SCIENCE ADVANCES | RESEARCH ARTICLE

CORONAVIRUS Copyright © 2021

The Authors, some
Vaccine optimization for COVID-19: rights reserved;
exclusive licensee
Who to vaccinate first? American Association
for the Advancement
of Science. No claim to
Laura Matrajt1*, Julia Eaton2†, Tiffany Leung1†, Elizabeth R. Brown1,3 original U.S. Government
Works. Distributed
Vaccines, when available, will likely become our best tool to control the COVID-19 pandemic. Even in the most under a Creative
optimistic scenarios, vaccine shortages will likely occur. Using an age-stratified mathematical model paired with Commons Attribution
optimization algorithms, we determined optimal vaccine allocation for four different metrics (deaths, symptom- NonCommercial
atic infections, and maximum non-ICU and ICU hospitalizations) under many scenarios. We find that a vaccine License 4.0 (CC BY-NC).
with effectiveness ≥50% would be enough to substantially mitigate the ongoing pandemic, provided that a high
percentage of the population is optimally vaccinated. When minimizing deaths, we find that for low vaccine effec-
tiveness, irrespective of vaccination coverage, it is optimal to allocate vaccine to high-risk (older) age groups first.
In contrast, for higher vaccine effectiveness, there is a switch to allocate vaccine to high-transmission (younger)
age groups first for high vaccination coverage. While there are other societal and ethical considerations, this work
can provide an evidence-based rationale for vaccine prioritization.

Downloaded from http://advances.sciencemag.org/ on July 30, 2021

INTRODUCTION than 65) were relatively more susceptible (8) (we also analyzed a
As of 22 September 2020, more than 960,000 people have died scenario assuming equal susceptibility across age groups; see the
because of the ongoing severe acute respiratory syndrome corona- Supplementary Materials). We assumed that both natural and vaccine-­
virus 2 (SARS-CoV-2) pandemic (1). Different countries have en- induced immunity last at least 1 year (our time horizon). At the
acted different containment and mitigation strategies, but the world beginning of our simulations, 20% of the population have already
awaits impatiently for the arrival of a vaccine as the ultimate tool to been infected and are immune (additional results for 10, 30, and 40%
fight this disease and to allow us to resume our normal activities. of the population can be found in Results and in the Supplementary
There are more than 100 vaccines under development (2, 3), with Materials), and all social distancing interventions have been lifted.
some currently undergoing phase 3 clinical trials (3). However, there To keep our results as general as possible, as each state/country will
are many unknowns surrounding a potential vaccine, including how have different vaccination rates, we did not model the vaccination
effective it would be, how long it would be protective, how effective campaigns in the main analysis. Hence, we assumed that at the be-
it would be in older individuals, how many doses would be immediately ginning of our simulations, vaccination has been carried out and that
available, and how long scaling up the vaccine production would vaccinated individuals have reached the full protection conferred by
take. Furthermore, should early vaccines have low effectiveness, what the vaccine. However, we also analyzed, in a separate scenario, how
are the potential trade-offs between using a low-effectiveness vaccine optimal allocation strategies changed when we modeled vaccination
and waiting for a vaccine with a more desirable vaccine effectiveness campaigns explicitly. Here, we consider that frontline health care
(VE)? With the hope of producing a vaccine in the near future comes workers and other essential personnel (e.g., firefighters and police)
the difficult task of deciding whom to vaccinate first, as vaccine who should obviously be prioritized have already been vaccinated.
shortages are inevitable (4–6). Here, we used a mathematical model For the main analysis, we considered a vaccine having an effect
paired with optimization algorithms to determine the optimal use of reducing susceptibility to infection (referred to VE throughout
of vaccine for 100 combinations of VE and number of doses avail- the text) only. In addition, as a separate analysis, we considered a
able under a wide variety of scenarios. vaccine that would also reduce the probability of COVID-19 disease
(referred to as VECOV below). This effect against COVID-19 disease
was modeled as a combination of the VE to prevent infection and
METHODS the VE to prevent disease-given infection. Because current vaccine
Briefly, we developed a deterministic age-structured mathematical clinical trial protocols have an expected VE against COVID-19 disease
model of SARS-CoV-2 transmission with a population stratified of 60% (9, 10), for this analysis, we set VECOV = 60 while varying the
into 16 age groups (fig. S1). Because, historically, vaccine is distributed relative contributions of the other two vaccine effects (see the
to each state in the United States proportional to its population and Supplementary Materials for full description).
the allocation strategy is then determined at the state level (7), we For the vaccine optimization, we collated the 16 age groups into
chose a state-level model with a population size similar to Washing- five vaccination groups: children (aged 0 to 19), adults between 20 and
ton state and demographics similar to those of the general U.S. popula- 49 years old, adults between 50 and 64 years old, adults between 65 and
tion; however, our results are generalizable to other populations. 74 years old, and those 75 and older. This stratification reflects our
We assumed that children were less susceptible to infection than current knowledge of disease severity and mortality based on age (11, 12).
middle-aged adults (20 to 65 years old), while older adults (older We developed an optimization routine that combined a coarse
global search algorithm with a fast optimizer to explore the entire
1
Vaccine and Infectious Disease Division, Fred Hutchinson Cancer Research Center, space of possible combinations of vaccine allocation. We compared
Seattle, WA, USA. 2University of Washington, Tacoma, WA, USA. 3Department of the optimal allocation strategy given by the optimizer to a pro rata
Biostatistics, University of Washington, Seattle, WA, USA.
*Corresponding author. Email: [email protected] allocation, where the vaccination coverage to each vaccination group
†These authors contributed equally to this work. is distributed proportionally to population size in each group. We

