What Testing Capacity Do We Need?

While there is near total agreement that the United States is not carrying out enough SARS-CoV-2 testing to enable it to safely relax social distancing measures, there is no agreed upon benchmark for how much testing we need to do to get there. Rather, several national-level estimates have been put forward based on different assumptions and targets. Here, we briefly review these estimates and compare them to current national and state testing levels. What we find is that while the national estimates vary wildly, the U.S. is still well off the mark no matter which benchmark you use. This is also true for most states, if the benchmark is applied to the state-level.

Existing Benchmarks

We looked at several national estimates, or potential benchmarks, that have been put forward:

1. A group from Harvard has looked at three scenarios for estimating daily testing capacity:
   1. Using the standard susceptible-infected-recovered model (SEIR)
   2. Using an “equilibrium model”, calibrated to protect hospital capacity
   3. Using the experiences of Taiwan and South Korea, two countries that have had success in controlling spread, to calculate a “best case scenario”
2. Scott Gottlieb and colleagues at AEI have talked about a roughly 1% of population estimate to derive the number of tests needed on a weekly basis (though have also suggested 750,000 as a minimum threshold once community transmission has been sufficiently reduced)
3. Paul Romer has estimated that a random selection of 7% of the population should be tested each day.

We summarize these in the table below and also include the share of the U.S. population that would need to be tested under each model (in a given week). As the table indicates, the models vary in their weekly testing target from 3 million to 160 million, accounting for 0.9% of the population for the lowest bound estimate to half of the population in the highest.

Another approach could be to estimate the amount of testing needed to identify all infections, based on reported cases. Currently, there are approximately 25,000 newly reported cases per day in the U.S. (the number reported on April 13, for example). We know that is an underestimate – maybe by a factor of 10, as some have suggested – which would mean that there could be 250,000 new infections per day. If we assume that about 20% of people tested are positive (according to the latest data from the COVID Tracking Project), that would mean conducting 1.25 million tests per day or 8.75 million per week, which is about 2.7% of the U.S. population being tested weekly.

(Note that none of these approaches includes any variation by region or state, positivity rate, doubling rate, or other variables, and is applied uniformly to the U.S. Obviously, these factors are quite important.)

Current Testing Levels Compared to Benchmarks

Now, comparing these ranges to current testing shows that the U.S. tested an additional 1 million people in the last week (April 6 to April 13). This is far below all of the estimated targets above and a smaller share of the population (0.3%) than even the lowest target (0.9%). Applying the population share targets to state testing capacity, we find that this is also the case in all states except one – Rhode Island, which tested 1.1%; Louisiana and New York each had tested 0.8% of their populations in the past week, just below the minimum benchmark. Because testing resources have been limited, the focus has been on people who are sick and hospitalized and frontline health care workers, so it is not surprising that Louisiana and New York have higher rates given the scale of infection in those communities.

There is not yet consensus over what approach to testing is required for social distancing measures to be loosened, or exactly how much capacity is needed. But, by any measure, it is clear that we are far from being able to do enough tests to enable us to move to the next phase of responding to the pandemic in states across the country.

Updated on April 27, 2020