Testing and social distancing measures by state
Summary: there is a HUGE patchwork of different responses nationally in the U.S. across states in terms of frequency of testing, how much mobility (as a proxy for social distancing) has slowed, and likely in the # and % of cases identified. This variation is going to represent an enormous challenge with reopening because of new cases coming into regions that are able to achieve local control of viral spread.
The degree of social distancing and availability of testing vary quite a bit from state to state. No one knows at this stage exactly how much testing and exactly how much social distancing are needed to control viral spread in an area, although in general more of both seems better as an impression.
We took state level testing data (from here), along with state level mobility data (from Google) to see how things are playing out in different parts of the country. The following graphs plot coronavirus tests done per 1,000 population on the x-axis against % change in different measures of mobility (presumably measuring some degree of social distancing uptake) in: 1) Transit, 2) Retail & Recreation, and 3) Workplaces from mid-February through April 5th. Each state is identified by its postal abbreviation, and the size of the letters corresponds to the # of deaths that each state has reported as of April 7th.
In general, it looks like states that have done more tests (further to the right) have also had more decreases in mobility data (so stricter social distancing, generally) in all three of these areas (transit, retail & recreation, & workplaces). A lot of the states with more deaths (larger sized text on the graph) seem to be testing more, although not all of them. It makes sense to see more testing in harder hit areas as they are the ones dealing with more severe outbreaks, and I’ve been imagining that ascertainment of hospitalized cases is typically better than non-hospitalized cases (or all those folks who have died at home without testing for that matter). Connecticut and Michigan stand out a bit here as having more deaths but fewer per capita tests. You can see states like Nevada and Florida with a lot of tourism really dropping their activity levels dramatically in the transit and workplace categories.
Another way we were thinking to assess how much testing is happening relative to the severity of an outbreak in an area is to look at how many tests are being done for each positive result. These graphs look at # of tests done per positive result, again graphed against the same mobility measures as above.
States all the way to the left are the curve are getting very frequent positive tests when people are testing, while those way to the right are doing many tests for each case identified. The states with more deaths have fewer tests to identify each positive case, which again is not surprising in more severely impacted areas because it’s easy to find cases when they’re in your local hospital. Outside of the hardest hit areas there is quite a bit of variation in how many tests are being done to identify each case in addition to a lot of variation in how much mobility decreased in different areas.
We don’t know the threshold of how much of a decrease in mobility is effective — just HOW MUCH does society need to shut down to get a handle on infections in a local area. Reports from other counties have suggested that with other public health measures in place some degree of activity can occur and maintain control, although the risk of importing cases remains high everywhere in the world right now, which can again seed local transmission even in areas where there was previously good control.
I think the implications of all this variation in testing rates, in # of tests per positive result, and in % decrease in mobility is that we’re going to have a HUGE patchwork of local and state experiences with the virus that are going to make re-opening challenging because the virus does not respect state or country boundaries.
Finally, one more graph directly comparing testing rates per population with # of tests per positive result:
The states I worry most about here are the ones with 1) less testing per capita that 2) have fewer tests to get a positive result and 3) more deaths — Michigan and Georgia, for example, are concerning here as likely undertesting, with Indiana, Colorado, Illinois, and Connecticut in the mix as well.
Summary: there is a HUGE patchwork of different responses nationally in the U.S. across states in terms of frequency of testing, how much mobility (as a proxy for social distancing) has slowed, and likely in the # and % of cases identified. This variation is going to represent an enormous challenge with reopening because of new cases coming into regions that are able to achieve local control of viral spread.
