How to Fix the Vaccine Rollout

At a moment when vaccines promise to end the coronavirus pandemic, emerging new variants threaten to accelerate it. The astonishingly fast development of safe and effective vaccines is being stymied by the glacial pace of actual vaccinations while 3,000 Americans die each day.

Minimizing death and suffering from COVID-19 requires vaccinating the most vulnerable Americans first and fast, but the vaccine rollout has been slow and inequitable. Prioritization algorithms have led to the most privileged being prioritized over the most exposed, and strict adherence to priority pyramids has been disastrously slow. Yet without prioritization, vaccines go to those with greatest resources rather than to those at greatest risk.

Algorithmic thinking can reduce chaos and slow-downs, identify gaps between what we are doing and need to do

As a new administration takes office, there is hope that the vaccination rollout will be infused with both common sense and urgency. But to best protect lives, livelihoods, and health, we need not just speed, but also a quantitative accounting of who is at risk and how to best save them.

As a computational biologist, I study scalable distribution systems in biology,¹ and I design algorithms for distributed robot systems that are fast, flexible, and robust.^2,3 The systems we build have important lessons for building a fast and effective distribution system for vaccines.

To beat COVID-19, we don’t need algorithms but we do need algorithmic thinking to effectively scale distribution of vaccines to 300 million Americans. This requires us to i) state our goals clearly, ii) provide step-by-step instructions, actions, and timelines for meeting those goals, and iii) test against quantitative benchmarks in order to adapt our actions to continuously move toward our goals. Algorithmic thinking can reduce chaos and slow-downs, identify gaps between what we are doing and what we need to do, and reveal when we need other strategies to augment vaccination in order to achieve our goals.

Algorithmic thinking leads to the following principles for a vaccine strategy.

A. Parallelization: Like algorithms, vaccination pipelines run fastest when they are run in parallel. Vaccinate multiple high priority groups at once.

B. Exponential growth in vaccinations: Linear solutions will never solve exponential problems. We must simultaneously slow the exponential growth of the virus and exponentially grow our vaccination rollout.

C. Match supply and demand: We need constant vigilance to accelerate whatever is slowing down vaccinations: the vaccine supply, the demand (or willingness) of the population to be vaccinated, and the pipelines that move vaccines from warehouses into arms.

Algorithms need goals, so first we must provide a quantitative description of our goals.

Goal #1: Slow the exponential growth of the virus

We must buy more time to vaccinate before new variants accelerate infections and deaths. The CDC predicts the new United Kingdom variant will become dominant in the United States by mid March.⁴ It is 50 percent more contagious and replicates weekly. If unchecked, it would generate five times faster spread each month leading to 1 million daily positive tests and 15,000 daily deaths. The vaccine supply is too small to appreciably slow viral growth before this catastrophic increase in spread. Vaccines are our best long-term solution, but in the short term we must ramp up every social distancing tool we have to slow viral growth.

We need not just speed but a quantitative accounting of who is at risk and how to best save them.

Goal #2: Vaccinate as fast as possible the populations most likely to die

U.S. deaths are highly concentrated in the very oldest age groups. Eighty percent of deaths are in those over 65. Vaccinating those 75 and older (who are four times more likely to die than those 65-74) would save 48,000 lives in 100 days given current daily death rates.⁵ However, racial minorities with younger populations and shorter life expectancies are a tiny fraction of those over 75. African-, Native- and Hispanic-Americans aged 25-54 have lost over seven times more years of life from COVID-19 than white Americans.⁶ Prioritizing vaccinations among minority populations saves more people in the prime of their lives.

Goal #3: Reduce the total number of infections, even in young and healthy populations

An estimated 10 percent of COVID-19 cases result in at least several months of physical and mental disability.⁷ Unlike severe disease, long-COVID poses a significant risk across age groups. We lack precise estimates on its prevalence, but if only 10 percent of these long-COVID cases lead to lifelong disability, over 2,000 people are doomed to this fate every day. We must speed up vaccinations of those most at risk of exposure (frontline workers, prisoners, students), as well as the young who have the most to lose from a lifelong disability.

Goal #4: End the pandemic as fast as possible

As long as the virus is uncontrolled, hundreds of millions of us will continue to suffer from isolation and the consequences of a partially hunkered down economy, and we will be at risk from emerging new variants that spread faster and may eventually escape vaccinations.

The three algorithmic principles show how we can achieve these goals.