Matrajt et al., Sci. Adv. 2021; 7 : eabf1374 3 February 2021 1 of 11

SCIENCE ADVANCES | RESEARCH ARTICLE

considered VE ranging from 10 to 100% and vaccination coverage (equivalently 40% vaccinated with a perfect vaccine under the
ranging from 10 to 100% of the total population. We evaluated four optimal allocation strategy for minimizing infections assuming
objective functions reflecting different metrics of disease burden that 20% of the population has immunity already) (Figs. 1J and 2A
could be considered by decision makers: minimization of the total and fig. S2).
number of symptomatic infections, total number of deaths, number The epidemic can be substantially slowed with any vaccine with
of cases requiring hospitalization [non–intensive care unit (ICU)] a VE ≥50%, as long as a majority of the population is vaccinated
at the epidemic peak, and number of cases requiring ICU hospital- (Figs. 1E and 2A), and more than 50% of deaths could be averted
ization at the epidemic peak. We chose to minimize symptomatic (in comparison to no vaccination and no nonpharmaceutical
infections as a key metric because symptomatic individuals are the intervention) with as little as 35% of the population optimally vac-
ones who are easier to identify and, presumably, particular inter- cinated (Fig. 2, A and B). If VE = 60%, then the epidemic is com-
ventions will be targeted to this group. In addition, minimizing pletely contained if we optimally vaccinate 70% of the population
symptomatic individuals minimizes the transmission of SARS-CoV-2, (or 50% for higher VE = 70%) (Figs. 1, F and I, and 2A). In our
and this is in line with current vaccine trials end points (9, 10). The model, only vaccines with VE ≥50% can maintain the number of
last two objective functions were chosen because hospital bed (non-ICU non-ICU hospitalizations below the established goal (≤10% hospital
and ICU) occupancy is a key metric currently used to determine bed occupancy by patients with COVID-19) and can prevent an
county/state/country readiness to move between different interven- overflow of the ICUs. With VE = 60%, optimally vaccinating 54%
tion strategies. Here, we used the total number of licensed ICU beds satisfies both hospital bed occupancy goals (Fig. 2, C and D, and
in Washington state and its current goal of staying below 10% of figs. S3F and S4F), compared with 67% under the pro rata alloca-
hospital beds occupied by COVID-19 cases (13, 14) as references tion (figs. S5 to S9). The optimal allocation strategy outperforms

Downloaded from http://advances.sciencemag.org/ on July 30, 2021

when interpreting our results. Full details of the methods can be the pro rata allocation most under low vaccine availability, with a
found in the Supplementary Materials. maximum difference of 32% deaths averted (for VE = 100% and with
enough vaccine to cover 20% of the population) and 32% symp-
tomatic infections averted (for VE = 60% and vaccination coverage
RESULTS of 60%) when compared with a pro rata allocation strategy (Fig. 3
Epidemic mitigation and containment and figs. S5 and S10). As VE increases, both strategies tend to per-
Our model suggests that, for a basic reproduction number R0 = 3, form similarly as vaccination coverage increases (Fig. 3 and figs. S5
herd immunity will be achieved once 60% of the population is infected and S10).

Fig. 1. Simulated prevalence of symptomatic infections. Simulated prevalence of symptomatic COVID-19 infections for VE ranging from 10% (A) to 100% (J) in 10%
increments. For each VE and each vaccination coverage, the optimal vaccine allocation for minimizing symptomatic infections was used in these simulations. Colors
represent different vaccination coverage, ranging from 0% (black, “baseline”) to 100% (magenta). For clarity, we present here epidemic curves for the main set of param-
eters only and show a complete figure with uncertainty bounds in fig. S2.

Matrajt et al., Sci. Adv. 2021; 7 : eabf1374 3 February 2021 2 of 11

SCIENCE ADVANCES | RESEARCH ARTICLE

Downloaded from http://advances.sciencemag.org/ on July 30, 2021

Fig. 2. Four key metrics of COVID-19 burden under optimal distribution of vaccine. Percentage of symptomatic infections (A) and deaths (B) averted, and number of
maximum non-ICU (C) and ICU (D) hospitalizations as a function of VE and vaccination coverage (total vaccine available as a percentage of the population). The dotted
lines correspond to VE = 50% and vaccine available to cover 50% of the population. The isoclines indicate the current goal for Washington state having 10% of licensed
general (non-ICU) hospital beds occupied by patients with COVID-19 in (C) and total ICU licensed hospital beds in Washington state in (D).

Optimal vaccine allocation changes with  Optimal vaccine allocation differs for 
VE and vaccination coverage different objective functions
The optimal allocation strategy to minimize deaths is identical for Next, we investigated how the optimal allocation strategy changed
VE between 10 and 50%: With low vaccination coverage, it is optimal for different objective functions and present results for VE = 60%.
to allocate vaccine first to the highest-risk group (people more than The optimal vaccine allocation for the four objectives differed the
75 years old) and then to the younger vaccination groups as more most when fewer vaccines are available (enough vaccine to cover
vaccine becomes available (Fig. 4, A to E). In contrast, there is a less than 30% of the total population). When minimizing symptomatic
threshold phenomenon observed for VE ≥60%: For low coverage, infections and peak non-ICU hospitalizations, priority was given to
the optimal allocation is still to vaccinate the high-risk groups first, the younger vaccination groups, as they have the most contacts in
but when there is enough vaccine to cover roughly half of the popula- our model and, hence, drive transmission (Fig. 5, A and B, and figs.
tion (60% for VE = 60%, 50% for VE = 70%, and 40% for VE≥80%), S11 and S12). As we move toward more severe outcomes (ICU
there is a switch to allocate vaccine to the high-transmission groups hospitalizations at peak and deaths), for which older individuals are
first (those aged 20 to 50 and children in our model). This is because most at risk, the optimal allocation strategy shifts toward those vac-
directly vaccinating those who are driving the epidemic results in a cination groups (Fig. 5, C and D, and fig. S13). Once more vaccine
much slower epidemic curve and, hence, in fewer deaths (Fig. 1, F to H). becomes available, the optimal allocation strategies are very similar
As more vaccine becomes available, the optimizer allocates it to for all objective functions. They are nearly identical for all the objec-
high-risk groups again (Fig. 4, F to J). tive functions when there is enough vaccine to cover 60 and 70% of

Matrajt et al., Sci. Adv. 2021; 7 : eabf1374 3 February 2021 3 of 11

SCIENCE ADVANCES | RESEARCH ARTICLE

Downloaded from http://advances.sciencemag.org/ on July 30, 2021

Fig. 3. Percentage of deaths averted under the optimal and the pro-rata strategies for different VE. Percentage of deaths averted for the optimal allocation strate-
gy (blue) and the pro rata strategy (green) for VE ranging from 10 (A) to 100% (J) in 10% increments and vaccination coverage ranging from 10 to 100% of the total
population. The shaded areas represent results of the 1000 simulations with the top and bottom 2.5% simulations removed.

Fig. 4. Optimal allocation strategies to minimize deaths for different VE. Optimal allocation strategies for minimizing deaths for VE ranging from 10 (A) to 100% (J) in
10% increments (additional figures for minimizing symptomatic infections, number of non-ICU hospitalizations at peak, and number of ICU hospitalizations at peak are
given in the Supplementary Materials). For each plot, each row represents the total vaccination coverage available (percentage of the total population to be vaccinated),
and each column represents a different vaccination group. Colors represent the percentage of the population in a given vaccination group to be vaccinated.

