Biden’s newly released pandemic strategy is organized around a central goal: to oversee administration of 100 million vaccines in 100 days. To do it, he’ll have to fix the mess.
Some critics have called his plan too ambitious; others have said it’s not ambitious enough. It’s guaranteed to be an uphill battle. But before we get to the solutions, we need to understand how the system operates at the moment—and which aspects of it should be ditched, replaced, or retained.
From manufacturer to patient
At the federal level, two core systems sit between the vaccine factories and the clinics that will administer the shots: Tiberius, the Department of Health and Human Services’ vaccine allocation planning system, and VTrckS, the Centers for Disease Control and Prevention’s vaccine ordering portal.
Tiberius takes data from dozens of mismatched sources and turns it into usable information to help state and federal agencies plan distribution. VTrckS is where states actually order and distribute shots.
The two are eons apart technologically. Whereas Palantir built Tiberius last summer using the latest available technology, VTrckS is a legacy system that has passed through multiple vendors over its 10-year existence. The two are largely tied together by people downloading files from one and uploading them to the other.
Dozens of other private, local, state, and federal systems are involved in allocating, distributing, tracking, and administering vaccines. Here’s a step-by-step explanation of the process.
Step one: Manufacturers produce the vaccine
HHS receives regular production updates from Pfizer and Moderna. The manufacturers communicate estimated volumes in advance to help HHS plan before confirming real production numbers, which are piped into Tiberius.
Both vaccines are made of messenger RNA, a biotechnology that’s never been produced at scale before, and they need to be kept extremely cold until just before they go into a needle: Moderna’s must be kept at -25 to -15 °C, while Pfizer’s requires even lower temperatures of -80 to -60 °C. In the fall, it became clear that manufacturers had overestimated how quickly they could distribute doses, according to Deacon Maddox, Operation Warp Speed’s chief of plans, operations, and analytics and a former MIT fellow.
“Manufacturing, especially of a nascent biological product, is very difficult to predict,” he says. “You can try, and of course everybody wants to you try, because everybody wants to know exactly how much they’re going to get. But it’s impossible.”
This led to some of the first stumbles in the rollout. While training the states on how to use Tiberius, Operation Warp Speed entered those inflated estimates into a “sandbox” version of the software so states could model different distribution strategies for planning purposes. When those numbers didn’t pan out in reality, there was confusion and anger.
“At the end of December, people were saying, ‘We were told we were going to get this and they cut it back.’ That was all because we put notional numbers into the exercise side, and folks assumed that was what they were going to get,” says Maddox. “Allocation numbers are highly charged. People get very emotional.”
Step two: The federal government sets vaccine allocations
Every week, HHS officials look at production estimates and inventory numbers and decide on the “big number”—how many doses of each vaccine will go out to states and territories in total. Lately, they’ve been sticking to roughly 4.3 million per week, which they’ve found “allows us to get through lows in manufacturing, and save through highs,” Maddox says.
That number goes into Tiberius, which divvies up vaccines on the basis of Census data. Both HHS and media reports have sometimes described this step as using an algorithm in Tiberius. This should not be confused with any kind of machine learning. It’s just simple math based on the allocation policy, Maddox says.
Thus far, the policy has been to distribute vaccines according to each jurisdiction’s adult (18+) population. Maddox says the logic in Tiberius could easily be updated should Biden decide to do it on another basis, such as elderly (65+) population.
Once Operation Warp Speed analysts confirm the official allocation numbers, Tiberius pushes the figures to jurisdictions within their version of the software. An HHS employee then downloads the same numbers in a file and sends them to the CDC, where a technician manually uploads it to set order limits in VTrckS. (You can think of VTrckS as something like an online store: when health departments go to order vaccines, they can only add so many to their cart.)
Even that hasn’t been an exact science. Shortly before the inauguration, in a phone call with Connecticut governor Ned Lamont, outgoing HHS secretary Alex Azar promised to send the state 50,000 extra doses as a reward for administering vaccines efficiently. The doses arrived the next week.