A) Vaccinate multiple priority groups in parallel

Computer algorithms run faster when we break problems into separate pieces and solve them simultaneously. Current plans to vaccinate many groups simultaneously will similarly increase vaccination speed (Goals #3 and #4). However, separate parallel efforts to vaccinate different groups can also give prioritized access for the most vulnerable (Goal #2).

With this acceleration, everyone in America could have their first shot by the Fourth of July.

We can protect the most vulnerable first (saving tens of thousands of lives) and also increase vaccination speed for everyone (reducing infections in the long and short term). Not only do we need pharmacies, health departments, church parking lots, and stadiums as venues for vaccinating multiple populations in multiple places, we must have faster access and tailored deliveries for those who are most vulnerable and hardest to reach. Access to vaccines in vulnerable neighborhoods is necessary but woefully inadequate; even prioritizing frontline workers has left more vulnerable racial groups less vaccinated. We must have high priority, easy access lines (like Walt Disney World FastPass) for those over 75, those who live in highly vulnerable zip codes, and those in frontline essential jobs.

Even while supply is limited we must ramp up at-home vaccinations for those who cannot travel and reach out to those unable to access online registrations. We must monitor who gets vaccinated, so we know if the line is moving slowest in poor and rural areas and then can provide support to counter the lack of pharmacies, doctors, and trust in the least resourced communities. Long before any priority group is fully vaccinated, new pipelines need to be established and tested to serve the next priority population. We must be prepared to shift; for example, if schools become a major source of spread of new variants, then children may need to move up in the priority list. If a regional outbreak overwhelms hospitals, more vaccinations may need to shift to that location. If vaccines are better than expected at blocking transmission, we may need to prioritize vaccinations for the most socially connected (potential superspreaders) along with the most vulnerable. We need to be ready to hit the accelerator on different vaccine pipelines as the need arises.

B) Fight an exponential problem with exponential solutions

Imagine you are paying a fixed amount on your credit card each month. If you are chipping away at the principal, then the interest payment also gets smaller over time. If suddenly your interest rate goes up 50 percent, but you continue the same fixed payments, then you get further in debt with each payment you make. We are already in deep COVID-19 debt. And we can see bearing down on us a 50 percent increase from the new COVID-19 variants. We must increase our prevention methods now with increased testing, high-quality masks, quarantining, and restrictions on travel and gathering to buy time for longer term vaccination strategies to work. Every week we delay, our job gets 50 percent harder and more people unnecessarily die.

While we initially slow viral growth (Goal #1) through social distancing, we ultimately win by exponentially growing vaccinations. A 50 percent increase in vaccinations each month can catch up to and then dampen viral growth. A linear plan for 100 million vaccines in 100 days is not enough. But, if we vaccinate 1 million per day by the end of January, 1.5 million per day in February and 2.25 million per day in March, then nearly 100 million of the most vulnerable would have at least their first shots (and some their second) by the end of March, saving lives and slowing viral spread as the new variant takes hold. With this acceleration, everyone in America (300 million people, 500 million shots) could be vaccinated by the Fourth of July.⁸ This massive vaccination effort will prevent tens of millions of infections, particularly those in the last priority groups who may be least vulnerable to death but most impacted by long-term debilitating effects of COVID-19.

Enlist every vaccinated grandparent to post Facebook pictures of hugs with their grandchildren.

Will we have the vaccine supply to pull this off? It is absolutely possible. If our capitalist system of production is good for anything, it is exponential growth. The richest man in the world made many more Teslas in 2020 as 2010. We can use the same strategies that created exponential growth in computing, cell phones, and electric cars to ramp up vaccine production.

Before July, we are scheduled to have 300 million doses of approved mRNA vaccines and 200 million doses of vaccines that should be approved in the coming weeks. But this supply can only get into arms quickly if we have ramped up our delivery capacity to take advantage of the burst of production in the spring.

The acceleration strategy must set aggressive targets and then adapt to meet them. If production goals aren’t met, the supply could be stretched by over 100 million more doses if vaccine trials validate single doses, delayed second doses, or half doses of vaccines for younger age groups who will be vaccinated in later months.