Matrajt et al., Sci. Adv. 2021; 7 : eabf1374 3 February 2021 4 of 11

SCIENCE ADVANCES | RESEARCH ARTICLE

against COVID-19 disease has a huge impact on symptomatic infections

and hospitalizations, even when this vaccine is marginally effective
against preventing infection (Fig. 7). For example, if VE = 10%, then
it can prevent 50% of the symptomatic infections when we optimally
vaccinate 64% of the population (Fig. 7A). Furthermore, for this VE
and this vaccination coverage, peak hospitalizations are substantially
reduced (6840 and 3392 non-ICU and ICU hospitalizations, respec-
tively; Fig. 7, B and C) when compared with a vaccine without an
effect in COVID-19 disease (15,091 and 8405 non-ICU and ICU
hospitalizations, respectively; Fig. 2, B and C).

Optimal vaccine allocation as a function of preexisting

immunity to SARS-CoV-2
As the COVID-19 pandemic dynamics have been markedly dif-
ferent across the globe, we expect to see a range of population-­
level naturally acquired immunity when vaccination campaigns
start. Hence, we investigated the optimal use of vaccine with 10,
30, and 40% of the population already immune at the beginning
of the simulations. For all of these scenarios, the same pattern is

Downloaded from http://advances.sciencemag.org/ on July 30, 2021

observed when minimizing deaths: For low coverage, it is optimal
to allocate all of the vaccine to the high-risk groups; for higher
coverage, the optimal vaccination strategy switches to allocate
more vaccine to the high-transmission groups. This threshold,
however, varies with the degrees of preexisting immunity in the
population. When only 10% of the population has natural immu-
nity, the switch occurs at 80% vaccination coverage, but when 40%
has natural immunity, the threshold is observed at 40% vaccination
coverage (Fig. 8).
In addition, under low preexisting immunity to SARS-CoV-2,
Fig. 5. Optimal allocation strategies for all objective functions analyzed. Optimal the optimal strategy favors vaccinating the older vaccination groups
allocation strategies for minimizing: Symptomatic infections (A), number of non-ICU (Fig. 8A), while under higher preexisting immunity, the optimal
hospitalizations at peak (B), number of ICU hospitalizations at peak (C), and total num- allocation strategy tends to distribute vaccine more evenly across
ber of deaths (D). Here, we assumed VE = 60%. For each plot, each row represents the vaccination groups (Fig. 8D).
total vaccination coverage available (percentage of the total population to be vacci-
nated), and each column represents a different vaccination group. Colors represent
Modeling the vaccination campaign
the percentage of the population in a given vaccination group to be vaccinated.
In this section, we modeled the vaccination campaign and determined
the optimal vaccine allocation. We extended our time horizon to
the population. For high coverage, the optimal allocation strategies 2 years and considered administering 75,000, 150,000, or 300,000 doses
for all objective functions shifted toward the high-transmission groups. of vaccine per week. This corresponds to vaccinating the entire popu-
Of note, we did not impose the optimizer to use all the available lation in 101, 50, or 25 weeks, respectively. The first vaccination rate
vaccine. As a result, the optimizer found allocation strategies using was chosen to roughly match the vaccination rate during the 2009
less than the total vaccine available while performing equally well. H1N1 pandemic in the United States (0.87% of the population weekly)
This was very prominent when VE and vaccination coverage were (15). Note that to avoid confounding, as in the rest of this work, we
very high. For example, when minimizing peak ICU hospitalizations did not assume any social distancing intervention. Similar to previ-
and VE = 90%, the optimizer used vaccine to cover 75% of the pop- ous results, when the vaccination campaign is modeled, the optimal
ulation even though there was vaccine available to cover the entire vaccine allocation strategies were very different depending on the
population. This is expected because complete containment is attained objective function: While the older age groups are prioritized when
once a high proportion of the population is vaccinated, and any vaccine minimizing deaths and maximum ICU hospitalizations (Fig. 9), the
used above that threshold will result in the same mathematical outcome. optimizer allocated vaccine to younger age groups when minimiz-
ing symptomatic infections or maximum non-ICU hospitalizations
Optimal vaccine allocation assuming VE against COVID-19 (fig. S17). Furthermore, older age groups were prioritized when
disease (VECOV) vaccination rate was slow (75,000 doses administered per week, when
In this section, we considered a vaccine that, in addition to reducing minimizing deaths and ICU peak hospitalization). For faster vacci-
the probability of acquiring infection, would reduce the probability nation campaigns, the optimizer allocated vaccine to younger age
of COVID-19 disease (see the Supplementary Materials for full de- groups in addition to the older age groups (Fig. 9). There were also
tails). We considered VECOV = 60%, in line with expected VE in the some important differences. First, for any VE and any vaccination
current trial protocols (9, 10). The optimal allocation strategies as- rate, we did not observe any threshold in vaccination coverage. In
suming VECOV = 60% are nearly identical to the ones without this addition, because the time frame of the vaccination campaigns occurs
effect (Fig. 6 and figs. S14 to S16). As expected, including this effect at a much slower speed than the epidemic, most of the vaccines under

Matrajt et al., Sci. Adv. 2021; 7 : eabf1374 3 February 2021 5 of 11

SCIENCE ADVANCES | RESEARCH ARTICLE

Downloaded from http://advances.sciencemag.org/ on July 30, 2021

Fig. 6. Optimal allocation strategies for different VE with a vaccine including an effect against COVID-19 disease. Optimal allocation strategies for minimizing total
symptomatic infections for VE ranging from 10% (A) to 60% (F) in 10% increments for VECOV = 60%. For each plot, each row represents the total vaccination coverage
available (percentage of the total population to be vaccinated), and each column represents a different vaccination group. Colors represent the percentage of the popu-
lation in a given vaccination group to be vaccinated.

A B C

Fig. 7. Three key metrics of COVID-19 burden under optimal distribution of vaccine for VECOV = 60%. Percentage of symptomatic infections averted (A) and number
of maximum non-ICU (B) and ICU (C) hospitalizations as a function of VE and vaccination coverage (total vaccine available as a percentage of the population). The dotted
lines correspond to VE = 50% and vaccine available to cover 50% of the population. The isoclines indicate the current goal for Washington state having 10% of licensed
general (non-ICU) hospital beds occupied by patients with COVID-19 in (B) and total ICU licensed hospital beds in Washington state in (C).

