The COVAX Allocation Algorithm for Vaccine Distribution
Newsletter Edition #36 [The Friday Deep Dives]
It is with great responsibility and pleasure that we are sharing our first post with the earliest supporters of Geneva Health Files.
We hope to not only meet your expectations but also deliver on the promise to report on the most significant matters in global health with care and commitment.
This engagement reveals two important insights: one that readers value original content; second that there is a demand for closer scrutiny of power relationships in global health.
We are truly excited to share this journey with you. Thank you for making this possible.
This week we take a look at the allocation algorithm determining the vaccine distribution for COVID-19 under the aegis of the COVAX Facility. Please feel free to comment and respond to our work.
Please note that we are making an exception and will make this exclusive edition public after a few days, in order to accommodate regular readers who are in the process of making a transition into paid subscriptions. Thank you for understanding.
It has been brought to our attention that some readers have questions about accessing our content via the paid subscription. Please write to us with any clarifications you may have. Here is a helpful link from Substack responding to reader queries.
Until next week!
Write to us: firstname.lastname@example.org or email@example.com; Follow us on Twitter: @filesgeneva
1. Story of the week
WHO’S ALLOCATION ALGORITHM: COVID-19 VACCINES DISTRIBUTION
The logic underpinning the distribution of COVID-19 vaccines secured by The COVAX Facility
The COVAX Facility is using an allocation algorithm to distribute COVID-19 vaccines globally. The underlying logic determines how countries line up to receive vaccines in a period of acute scarcity when less than 250 million vaccine doses expected to be distributed among more than 7 billion people worldwide in the first half of 2021 through the Facility.
COVAX published its first round of allocations earlier this week that seeks to deliver 237 million doses of AstraZeneca vaccines to 142 countries.
According to WHO’s allocation mechanism, the first phase will aim for countries to receive doses covering the same proportion of their population over time. The initial goal is to reach a coverage of 3% of the population in countries and eventually extending vaccination up to 20% of the population.
We bring you an analysis on the assumptions in the algorithm and how it allocates vaccines among countries during the current first phase of the roll-out of the vaccines globally. The allocation algorithm follows the WHO allocation framework first released in September 2020.
WHO has explained how the allocation of COVAX vaccines is structured, why an algorithm is needed to allocate vaccines, when and how the algorithm is run, the logic of the algorithm, how it is optimized and the implications of pricing, timelines and delivery for vaccines. It also addresses aspects of governance, transparency and safety of data.
Image Credit: © UNICEF/UN0420494/Krishnan
Employees packing boxes with COVID-19 vaccine at the packaging and dispatch department in Pune, Maharashtra, India, on Tuesday, Feb. 23, 2021. Source: Gavi Website
THE NEED FOR AN ALLOCATION MECHANISM
A document dated February 15, 2021, classified as internal, clarifies that this description of allocation operationalization is likely subject to changes on account of shifts in supply, demand, or policy. When Geneva Health Files asked WHO for information on the algorithm, this document was shared.
During the first phase, the COVAX Facility seeks to proportionally allocate doses to up to 20% of the total population in participating countries. It is understood that “The rate at which participants receive vaccines depends on country readiness and the availability of doses (not on threat and vulnerability),” according to the document explaining the algorithm.
As per current supply projections, the first phase will unfold for most of 2021. According to WHO, in the second phase, there will be weighted allocation beyond 20%, where participants will receive doses at variable rates, based on consideration of vulnerability and COVID-19 threat. (This algorithm in its current form will not be applicable for the second phase. Adjustments are expected.)
WHY THE 20% TARGET?
The evolution of the pandemic suggests that countries will likely need full coverage to get to herd immunity. Critics have questioned the 20% target of coverage and want to know how the COVAX Facility seeks to address the need for covering a greater percentage of the population in affected countries.
According to WHO, based on information including the epidemiological impact of the virus, in terms of mortality rates across age groups and frontline workers, the groups most at risk were: adults above 65 years of age and adults with a series of health conditions. Considering the distribution across countries of these age groups and risk groups, and frontline healthcare workers, it was agreed that an initial objective to vaccinate 20% of the population would help to reach these high risk and essential populations.
In response to our queries on this, WHO said:
“WHO recognized ever since the early development of the global framework that the percentage of at-risk populations is variable across different countries; and it is important to note that the 20% target is considered a floor for an initial allocation that should increase as soon as more product becomes available. In choosing national prioritized target groups, countries perform their own assessment and are advised to rely on latest evidence-based SAGE recommendations. SAGE will also periodically provide advice in light of new evidence on variants impacts on COVID-19 vaccines.”
(SAGE is Strategic Advisory Group of Experts on Immunization advises WHO.)
Asked whether WHO will revisit this 20% figure in light of the variants of SARS-Cov-2, WHO said, “There is no need to review phase 1 allocation since in the meantime SAGE has provided recommendations for priority groups vaccination. WHO is monitoring the evolution and the impact of variants on product’s efficacy and the allocation group stands ready to adapt the allocation logic should evidence arise and recommendations from SAGE or WHO Prequalification (PQ) team arrive.”