The deal was representative of “the rather loose nature of the vaccine distribution process from the federal level,” Lamont’s press secretary, Max Reiss, told us in an email.
Step three: States and territories distribute the vaccine locally
State and territory officials learn how many vaccines they’ve been allotted through their own version of Tiberius, where they can model different distribution strategies.
Tiberius lets officials put data overlays on a map of their jurisdiction to help them plan, including Census data on where elderly people and health-care workers are clustered; the CDC’s so-called social vulnerability index of different zip codes, which estimates disaster preparedness on the basis of factors like poverty and transportation access; and data on hospitalizations and other case metrics from Palantir’s covid surveillance system, HHS Protect. They can also enter and view their own data to see where vaccination clinics and ultra-cold freezers are located, how many doses different sites have requested, and where vaccines have already gone.
Once states decide how many doses of each vaccine they want to send to each site, they download a file with addresses and dose numbers. They upload it into VTrckS, which transmits it to the CDC, which sends it to manufacturers.
Last week, Palantir rolled out a new “marketplace exchange” feature, effectively giving states the option to barter vaccines. Since the feds divvy up both Moderna and Pfizer vaccines without regard to how many ultra-cold freezers states have, rural states may need to trade their Pfizer allotment for another state’s Moderna shots, Maddox says.
When thinking about the utility of the system, it’s worth noting that many health departments have a shallow bench of tech-savvy employees who can easily navigate data-heavy systems.
“It’s a rare person who knows technology and the health side,” says Craig Newman, who researches health system interoperability at the Altarum Institute. “Now you throw in large-scale epidemiology…it’s really hard to see the entire thing from A to Z.”
Step four: Manufacturers ship the vaccines
Somehow, shipping millions of vaccines to 64 different jurisdictions at -70 °C is the easy part.
The CDC sends states’ orders to Pfizer and to Moderna’s distribution partner McKesson. Pfizer ships orders directly to sites by FedEx and UPS; Moderna’s vaccines go first to McKesson hubs, which then hand them off to FedEx and UPS for shipping.
Tracking information is sent to Tiberius for every shipment so HHS can keep tabs on how deliveries are going.
Step five: Local pharmacies and clinics administer the vaccine
At this point, things really start to break down.
With little federal guidance or money, jurisdictions are struggling with even the most basic requirements of mass immunization, including scheduling and keeping track of who’s been vaccinated.
Getting people into the clinic may intuitively seem easy, but it’s been a nightmare almost everywhere. Many hospital-based clinics are using their own systems; county and state clinics are using any number of public and private options, including Salesforce and Eventbrite. Online systems have become a huge stumbling block, especially for elderly people. Whenever jurisdictions set up hot lines for the technologically unsavvy, their call centers are immediately overwhelmed.
Even within states, different vaccination sites are all piecing together their own hodgepodge solutions. To record who’s getting vaccines, many states have retrofitted existing systems for tracking children’s immunizations. Agencies managing those systems were already stretched thin trying to piece together messy data sources.
It may not even be clear who’s in charge of allocating doses. Maddox described incidents when state officials contacted HHS to say their caps were too low in VTrckS, only to realize that someone else within their office had transferred doses to a federal program that distributes vaccines to long-term care homes, without telling other decision makers.
“Operation Warp Speed was an incredible effort to bring the vaccine to market quickly,” and get it to all 50 states, says Hana Schank, the director of strategy for public interest technology at the think tank New America. “All of that was done beautifully.” But, she says, the program paid little attention to how the vaccines would actually get to people.
Many doctors, frustrated by the rollout, agree with that sentiment.
These robots know when to ask for help
A new training model, dubbed “KnowNo,” aims to address this problem by teaching robots to ask for our help when orders are unclear. At the same time, it ensures they seek clarification only when necessary, minimizing needless back-and-forth. The result is a smart assistant that tries to make sure it understands what you want without bothering you too much.
Andy Zeng, a research scientist at Google DeepMind who helped develop the new technique, says that while robots can be powerful in many specific scenarios, they are often bad at generalized tasks that require common sense.