Matrajt et al., Sci. Adv. 2021; 7 : eabf1374 3 February 2021 6 of 11

SCIENCE ADVANCES | RESEARCH ARTICLE

to one assuming equal susceptibility across age groups (figs. S19 to

S22), as suggested in (17, 18). The optimal allocation strategies under
both equal and differential susceptibility were remarkably consistent,
but, as expected, assuming equal susceptibility resulted in strategies
allocating slightly more vaccine to children (assuming equal sus-
ceptibility) as VE increases (VE ≥60) and more vaccine becomes
available (coverage to vaccinate 70% or higher) (figs. S19 to S22).
The major differences were observed when minimizing peak non-
ICU hospitalizations for low VE and low vaccination coverage [less
than, assuming equal susceptibility resulted in favoring adults aged
50 to 65 over younger adults (fig. S21)].
Susceptibility to symptomatic disease. While we know that children
are much less susceptible to develop severe disease (19), the role of
age in the probability of developing any kind of COVID-19 symp-
toms remains currently unclear. Hence, we compared optimal allo-
cation strategies assuming equal probability of all age groups to
develop symptoms (presented throughout the text), as suggested in
(19, 20), to ones assuming different probabilities by age (children
being less susceptible and older adults being more susceptible), as

Downloaded from http://advances.sciencemag.org/ on July 30, 2021

suggested in (21). The optimal allocation strategies for minimizing
deaths and peak ICU are very consistent (figs. S23 and S24), but, as
expected, we observed some changes when minimizing symptom-
atic infections and peak non-ICU hospitalizations (figs. S25 and
S26). When we minimized symptomatic infections, the optimal
allocation strategies were very similar for vaccines with VE ≥60. For
lower VE, the optimal allocation strategies tended to favor more the
middle- and older-age adult groups (fig. S25). When minimizing
maximum non-ICU hospitalizations, the optimal allocation strate-
gies were nearly identical for high vaccination coverage (≥70%) but
Fig. 8. Optimal allocation strategies for minimizing deaths assuming different switched to adults aged 50 to 75 (as opposed to younger adults) if
levels of pre-existing immunity in the population. Optimal allocation strategies vaccination coverage is lower (fig. S26).
for minimizing deaths assuming 10% (A), 20% (B), 30% (C), and 40% (D) of the pop- Basic reproduction number R0. Many regions in the world have
ulation has natural immunity to COVID-19 at the start of the simulations. Here, we controlled the epidemic using social distancing interventions and
assumed VE = 60%. For each plot, each row represents the total vaccination cover- have reduced the effective reproduction number Rt to hover around
age available (percentage of the total population to be vaccinated), and each column 1. Hence, we analyzed the optimal allocation strategies in this sec-
represents a different vaccination group. Colors represent the percentage of the popula-
tion with R0 set to 1.5, 2, and 2.5, assuming still that at the beginning
tion in a given vaccination group to be vaccinated.
of our simulations, 20% of the population have been infected and
are now recovered. For minimizing deaths and ICU peak hospital-
this scenario are given after the epidemic (this was especially true izations, when R0 = 1.5, the optimal allocation strategy favors vacci-
when considering the 75,000 campaign). This resulted in optimal nating children for low VE (Fig. 10A and fig. S27), is more equally
allocation strategies that, for a given VE, were identical beyond a distributed if VE = 60%, and favors vaccinating the older age groups
certain vaccination coverage (Fig. 9 and fig. S17), and the measured for higher VE (Fig. 10, D and G). If R0 = 2, then we observe the same
outcomes did not improve beyond that coverage (fig. S18). This points threshold phenomenon described for higher R0, but at a lower VE:
to the fact that even in the very optimistic scenario of having a highly For VE = 30%, the optimal use of vaccine is to allocate it to high-risk
efficacious vaccine given at very high rates, additional interventions groups under low vaccine coverage and to the younger groups once
would be necessary to control the epidemic while the vaccination there is enough vaccine to cover 60% of the population. As VE is higher,
campaign takes place. this threshold moves up (Fig. 10, C, G, and K). Last, for R0 = 2.5,
the optimal allocation strategy is identical to the one for R0 = 3. For
Robustness of optimal allocation strategies minimizing symptomatic infections and non-ICU peak hospitalizations,
around major parameters the optimal allocation strategies were more homogeneous, favoring
One-way robustness analysis more the younger age groups in alignment with the results presented
We explored the robustness of the optimal allocation strategies around above (figs. S28 and S29). Note here that social distancing interven-
key features of the transmission and natural history of SARS-CoV-2. tions have kept the effective reproduction number low, but that, if those
For each of these features, we investigated how changing that par- measures are lifted, then Rt would go up again. Hence, any vaccination
ticular parameter would change the optimal allocation strategy. program should take this into account.
Susceptibility to infection. Because the effect of age on suscepti- Distribution of preexisting immunity in the population. Here, we
bility to SARS-CoV-2 infection remains unclear, we compared the assumed a different distribution of preexisting immunity in the
optimal allocation strategy under the assumption of differential population. For this section of the analysis, we ran the simulations
susceptibility, as suggested in (8, 16) (presented throughout the text), without any vaccination until the recovered compartments reached

Matrajt et al., Sci. Adv. 2021; 7 : eabf1374 3 February 2021 7 of 11

SCIENCE ADVANCES | RESEARCH ARTICLE

Downloaded from http://advances.sciencemag.org/ on July 30, 2021

Fig. 9. Optimal allocation strategies for minimizing deaths assuming different vaccination rates. Optimal allocation strategies for minimizing deaths for two
VE = 50% (A to C) and 90% (D to F) and for three different vaccination rates: 75,000 (A and D), 150,000 (B and E), and 300,000 (C and F) vaccine doses administered per
week. For each plot, each row represents the total vaccination coverage available (percentage of the total population to be vaccinated), and each column represents a
different vaccination group. Colors represent the percentage of the population in a given vaccination group to be vaccinated.

20% of the population. We then used this composition of the popula- Number of current infections. We compared the optimal alloca-
tion as the initial distribution of preexisting immunity (this composi- tion strategies when the simulations were started with a higher number
tion will depend on the contact matrix and the demographics used). of infected individuals (10,000 current infections). This would re-
The optimal allocation strategies were very similar to those obtained in flect a situation where the epidemic is in full exponential growth
the main analysis, with some notable differences. First, under this sce- when vaccination becomes available. The optimal allocation strategy
nario, the optimal allocation strategies tended to protect vaccination was unexpectedly robust under this scenario, with nearly identical
groups in full before prioritizing other groups. This was very apparent allocation strategies for all objective functions (fig. S35).
when minimizing deaths or peak ICU hospitalizations: For low VE Multiway robustness analysis
(irrespective of vaccination coverage) or high VE but low vaccination In addition, we selected four parameters (R0, proportion of infec-
coverage, the optimal allocation strategies prioritized the highest-risk tions that are asymptomatic, relative infectiousness of asymptomatic
age groups (individuals aged 65 to 75 and those more than 75 years infections, and relative infectiousness of presymptomatic infections)
old) to get fully vaccinated before allocating to other groups (figs. S30 for which there is the most uncertainty and reran the optimization
and S31). Second, the threshold observed when minimizing deaths routine for several combinations of them (full details in the Supple-
occurred at higher vaccination coverage under this scenario. For ex- mentary Materials). The optimal allocation strategies were very
ample, if VE = 60%, then this threshold occurs when there is enough robust under this analysis (Supplemental Files SF1 to SF4).
vaccine to cover 70% of the population (fig. S30, F to J). Last, when
minimizing symptomatic infections or non-ICU peak hospitaliza-
tions, the optimal allocation strategies shifted away from young adults DISCUSSION
toward adults in the 50-to-65 vaccination groups (figs. S32 and S33). The COVID-19 pandemic has devastated families and societies
Duration of the incubation period. On the basis of the early studies around the world. A vaccine, when available, would most likely be-
(17, 22–24), we presented results assuming an incubation period of come our best tool to control the spread of SARS-CoV-2. However,
5.1 days. However, a more recent study (25) has suggested that the in the short term, even in the most optimistic scenarios, vaccine
incubation period for COVID-19 might be longer (7.76 days). We production would likely be insufficient. In this work, we paired a
found no difference in the optimal allocation strategies assuming this mathematical model of SARS-CoV-2 transmission with optimization
longer incubation period (fig. S34). algorithms to determine optimal vaccine allocation strategies. Given