USING AN ALGORITHM TO ALLOCATE VACCINES
According to the document explaining the algorithm, “The allocation is relying on rapid computations of various data streams which necessitate an IT solution…. An allocation round is a complex optimization problem requiring the distribution of multiple vaccines for [nearly] 190 participants.”
The algorithm processes the country and vaccine data, and optimizes the allocation of vaccines driven by three objectives: equality of doses received proportionally to population; matching countries’ preferences for products; and the consistency of product received (allocating a single product when possible.)
“This complexity increases as the COVAX portfolio grows in supply and in number of vaccines to be allocated, therefore an allocation algorithm was warranted to support allocation processes,” the document says.
THE ALLOCATION ALGORITHM: HOW DOES IT WORK?
The Allocation Algorithm is run by the Joint Allocation Taskforce (JAT) that has representation from both WHO and Gavi - The Vaccine Alliance. Each allocation is determined by the Global Regulatory Approval (EUL/PQ and/or SRA), a SAGE policy recommendation and available supply, the explainer says.
The Allocation Algorithm Logic involves establishing a supply forecast, mapping demand constraints, assessing the “demand envelope”, matching supply to demand preferences and finally, working out a delivery sequence.
In order to arrive at an optimal solution, a number of factors feed into the system.
One is the kind of information on vaccines in an allocation round including the supply forecast and the characteristics of vaccines (platform, regulatory status, cold chain storage requirement, dose regimen, price per dose, and vial size). The limitations on “destinations for certain manufacturers or deals (e.g. based on licensing), and minimum shipment and batch/pallet size,” are also factors, according to the document.
Indicators on demand constraints include country readiness to receive doses and kinds of prices and conditions in countries’ contracts with COVAX. (Committed purchase agreements versus optional opt-outs as exercised by self-financing countries.)
Similarly establishing a “demand envelope” is subject to the number of doses requested by countries, the total population, minimum shipment size among other considerations.
The final two steps of the algorithm include matching supply to demand preferences and ensuring a delivery sequence.
For optimal results, the algorithm gives most weight to the equality objectives over product consistency and product preference, the document explains.
“This means that if a solution is more equal and a bit less aligned with another objective, the algorithm will prefer it over a solution which is a bit less equal but more aligned with the other objectives.” (For the first allocation round, the weights for the equality, consistency and preference objectives are set at 50%, 25% and 25% respectively.)
Optimizing for minimizing inequality could mean population coverage differences between participants over time, an example cited in the explainer.
On working out a delivery sequence, the document says, “The algorithm output will contain a ranking of participants to serve as a prioritization list for the delivery sequence. This sequence will be followed to fulfil and ship allocations to participants in case of competition for capacity.”
THE MECHANICS OF OPTIMIZATION:
This algorithm seeks to determine vaccine allocation mathematically while optimizing for goals on equality, preference, and consistency, the document says.
“Optimization is a quantitative method for finding the best solution to a problem that has many possible solutions. Three components are used to define the problem’s goals, feasible solutions, and decision variables,” according to the explainer.
“The general approach to solving an optimization problem involves changing the variables to find solutions that satisfy all constraints. The objectives are then used to evaluate the best solution out of all possible solutions,” the document adds. (In other words, this process involves streamlining ways of finding a feasible solution by laying down rules on how goals can be achieved instead of specifying a goal itself.)
The allocation algorithm uses linear programming to find an optimal solution. The details of the search algorithm are proprietary, the document says.
In response to our queries on where and how was the algorithm designed, a WHO spokesperson said, “The code has been developed by Linksbridge SPC (a research institute in Seattle in USA) following directions by WHO.” The algorithm is the translation of the logic that the Fair Allocation group (composed of representative from WHO, Gavi, CEPI and lately UNICEF and PAHO RF) has discussed and agreed during the months of the design phase, the spokesperson added.
“…The algorithm converts the input data (supply forecasts, vaccine characteristics, demand constraints, and demand envelopes) and optimization goals into linear functions to act as constraints and objectives, respectively. It then defines the variables that can change,” the document says. (On how the algorithm actual solves the optimization problem, refer to the document for details.)
THE ALGORITHM AND EQUITABLE ALLOCATION
The algorithm accords priority to equitable access as an outcome.
“Given an allocation round supply forecast, the algorithm is designed to identify the highest total coverage level it can achieve in as many participants as possible. This coverage level is defined as the achievable equity line…”, the document says.
According to the explainer, “For most participants, the algorithm will increase their total coverage at roughly equal rates.” Some countries may fall behind if they miss a round, for example if a country is not ready. But “the algorithm will allocate extra doses to bring the participant’s total coverage up to the current round’s achievable equity line.” There are a number of other scenarios described that are influenced by factors such as shipment size or countries with small populations, for example.