For example, when asked to bring you a Coke, the robot needs to first understand that it needs to go into the kitchen, look for the refrigerator, and open the fridge door. Conventionally, these smaller substeps had to be manually programmed, because otherwise the robot would not know that people usually keep their drinks in the kitchen.
That’s something large language models (LLMs) could help to fix, because they have a lot of common-sense knowledge baked in, says Zeng.
Now when the robot is asked to bring a Coke, an LLM, which has a generalized understanding of the world, can generate a step-by-step guide for the robot to follow.
The problem with LLMs, though, is that there’s no way to guarantee that their instructions are possible for the robot to execute. Maybe the person doesn’t have a refrigerator in the kitchen, or the fridge door handle is broken. In these situations, robots need to ask humans for help.
KnowNo makes that possible by combining large language models with statistical tools that quantify confidence levels.
When given an ambiguous instruction like “Put the bowl in the microwave,” KnowNo first generates multiple possible next actions using the language model. Then it creates a confidence score predicting the likelihood that each potential choice is the best one.
The Download: inside the first CRISPR treatment, and smarter robots
The news: A new robot training model, dubbed “KnowNo,” aims to teach robots to ask for our help when orders are unclear. At the same time, it ensures they seek clarification only when necessary, minimizing needless back-and-forth. The result is a smart assistant that tries to make sure it understands what you want without bothering you too much.
Why it matters: While robots can be powerful in many specific scenarios, they are often bad at generalized tasks that require common sense. That’s something large language models could help to fix, because they have a lot of common-sense knowledge baked in. Read the full story.
Medical microrobots that travel inside the body are (still) on their way
The human body is a labyrinth of vessels and tubing, full of barriers that are difficult to break through. That poses a serious hurdle for doctors. Illness is often caused by problems that are hard to visualize and difficult to access. But imagine if we could deploy armies of tiny robots into the body to do the job for us. They could break up hard-to-reach clots, deliver drugs to even the most inaccessible tumors, and even help guide embryos toward implantation.
We’ve been hearing about the use of tiny robots in medicine for years, maybe even decades. And they’re still not here. But experts are adamant that medical microbots are finally coming, and that they could be a game changer for a number of serious diseases. Read the full story.
5 things we didn’t put on our 2024 list of 10 Breakthrough Technologies
We haven’t always been right (RIP, Baxter), but we’ve often been early to spot important areas of progress (we put natural-language processing on our very first list in 2001; today this technology underpins large language models and generative AI tools like ChatGPT).
Every year, our reporters and editors nominate technologies that they think deserve a spot, and we spend weeks debating which ones should make the cut. Here are some of the technologies we didn’t pick this time—and why we’ve left them off, for now.
New drugs for Alzheimer’s disease
Alzmeiher’s patients have long lacked treatment options. Several new drugs have now been proved to slow cognitive decline, albeit modestly, by clearing out harmful plaques in the brain. In July, the FDA approved Leqembi by Eisai and Biogen, and Eli Lilly’s donanemab could soon be next. But the drugs come with serious side effects, including brain swelling and bleeding, which can be fatal in some cases. Plus, they’re hard to administer—patients receive doses via an IV and must receive regular MRIs to check for brain swelling. These drawbacks gave us pause.
Sustainable aviation fuel
Alternative jet fuels made from cooking oil, leftover animal fats, or agricultural waste could reduce emissions from flying. They have been in development for years, and scientists are making steady progress, with several recent demonstration flights. But production and use will need to ramp up significantly for these fuels to make a meaningful climate impact. While they do look promising, there wasn’t a key moment or “breakthrough” that merited a spot for sustainable aviation fuels on this year’s list.
One way to counteract global warming could be to release particles into the stratosphere that reflect the sun’s energy and cool the planet. That idea is highly controversial within the scientific community, but a few researchers and companies have begun exploring whether it’s possible by launching a series of small-scale high-flying tests. One such launch prompted Mexico to ban solar geoengineering experiments earlier this year. It’s not really clear where geoengineering will go from here or whether these early efforts will stall out. Amid that uncertainty, we decided to hold off for now.