Matrajt et al., Sci. Adv. 2021; 7 : eabf1374 3 February 2021 8 of 11

SCIENCE ADVANCES | RESEARCH ARTICLE

Downloaded from http://advances.sciencemag.org/ on July 30, 2021

Fig. 10. Optimal allocation strategies for minimizing deaths for different values of R0. Optimal allocation strategies for minimizing deaths for three different VE: 30% (A to C),
60% (D to F), and 90% (G to I) for three different values of R0 = 1.5, 2, and 2.5 (additional figures for minimizing symptomatic infections, number of non-ICU hospitalizations at peak,
and number of ICU hospitalizations at peak are given in the Supplementary Materials). For each plot, each row represents the total vaccination coverage available (percentage of
the total population to be vaccinated), and each column represents a different vaccination group. Colors represent the percentage of the population in a given vaccination group
to be vaccinated.

the current uncertainties surrounding such a vaccine (we do not yet vaccine modeling studies (26, 27). Furthermore, we showed that much
know whether and when this vaccine would be available, how effi- can be achieved even with low vaccination coverage; With medium
cacious it will be, or the number of doses immediately available), we VE, more than half of deaths can be averted by optimally vaccinat-
explored 100 combinations of VE and vaccination coverage under a ing only 35% of the population. When minimizing deaths, for low
wide variety of scenarios minimizing four metrics of disease burden. VE and a low supply of vaccine, our results suggest that vaccines
Our results suggest that, assuming R0 = 3, any vaccine with should be given to the high-risk groups first. For high VE and high
medium to high effectiveness (VE ≥50%) would be able to consid- vaccination coverage, the optimal allocation strategy switched to
erably slow the epidemic while keeping the burden on health care vaccinating the high-transmission groups (younger adults and
systems manageable, as long as a high proportion of the population children). This remained true under equal or reduced susceptibility
is optimally vaccinated. Moreover, once VE = 70%, full containment to infection for children, pointing to the importance of children as
of the epidemic would be possible. This is in agreement with key players in disease transmission. This finding is consistent with

Matrajt et al., Sci. Adv. 2021; 7 : eabf1374 3 February 2021 9 of 11

SCIENCE ADVANCES | RESEARCH ARTICLE

previous work for other respiratory viruses (28–30) that found that coverage could achieve the same goals, because deterministic models
protecting the high-transmission groups indirectly protects the tend to overestimate the transmission dynamics. We computed the
high-risk groups and is the optimal use of resources. optimal allocation strategies using age as the sole risk factor. How-
Furthermore, the optimal allocation strategies were identical when ever, other factors, such as occupation, have been linked to an increased
we considered a vaccine that would also reduce symptomatic infec- risk of acquisition and severe disease (33, 34). Furthermore, several
tions, but the impact of such a vaccine would of course be greater in studies (35, 36) have shown that, as a result of health systems with
reducing COVID-19 disease and health care burden. Our results show systemic health and social inequalities, people from racial and ethnic
that even if this vaccine had a marginal effect in preventing infec- minority groups are at increased risk of getting sick and dying from
tion, it would still be very beneficial to reduce the number of hospital- COVID-19 in certain countries. These are crucial considerations that
izations and symptomatic infections. It is expected that once a vaccine will be included in further studies and can point toward who, within
is proven to be effective, more information about its mechanisms of a given age group, should get the vaccine first. We believe that these
action, including how it affects the viral load trajectory and the rela- results can provide a quantification of the effectiveness of different
tionship between that trajectory and infectiousness, will be available, allocation scenarios under four metrics of disease burden and can
allowing us to expand and refine our projections. be used as an evidence-based guidance to vaccine prioritization.
Here, we used mathematical optimization to determine the opti-
mal vaccine allocation and, by design, did not impose any restric- SUPPLEMENTARY MATERIALS
tions in the allocation strategies. Supplementary material for this article is available at http://advances.sciencemag.org/cgi/
content/full/7/6/eabf1374/DC1
However, in practice, implementation of optimal strategies must
also account for other factors (ethical, political, and societal). When