Other optimization goals of the algorithm include vaccine preferences and product consistency. These goals are scored and incentivized to achieve an overall optimal outcome. (A higher score means a vaccine product is more desirable to a participant, the document explains. In order to prevent countries from receiving multiple vaccines with different characteristics, the algorithm imposes a penalty for assigning multiple products to a country, incentivizing the model to select single products whenever possible, according to the document.)
PRICING OF DOSES AND THE ALGORITHM
COVAX explains that when countries communicated their preferences, price was one of the 12 characteristics that were ranked by countries. This is taken into account when a country and a vaccine are matched.
Therefore, if price considerations were ranked high in the preferences provided by countries, the algorithm will aim to match the participant with lower priced vaccines.
But the document also clarifies: “However, as the objectives of equity of coverage, vaccine consistency and vaccine preference match can be competing objectives, there is no guarantee that the algorithm will allocate a less expensive vaccine based on preference alone.”
Finally, delivery of the allocated vaccines is determined by regulatory timelines in countries, arrangements with manufacturers and other variables. (WHO clarified to us that “the stratified randomization of countries for shipment purpose happens post allocation and does not consider the price of products.”)
The Vaccine Allocation Decision prepared by the Joint Allocation Taskforce contains both the allocated number of doses per vaccine for each participant as well as a ranking of participants for the delivery sequence, the document explains.
GOVERNANCE AND TRANSPARENCY
The Independent Allocation Validation Group (IAVG) and the Joint Allocation Taskforce (JAT) work on vaccines distribution within the COVAX Facility.
The document lays out how the governance processes tie into the determination of vaccine distribution:
“The JAT prepares Vaccine Allocation Decision (VAD) proposals based on the Allocation Algorithm output. This is passed on to the IAVG. The IAVG then validates this proposal ensuring it is technically informed and free from conflict of interest. The validated VAD would then be passed on to the COVAX Facility, procurement agencies and self-procuring participants to be implemented.” (To know more about these governance bodies in the COVAX Facility, see our earlier story: A peek into the COVAX Machine.)
According to WHO, the finalised code of the Allocation Algorithm is expected to be published in the coming months.
It is not clear for example, how dose-sharing, or trading in vaccines - other mechanisms under the COVAX Facility will affect this allocation mechanism or how it feeds into this algorithm.
Fundamental questions on how countries were classified into categories under the COVAX Facility have long been raised. According to GAVI, the 92 AMC-eligible economies includes all economies with Gross National Income (GNI) per capita under US$ 4,000 plus other World Bank International Development Association (IDA)-eligible economies. But experts say that GNI per capita only captures income in a country and not how much of that income is spent on health for example.
Sara L M Davis, Senior Researcher at the Global Health Centre at The Graduate Institute in Geneva, said, “Any allocation methodology is essentially rationing healthcare; it's deciding who lives and who dies. It's a terrible problem to have.”
Davis who authored, The Uncounted: Politics of Data in Global Health, raised some questions. “If the allocation methodology does not address some aspects, it should perhaps be publicly explained by the ACT Accelerator: How will ACT-A verify that countries that get COVAX vaccines are prioritizing health care workers, the elderly, and other vulnerable groups, and that publicly-financed vaccines are not being captured by elites? What accountability measures are in place?”
MSF had also raised questions on the implications of certain countries with rising burden on COVID-19 being excluded from receiving allocations on account of limited cold chain facilities for example.
“MSF calls on vaccine manufacturers to ensure that priority is given to those countries that are in urgent need of protecting their healthcare staff,” says Isabelle Defourny, MSF Director of Operations. “MSF stands ready to provide logistical support to high priority countries that were denied access to the Pfizer/BioNTech vaccine through COVAX because of their limited cold chain management capacity.”
COVAX: THE DEMAND SIDE OF THE EQUATION
Ultimately the COVAX Facility addresses only the demand side of the vaccines access problem.
In parallel, political, discussions on the TRIPS Waiver proposal at the WTO, India, one of the proponents of the proposal, reportedly highlighted that COVAX is a demand side initiative and does not address supply side constraints.
At the General Council meeting earlier this week, India cautioned that if members “do not address supply side issues, then we will not be able to increase the production of vaccines,” arguing that the “Waiver will help the COVAX mechanism by augmenting the manufacturing capacity globally”, the Third World Network reported.
This week Gavi was asked whether it would support mechanisms such as the intellectual property waiver to address vaccine shortages. The question went unanswered.
Like this story? Tell us what you think.
2. A Photo Essay
Health by train in the Russian Federation: WHO/Europe
“Travel with those giving and receiving vital health services on board the Saint Lukas medical train in the Kashtan Bogotol region of Siberia.”
Thank you for being a committed reader of Geneva Health Files. Your engagement helps us remain independent and powers this investigative journalistic initiative at the forefront of global health policy-making.