C) Rapid vaccination requires matching supply and demand with an adaptive pipeline

With much of the current supply sitting in freezers, it is now clear that we must exponentially grow not just supply, but also pipelines that move vaccines from factories to the “last mile” (vaccination clinics) and the “last inch” (people’s arms). Parallelization sets up multiple vaccination streams, and we must increase their combined capacity by 50 percent each month. Every venue that successfully vaccinates in January should be copied to vaccinate more people in February. If a clinic is open 12 hours per day in February, open it 24/7 in March. If there is a lack of vaccinators in March, recruit medical students, dentists, veterinarians, and anyone who can give a shot, and call in the National Guard in every state by April. Federal advice, money, coordination, and accurate communication of vaccine availability will help, but we also need every public health worker to be empowered to act with urgency to pull from the supply as fast as possible. Governors must simultaneously aid their pharmacies, hospitals, and clinics to pull faster, while demanding a faster push from the Federal supply.

In the long run, our most difficult problem may be demand. The initial vaccination allocations were slowed in part because health care workers did not want them.⁹ This problem will appear to wane as vaccination programs open up to larger groups, but it will persist and reappear when we run out of people willing to be vaccinated in the spring. An untrustworthy system has created many who are understandably vaccine hesitant, while online disinformation campaigns are fueled by conspiracies and distrust.

We must engage community leaders, churches, activists, doctors, and people in every neighborhood in America as well as celebrities of every stripe to get truthful vaccine messages out to the public. Social media is an exponential accelerator that must counter online disinformation with coordinated messaging in multiple parallel domains: public health messages on the nightly news targeting the elderly, and TikTok, YouTube, and Twitter influencers speaking to the young. Enlist every vaccinated grandparent to post Facebook pictures of hugs with their grandchildren. Replicate successful “get out the vote” campaigns to “get out the vax.” Flood the airways with marketing campaigns to galvanize us all to do our part so we can celebrate freedom from the virus this Fourth of July.

The movie War Games offers a final important algorithmic lesson. Its fictional algorithm faced repeated defeats and concluded “sometimes the only winning move is not to play.” As long as we repeatedly let the virus come back stronger each time we knock it back, we lose. We must use the ramped-up vaccine production capacity we will have developed by summer to produce enough vaccines to beat the virus at a global scale. The WHO reports, “More than 39 million doses of vaccine have now been administered in at least 49 higher-income countries. Just 25 doses have been given in one lowest-income country. Not 25 million; not 25,000; just 25.” Allow international licensing of vaccines to aid efforts to parallelize and exponentially grow global vaccine capacity. Not only can the U.S. help global efforts to prevent suffering and death around the world, we can build a massive vaccine production capacity to combat new variants that may escape our first round of vaccinations. We do not want to beat the virus in 2021 just to import a faster-spreading, vaccine-resistant strain in 2022. Only through a massive global vaccination campaign can we end this game.

Melanie E. Moses is a professor of computer science and biology at the University of New Mexico and an external faculty member of the Santa Fe Institute. She builds computer models of how the immune system responds to viral lung infections and designs scalable swarms of robots using bio-inspired algorithms.

References

1. Banavar, J.R., et al. A general basis for quarter-power scaling in animals. Proceedings of the National Academy of Sciences 107, 15816-158120 (2010).

2. Lu, Q. , Hecker, J.P., & Moses, M.E. Multiple-place swarm foraging with dynamic depots. Autonomous Robots 42, 909-926 (2018).

3. Lu, Q., Fricke, G.M., Tsuno, T., & Moses, M.E. A bio-inspired transportation network for scalable swarm foraging. 2020 IEEE International Conference on Robotics and Automation (ICRA), Paris, France, (2020).

4. Mandavilli, A. & Rabin, R.C. CDC warns the new virus variant could fuel huge spikes in COVID-19 cases. The New York Times (2021).

5. Of 3,000 deaths per day, 2,400 are in those 65 and older: 480 deaths in 65-74 and 1,920 deaths in 75+. Compared to vaccinating this group at random over 100 days, vaccinating those 75+ in the first 40 days saves 57,600 in that population. The delay in vaccinating the 65+ group costs 9,600 lives, for a net gain of 48,000 lives.

6. Bassett, M.T., Chen, J.T., & Krieger, N. Variation in racial.ethnic disparities in COVID-19 mortality by age in the United States: A cross-sectional study. PLoS Medicine 17, e1003402 (2020).

7. Venkatesan, P. NICE guideline on long COVID. The Lancet (2021).

8. This assumes 100 million Johnson & Johnson vaccines will only require one shot.

9. Beer, T. Large numbers of health care and frontline workers are refusing COVID-19 vaccine. Forbes (2021).

Lead image: Helenshi / Shutterstock

Get the Nautilus newsletter

Cutting-edge science, unraveled by the very brightest living thinkers.

NL – In Page Mobile

Email: *

Captcha

Sign up for free