Downloaded from http://advances.sciencemag.org/ on July 30, 2021

large quantities of vaccine are available, a feasible solution could REFERENCES AND NOTES
1. Johns Hopkins University and Medicine, Coronavirus COVID-19 global cases by the
involve first vaccinating the high-risk groups and then allocating the Center for Systems Science and Engineering (CSSE) at Johns Hopkins University (JHU)
remaining vaccine to the high-transmission groups. (2020); https://coronavirus.jhu.edu/map.html.
This study has several limitations. Our model assumes that both 2. A. Mullard, COVID-19 vaccine development pipeline gears up. Lancet 395, 1751–1752
natural and vaccine-acquired immunity will last for at least 1 year. (2020).
3. C. Zimmer, J. Corum, S.-L. Wee, Coronavirus Vaccine Tracker (2020); https://www.nytimes.
We do not yet know how long immunity against SARS-CoV-2 will
com/interactive/2020/science/coronavirus-vaccine-tracker.html.
last. There is some evidence that neutralizing antibodies become 4. R. Cohen, C. Jung, N. Ouldali, A. Sellam, C. Batard, F. Cahn-Sellem, A. Elbez, A. Wollner,
undetectable a few weeks following infection (31), although it is O. Romain, F. Corrard, S. Aberrane, N. Soismier, R. Creidy, M. Smati-Lafarge, O. Launay,
unclear how this correlates with immunity. If immunity were short- S. Bechet, E. Varon, C. Levy, Assessment of spread of SARS-CoV-2 by RT-PCR
lived, then these results would only be applicable for that duration. and concomitant serology in children in a region heavily affected by COVID-19 pandemic.
medRxiv 2020.06.12.20129221 (2020).
Furthermore, we assumed that asymptomatic and symptomatic in- 5. A. D. Usher, COVID-19 vaccines for all? Lancet 395, 1822–1823 (2020).
fections would confer equal immunity. However, it is conceivable 6. M. Twohey, Who Gets a Vaccine First? U.S. Considers Race in Coronavirus Plans (2020).
that asymptomatic infections might result in a weaker immune re- 7. Centers for Disease Control and Prevention, 2009 H1N1 Flu (2009); www.cdc.gov/
sponse (32). We chose four metrics of disease burden to minimize. h1n1flu/vaccination/vaccinesupply.htm.
8. J. Zhang, M. Litvinova, Y. Liang, Y. Wang, W. Wang, S. Zhao, Q. Wu, S. Merler, C. Viboud,
However, other metrics, such as minimization of asymptomatic in-
A. Vespignani, M. Ajelli, H. Yu, Changes in contact patterns shape the dynamics
fections, or a combination of all of these metrics might be the key to of the COVID-19 outbreak in China. Science 368, 1481–1486 (2020).
stop the spread of the epidemic. We have identified optimal alloca- 9. Moderna, A phase 3, randomized, stratified, observer-blind, placebo-controlled study to
tion strategies, and once more information about a vaccine charac- evaluate the efficacy, safety, and immunogenicity mRNA-1273 SARS-CoV-2 vaccine in
teristics is known, validating our allocation strategies with more adults aged 18 years and older (2020).
10. Pfizer, A phase 1/2/3, placebo-controlled, randomized, observer-blind, dose-finding
complex models is welcome. To avoid confounding effects from study to evaluate the safety, tolerability, immunogenicity, and efficacy of SARS-CoV-2
different interventions, we optimized vaccine allocation assuming RNA vaccine candidates against COVID-19 in healthy individuals (2020), pp. 1–137.
no social distancing interventions in place. In reality, vaccination, at 11. CDC, COVID-19 Pandemic Planning Scenarios (2020); www.cdc.gov/coronavirus/2019-
least at the beginning, would take place while some social distancing ncov/hcp/planning-scenarios.html.
12. N.M. Ferguson, D. Laydon, G. Nedjati-Gilani, N. Imai, K. Ainslie, M. Baguelin, S. Bhatia,
interventions remain in effect. Under those circumstances, we would
A. Boonyasiri, Z. Cucunubá, G. Cuomo-Dannenburg, A. Dighe, I. Dorigatti, H. Fu,
need less vaccine to control the epidemic. In that sense, our results K. Gaythorpe, W. Green, A. Hamlet, W. Hinsley, L. C. Okell, S. van Elsland, H. Thompson,
are conservative. To keep the optimization from being unreasonably R. Verity, E. Volz, H. Wang, Y. Wang, Patrick GT Walker, C. Walters, P. Winskill, C. Whittaker,
long, our model assumes that vaccination is given all at once and C. A. Donnelly, S. Riley, A. C. Ghani, Impact of non-pharmaceutical interventions (NPIs) to
does not capture geographical differences or other heterogeneities. reduce COVID-19 mortality and healthcare demand (2020); https://doi.
org/10.25561/77482.
We used mortality and hospitalization rates published by 13. Washington State Coronavirus Response, COVID-19 risk assessment dashboard (2020);
Ferguson et  al. (12) that were based on the epidemic in Wuhan, https://coronavirus.wa.gov/what-you-need-know/covid-19-risk-assessment-dashboard.
China, but these rates may vary vastly in different regions. In particular, it 14. State of California, COVID-19 data and tools (2020); https://covid19.ca.gov/data-and-tools/.
is now known that certain underlying conditions (e.g., obesity, 15. Centers for Disease Control and Prevention, Final estimates for 2009–10 seasonal
influenza and influenza A (H1N1) 2009 monovalent vaccination coverage, August 2009
diabetes, and heart disease) are important risk factors for develop-
to May 2010 (2010); www.cdc.gov/flu/fluvaxview/coverage_0910estimates.htm.
ing severe COVID-19 hospitalization and death. As the population 16. Q.-L. Jing, M.-J. Liu, J. Yuan, Z.-B. Zhang, A.-R. Zhang, N. E. Dean, L. Luo, M. Ma,
makeup of individuals with underlying conditions can be very dif- I. Longini, E. Kenah, Y. Lu, Y. Ma, N. Jalali, L.-Q. Fang, Z.-C. Yang, Y. Yang, Household
ferent in different countries, it is then key to determine region-based secondary attack rate of COVID-19 and associated determinants. medRxiv
estimates of COVID-19–related hospitalizations and deaths, so that 2020.04.11.20056010 (2020).
17. Q. Bi, Y. Wu, S. Mei, C. Ye, X. Zou, Z. Zhang, X. Liu, L. Wei, S. A. Truelove, T. Zhang, W. Gao,
models can be adequately parameterized. Furthermore, we compared C. Cheng, X. Tang, X. Wu, Y. Wu, B. Sun, S. Huang, Y. Sun, J. Zhang, T. Ma, J. Lessler,
modeled peak hospitalizations to current state goals for hospital bed T. Feng, Epidemiology and transmission of COVID-19 in 391 cases and 1286 of their close
occupancy, but it is possible that a lower VE or a lower vaccination contacts in Shenzhen, China: A retrospective cohort study. Lancet 20, 911–919 (2020).

