Considering the role of data in achieving vaccine equity
As agencies throughout the country distribute vaccines for COVID-19, the list of people with important roles to play in the process is long. In addition to the officials planning it all and the frontline workers actually distributing the vaccine, these roles include transportation and logistics personnel, warehousing, security, and so forth. There is even a fractional role for people like me — data scientists — because data management and analysis will allow everyone to plan better, operate more efficiently, and vaccinate more equitably. (It’s that last part, as my font formatting implies, that prompts me to write this blog post).
There’s one component of the challenge of COVID-19 vaccine data that I worry is not being discussed enough. It has less to do with who has received a vaccine and more to do with who is eligible — or more specifically who is eligible in a given geography for a given phase of the distribution plan. If we don't know how many people are eligible, we won't know who is being missed. In other words, we need to estimate the denominator.
This is a timely question that is vital to eliminating disparities in vaccination, including but not limited to racial disparities. It is surely not enough to assume that vaccine utilization rates correspond neatly to eligibility rates, nor should we dismiss the question by simply attributing any discrepancy between utilization and eligibility to vaccine refusal. There are many confounding factors — including varying methods of information dissemination, physical/transportation barriers, constraints like childcare access, and so forth — that could contribute to underutilization in a given community.
Some elements of the process of quantifying eligibility are fairly straightforward. The American Community Survey from the US Census Bureau is an excellent survey product that provides detailed estimates of demographic data. It is relatively simple, therefore, to estimate how many people will qualify in a given area by age, and even to look at those age groups across race, income, and other characteristics. We need to do that. But we also need to do more.
Who Are Essential Workers?
A much trickier part of the puzzle, as we approach phase 1B and 1C of the distribution plan, would require us to estimate the number of people who qualify for vaccination through their jobs, and to associate race and other economic and demographic characteristics in the process. Broadly speaking, these people are the so-called “essential workers” — often referenced as such by journalists and public officials alike. The problem is, while that phrase has come to be widely used, those words do not come with an obvious or widely understood definition.
Earlier in the pandemic, determining what and who was economically essential was a task that effectively fell to state and local governments, as they created specific business closure orders and other operational restrictions. The resulting definitions varied widely in both their content and the clarity with which they were conveyed.
Now, as we move into the vaccine distribution, the task of setting definitions for essential work is being led by the federal government, and it has fallen (not very intuitively, in my opinion) to an agency within the Department of Homeland Security called CISA — the Cybersecurity and Infrastructure Security Agency. They have set a textual definition of essential workers that is 16 pages in length. You can read it here.
Although they are admirably detailed in words, there are two issues that make it challenging for analysts to quantify the corresponding number of workers in a given place who fall into these descriptions:
The definitions involve the intersection of two economic concepts, “occupations” and “industries” — i.e., what one does and where one does it — either of which might make a worker essential. This interaction makes analysis difficult, especially for small areas.
The definitions aren’t mapped to standard occupational or industry codes, making the process of analysis both tedious and challenging.
There are workarounds to be had. One simple proxy for essential status comes from the state of Pennsylvania. Governor Wolf’s well-organized March/April 2020 closure orders are mapped to standard industry codes (NAICS). This gives us an opportunity to use publicly available economic data to put together estimates at a county scale of how many workers are currently employed in those essential (or, in the language of the original order, “life-sustaining”) industries.
Those estimates allow us to say a few interesting things about the characteristics and distribution of the essential workforce in the state. One important takeaway that rises to the top of that list is that jobs held by people of color are more likely to be “essential” than jobs held by white workers. This is true statewide, as well as specifically in 61 of 67 counties, despite wide variation in industry concentration and demographics.
The map below captures the corresponding data.
(Click on a county to see specific estimates and demographic data.)
A Question of Equity
The above analysis can help us to look for disparities in Pennsylvania during the vaccination process across race as well as income and other characteristics, but these estimates are far from ideal. We need more methodologically robust numbers. We need them across the country, developed in partnership with the officials in charge of vaccine distribution, to guide our understanding of gaps in the process. This needs to happen quickly. Many aspects of the response to this pandemic — including the provision of both public health and economic resources — have not been handled equitably. People of color and low-income workers have disproportionately borne the health and economic impacts of the crisis, and absent much more active and thoughtful accounting, there is no reason to assume those trends will change with the vaccine distribution.
But we can do better now than we did before, and this seems like an altogether appropriate day for that message. 🇺🇸