There’s a New York Times article today that makes some comments about percentage of positive tests in each state. It’s quite critical, and I think it whitewashes the challenge of building the surge capacity that’s needed to respond quickly. Most states had testing pretty well under control at the end of May, but with the massive increase in infections in the south, the sheer number of tests has become overwhelming. The US is currently running about 650K tests per day, i.e., about 4.5 million tests per week–and the areas that need the tests keep moving around the country.
In addition to the testing itself, I’ve been told the data management is becoming a challenge. That’s being conducted by government agencies that are perennially short staffed. I would imagine that social distancing requirements make that work even more challenging.
The percentage of virus tests that show positive results (“positivity”) is an indicator of the percentage of the total infections that is being identified through testing. The higher the positivity is, the lower the percentage of overall infections being detected is, all other factors being held constant. Most states started with high positivity (in some cases 30-40%), and many got down to 5% positivity or lower. Now, many states are seeing increases in positivity–in some cases, significant increases.
These graphs show the history of positivity in each state.
One interesting speculation, as testing has become so much more extensive, is that positivity becomes less accurate as a proxy for the percentage of infections tested, but it starts to become more representative of the level of infections in the general population. At this point that’s just speculation, but it’s something to watch.
Graph Anomalies. To a greater degree than any other type of graph I publish, these graphs are sensitive to anomalies in states’ data. Very high positivity in March through early April is usually a result of the data being reported and is accurate. High spikes in positivity in mid-May through the present are usually the result of the state adjusting its reporting — including removing large numbers of positive tests from its results, removing large numbers of negative tests, or removing large numbers of total tests. In other words, on one day a state will report a cumulative total of 250,000 tests. The next day it will report a cumulative total of 220,000 tests, implying that the day before it conducted -30,000 tests. These types of anomalies show up as spikes on the graphs. Days with negative numbers of tests reported are noted on the “Raw data” graphs.