Matrajt et al., Sci. Adv. 2021; 7 : eabf1374 3 February 2021 10 of 11

SCIENCE ADVANCES | RESEARCH ARTICLE

18. F. P. Havers, C. Reed, T. Lim, J. M. Montgomery, J. D. Klena, A. J. Hall, A. M. Fry, 39. K. Prem, A. R. Cook, M. Jit, Projecting social contact matrices in 152 countries using
D. L. Cannon, C.-F. Chiang, A. Gibbons, I. Krapiunaya, M. Morales-Betoulle, K. Roguski, contact surveys and demographic data. PLOS Comput. Biol. 13, e1005697 (2017).
M. A. U. Rasheed, B. Freeman, S. Lester, L. Mills, D. S. Carroll, S. M. Owen, J. A. Johnson, 40. Z. Du, X. Xu, Y. Wu, L. Wang, B. J. Cowling, L. A. Meyers, The serial interval
V. Semenova, C. Blackmore, D. Blog, S. J. Chai, A. Dunn, J. Hand, S. Jain, S. Lindquist, of COVID-19 from publicly reported confirmed cases. Emerg. Infect. Dis. 26,
R. Lynfield, S. Pritchard, T. Sokol, L. Sosa, G. Turabelidze, S. M. Watkins, J. Wiesman, 1341–1343 (2020).
R. W. Williams, S. Yendell, J. Schiffer, N. J. Thornburg, Seroprevalence of antibodies 41. H. Nishiura, N. M. Linton, A. R. Akhmetzhanov, Serial interval of novel coronavirus
to SARS-CoV-2 in 10 sites in the United States, March 23-May 12, 2020. JAMA Intern. Med. (COVID-19) infections. Int. J. Infect. Dis. 93, 284–286 (2020).
180, 1576–1586 (2020). 42. M. E. Halloran, M. Haber, I. M. Longini, Interpretation and estimation of vaccine efficacy
19. X. Lu, L. Zhang, H. Du, J. Zhang, Y. Y. Li, J. Qu, W. Zhang, Y. Wang, S. Bao, Y. Li, C. Wu, under heterogeneity. Am. J. Epidemiol. 136, 328–343 (1992).
H. Liu, D. Liu, J. Shao, X. Peng, Y. Yang, Z. Liu, Y. Xiang, F. Zhang, R. M. Silva, K. E. Pinkerton, 43. M. E. Halloran, I. M. Longini, C. J. Struchiner, Design and Analysis of Vaccine Studies
K. Shen, H. Xiao, S. Xu, G. W. K. Wong; Chinese Pediatric Novel Coronavirus Study Team, (Springer, 2010).
SARS-CoV-2 infection in children. N. Engl. J. Med. 382, 1663–1665 (2020). 44. C. Audet, W. Hare, Derivative-Free and Blackbox Optimization (Cham, Switzerland,
20. K.-Q. Kam, C. F. Yung, L. Cui, R. T. P. Lin, T. M. Mak, M. Maiwald, J. Li, C. Y. Chong, K. Nadua, Springer, 2017).
N. W. H. Tan, K. C. Thoon, A well infant with coronavirus disease 2019 (COVID-19) 45. The Sage Developers, SageMath, the Sage Mathematics Software System (Version 9.1)
with high viral load. Clin. Infect. Dis. 71, 847–849 (2020). (2020); www.sagemath.org.
21. N. G. Davies, A. J. Kucharski, R. M. Eggo, A. Gimma; CMMID COVID 19 working group, 46. S. Kotz, N. Balakrishnan, N.L. Johnson, Continuous Multivariate Distributions (Models and
W. J. Edmunds, The effect of non-pharmaceutical interventions on COVID-19 cases, Applications, Wiley, 2000), vol. 1.
deaths and demand for hospital services in the UK: A modelling study. Lancet 5, 47. F. Gao, L. Han, Implementing the Nelder-Mead simplex algorithm with adaptive
E375–E385 (2020). parameters. Comput. Optimiz. Appl. 51, 259–277 (2012).
22. S. A. Lauer, K. H. Grantz, Q. Bi, F. K. Jones, Q. Zheng, H. R. Meredith, A. S. Azman, 48. P. Virtanen, T. E. Oliphant, M. Haberland, T. Reddy, D. Cournapeau, E. Burovski,
N. G. Reich, J. Lessler, The incubation period of coronavirus disease 2019 (COVID-19) P. Peterson, W. Weckesser, J. Bright, S. J. van der Walt, M. Brett, J. Wilson, K. J. Millman,
from publicly reported confirmed cases: Estimation and application. Ann. Intern. Med. N. Mayorov, A. R. J. Nelson, E. Jones, R. Kern, E. Larson, C. J. Carey, İ. Polat, Y. Feng,
172, 577–582 (2020). E. W. Moore, J. V. Plas, D. Laxalde, J. Perktold, R. Cimrman, I. Henriksen, E. A. Quintero,

