that also uses Israeli data, one published a few weeks ago in The Lancet. Why is that?

The Risk Tracker calculates not only how much risk you’ve accumulated over the week or the year, but it also uses a model for when symptoms appear to help you predict how dangerous you are to other people. Within our community, that level of riskiness became almost a currency.

Model updates are big lifts for us, and the team has definitely wound down their involvement in the last month or so. We’re all vaccinated at this point. And the general feeling in the United States is that once you're vaccinated, you can go back to living an almost normal life, which has taken a lot of the urgency out of doing work on the project.

But I took today off of work, and I'm doing a data ingest of a vaccine data. That’s necessary because we've been bombarded with requests to be able to calculate what the risk profile is for a random vaccinated person in a given area. And that's impossible to do without knowing how many people in the region are vaccinated.

Could you describe the other tool you offer on the website, the Risk Tracker?

The Risk Tracker is the original form of microCOVID calculator. It takes the budget idea to the next level: You have a budget, so you can spend more than one week’s worth of that budget in a week so long as you don’t outspend your budget over the course of a year.

The tracker allows you to enter your activities into it. And it calculates not only how much risk have you accumulated over the week or the year, but it also uses a model for when symptoms appear to help you predict how dangerous you are to other people. Within our community, that level of riskiness became almost a currency.

Has anyone else produced similar tools?

In developing this, we looked at was a tool called COVID: Can I do it?, which has a beautifully slick site with thousands, if not millions, of activities. It rates on a five-point scale how safe the activity is. Our main critique of that was it doesn't take into account how prevalent COVID is in your area. But otherwise I think it’s a wonderful tool.

You said you are winding down because you’re all now vaccinated. But people with little kids are still very much trying to figure out what's safe to do.

I’ll totally admit that’s a result of none of us being parents. There’s a lot about child-transmission dynamics that isn't really represented in the tool. There’s just lot for someone to sift through, to come up with something really comprehensive about what children should and shouldn’t be doing.

And there’s the ongoing question of the consequences if you get COVID but you don't get serious COVID. That's kind of the specter in the room. Very few papers that we’ve found do a really detailed analysis of the risks associated with mild COVID or how bad mild COVID might be for children.

I want to thank you because your microCOVID calculator has been very helpful for my own family. We’re all vaccinated now, and the risks are abating. But if down the road there's a breakout variant for which the vaccine is less effective, we have a great tool to use to help navigate the future.

Thanks for saying that.

The Conversation (0)

This CAD Program Can Design New Organisms

Genetic engineers have a powerful new tool to write and edit DNA code

11 min read
A photo showing machinery in a lab

Foundries such as the Edinburgh Genome Foundry assemble fragments of synthetic DNA and send them to labs for testing in cells.

Edinburgh Genome Foundry, University of Edinburgh

In the next decade, medical science may finally advance cures for some of the most complex diseases that plague humanity. Many diseases are caused by mutations in the human genome, which can either be inherited from our parents (such as in cystic fibrosis), or acquired during life, such as most types of cancer. For some of these conditions, medical researchers have identified the exact mutations that lead to disease; but in many more, they're still seeking answers. And without understanding the cause of a problem, it's pretty tough to find a cure.

We believe that a key enabling technology in this quest is a computer-aided design (CAD) program for genome editing, which our organization is launching this week at the Genome Project-write (GP-write) conference.

With this CAD program, medical researchers will be able to quickly design hundreds of different genomes with any combination of mutations and send the genetic code to a company that manufactures strings of DNA. Those fragments of synthesized DNA can then be sent to a foundry for assembly, and finally to a lab where the designed genomes can be tested in cells. Based on how the cells grow, researchers can use the CAD program to iterate with a new batch of redesigned genomes, sharing data for collaborative efforts. Enabling fast redesign of thousands of variants can only be achieved through automation; at that scale, researchers just might identify the combinations of mutations that are causing genetic diseases. This is the first critical R&D step toward finding cures.

Keep Reading ↓ Show less