Downloaded from http://advances.sciencemag.org/ on July 30, 2021

23. Q. Li, X. Guan, P. Wu, X. Wang, L. Zhou, Y. Tong, R. Ren, K. S. M. Leung, E. H. Y. Lau, C. R. Harris, A. M. Archibald, A. H. Ribeiro, F. Pedregosa, P. van Mulbregt; SciPy 1.0
J. Y. Wong, X. Xing, N. Xiang, Y. Wu, C. Li, Q. Chen, D. Li, T. Liu, J. Zhao, M. Liu, W. Tu, Contributors, SciPy 1.0: Fundamental algorithms for scientific computing in python.
C. Chen, L. Jin, R. Yang, Q. Wang, S. Zhou, R. Wang, H. Liu, Y. Luo, Y. Liu, G. Shao, H. Li, Nat. Methods 17, 261–272 (2020).
Z. Tao, Y. Yang, Z. Deng, B. Liu, Z. Ma, Y. Zhang, G. Shi, T. T. Y. Lam, J. T. Wu, G. F. Gao, 49. J. Kennedy, R. Eberhart, Particle swarm optimization. Proc. IEEE Int. Conf. Neural Net. 4,
B. J. Cowling, B. Yang, G. M. Leung, Z. Feng, Early transmission dynamics in Wuhan, China, 1942–1948 (1995).
of novel coronavirus-infected pneumonia. N. Engl. J. Med. 382, 1199–1207 (2020). 50. L. J. V. Miranda, PySwarms: A research toolkit for Particle Swarm Optimization in Python.
24. N. M. Linton, T. Kobayashi, Y. Yang, K. Hayashi, A. R. Akhmetzhanov, S.-m. Jung, B. Yuan, J. Open Source Soft. 3, 433 (2018).
R. Kinoshita, H. Nishiura, Incubation period and other epidemiological characteristics 51. W. E. Wei, Z. Li, C. J. Chiew, S. E. Yong, M. P. Toh, V. J. Lee, Presymptomatic transmission
of 2019 novel coronavirus infections with right truncation: A statistical analysis of publicly of SARS-CoV-2—Singapore. Morb. Mortal. Wkly. Rep. 69, 411–415 (2020).
available case data. J. Clin. Med. 9, 538 (2020). 52. Y.-H. Lee, C. M. Hong, D. H. Kim, T. H. Lee, J. Lee, Clinical course of asymptomatic
25. J. Qin, C. You, Q. Lin, T. Hu, S. Yu, X.-H. Zhou, Estimation of incubation period distribution and mildly symptomatic patients with coronavirus disease admitted to community
of COVID-19 using disease onset forward time: A novel cross-sectional and forward treatment centers, South Korea. Emerg. Infect. Dis. 26, 2346–2352 (2020).
follow-up study. Sci. Adv. 6, eabc1202 (2020). 53. K. Mizumoto, K. Kagaya, A. Zarebski, G. Chowell, Estimating the asymptomatic proportion
26. S. M. Bartsch, K. J. O’Shea, M. C. Ferguson, M. E. Bottazzi, S. N. Cox, U. Strych, of coronavirus disease 2019 (COVID-19) cases on board the Diamond Princess cruise ship,
J. A. McKinnell, P. T. Wedlock, S. S. Siegmund, P. J. Hotez, B. Y. Lee, How efficacious must Yokohama, Japan, 2020. Euro Surveillance 25, 2000180 (2020).
a COVID-19 coronavirus vaccine be to prevent or stop an epidemic by itself. medRxiv 54. X. He, E. H. Y. Lau, P. Wu, X. Deng, J. Wang, X. Hao, Y. C. Lau, J. Y. Wong, Y. Guan, X. Tan,
2020.05.29.20117184 (2020). X. Mo, Y. Chen, B. Liao, W. Chen, F. Hu, Q. Zhang, M. Zhong, Y. Wu, L. Zhao, F. Zhang,
27. M. Makhoul, H. H. Ayoub, H. Chemaitelly, S. Seedat, G. R. Mumtaz, S. Al-Omari, B. J. Cowling, F. Li, G. M. Leung, Temporal dynamics in viral shedding and transmissibility
L. J. Abu-Raddad, Epidemiological impact of SARS-CoV-2 vaccination: Mathematical of COVID-19. Nat. Med. 26, 672–675 (2020).
modeling analyses. medRxiv 2020.04.19.20070805 (2020). 55. J. T. Wu, K. Leung, G. M. Leung, Nowcasting and forecasting the potential domestic
28. I. M. Longini, E. Ackerman, L. R. Elveback, An optimization model for influenza and international spread of the 2019-nCoV outbreak originating in Wuhan, China:
A epidemics. Math. Biosci. 38, 141–157 (1978). A modelling study. Lancet 395, 689–697 (2020).
29. J. Medlock, A. P. Galvani, Optimizing influenza vaccine distribution. Science 325, 56. S. Zhao, Q. Lin, J. Ran, S. S. Musa, G. Yang, W. Wang, Y. Lou, D. Gao, L. Yang, D. He,
1705–1708 (2009). M. H. Wang, Preliminary estimation of the basic reproduction number of novel
30. L. Matrajt, I. M. Longini Jr., Optimizing vaccine allocation at different points in time coronavirus (2019-nCoV) in China, from 2019 to 2020: A data-driven analysis in the early
during an epidemic. PLOS ONE 5, e13767 (2010). phase of the outbreak. Int. J. Infect. Dis. 92, 214–217 (2020).
31. F. J. Ibarrondo, J. A. Fulcher, D. Goodman-Meza, J. Elliott, C. Hofmann, M. A. Hausner,
K. G. Ferbas, N. H. Tobin, G. M. Aldrovandi, O. O. Yang, Rapid decay of anti–SARS-CoV-2 Acknowledgments: L.M. acknowledges M. Moore for helpful discussions regarding the
antibodies in persons with mild covid-19. N. Engl. J. Med. 383, 1085–1087 (2020). model structure and B. E. McGough for help with cluster computing. Funding: This work was
32. Q.-X. Long, X.-J. Tang, Q.-L. Shi, Q. Li, H.-J. Deng, J. Yuan, J.-L. Hu, W. Xu, Y. Zhang, F.-J. Lv, supported by the National Institute of Allergy and Infectious Diseases, 1 UM1 AI148684-01
K. Su, F. Zhang, J. Gong, B. Wu, X.-M. Liu, J.-J. Li, J.-F. Qiu, J. Chen, A.-L. Huang, Clinical (L.M., T.L., and E.R.B.). Scientific Computing Infrastructure at Fred Hutchinson Cancer Research
and immunological assessment of asymptomatic SARS-CoV-2 infections. Nat. Med. 26, Center was funded by NIH ORIP grant S10OD028685. Author contributions: L.M. and
1200–1204 (2020). E.R.B. conceived the study. L.M. developed the model. L.M. and J.E. wrote the optimization
33. Centers for Disease Control and Prevention, Assessing Risk Factors for Severe COVID-19 algorithms. L.M. and T.L. analyzed the data. L.M. produced the first draft of the manuscript. All
Illness (2020); https://www.cdc.gov/coronavirus/2019-ncov/covid-data/investigations- authors contributed to the final draft. Competing interests: The authors declare that they
discovery/assessing-risk-factors.html. have no competing interests. Data and materials availability: All data needed to evaluate
34. European Centre for Disease and Prevention and Control, COVID-19 Clusters and the conclusions in the paper are present in the paper and/or the Supplementary Materials.
Outbreaks in Occupational Settings in the EU/EEA and the UK (2020); www.ecdc.europa.eu/ Code is available at https://github.com/lulelita/vaccine_optimization. Additional data related
sites/default/files/documents/COVID-19-in-occupational-settings.pdf. to this paper may be requested from the authors.
35. M. W. Hooper, A. M. Nápoles, E. J. Pérez-Stable, COVID-19 and racial/ethnic disparities.
JAMA 323, 2466–2467 (2020).
Submitted 6 October 2020
36. L. S. Muñoz-Price, A. B. Nattinger, F. Rivera, R. Hanson, C. G. Gmehlin, A. Perez, S. Singh,
Accepted 17 December 2020
B. W. Buchan, N. A. Ledeboer, L. E. Pezzin, Racial disparities in incidence and outcomes
Published 3 February 2021
among patients with COVID-19. JAMA Netw. Open 3, e2021892 (2020).
10.1126/sciadv.abf1374
37. United States Census Bureau, Washington (2020); https://data.census.gov/cedsci/
profile?q=Washington&g=0400000US53.
38. Population Pyramid, Population of United States of America 2020 (2020); www. Citation: L. Matrajt, J. Eaton, T. Leung, E. R. Brown, Vaccine optimization for COVID-19: Who to
populationpyramid.net/united-states-of-america/2020/. vaccinate first?. Sci. Adv. 7, eabf1374 (2021).

Matrajt et al., Sci. Adv. 2021; 7 : eabf1374 3 February 2021 11 of 11

Vaccine optimization for COVID-19: Who to vaccinate first?
Laura Matrajt, Julia Eaton, Tiffany Leung and Elizabeth R. Brown

Sci Adv 7 (6), eabf1374.

DOI: 10.1126/sciadv.abf1374

Downloaded from http://advances.sciencemag.org/ on July 30, 2021

ARTICLE TOOLS http://advances.sciencemag.org/content/7/6/eabf1374

SUPPLEMENTARY http://advances.sciencemag.org/content/suppl/2021/02/01/7.6.eabf1374.DC1
MATERIALS

REFERENCES This article cites 33 articles, 3 of which you can access for free
http://advances.sciencemag.org/content/7/6/eabf1374#BIBL

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Use of this article is subject to the Terms of Service

Science Advances (ISSN 2375-2548) is published by the American Association for the Advancement of Science, 1200 New
York Avenue NW, Washington, DC 20005. The title Science Advances is a registered trademark of AAAS.
Copyright © 2021 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of
Science. No claim to original U.S. Government Works. Distributed under a Creative Commons Attribution NonCommercial
License 4.0 (CC BY-NC).