In the last of our rolling updates to our totalTestResults API field before we end data collection tomorrow, we switched totalTestResults for MT, NM, and WV from summing positive+negative to drawing data directly from totalTestsViral. covidtracking.com/about-data/tot…
These changes close the book on work we began in August 2020 to improve the data in totalTestResults. When our project began, most states shared positive and negative results only, so we summed those figures to calculate totals for every state. covidtracking.com/analysis-updat…
Switching each state’s totalTestResults from calculated to explicit numbers required getting complete historical testing data from states. We tried to get the data in units of tests—not people—because counting tests better captures testing volume.
In all, we managed to switch 45 of 56 US jurisdictions to explicit totals in totalTestResults. And 54 jurisdictions now post at least current (if not historical) data in units of tests. We are grateful to the efforts of state health officials who made these improvements possible.
More broadly, we are so grateful to the health officials who have worked inhuman hours for more than a year to collect and publish the best data they could, despite our nation’s rickety public health infrastructure—and who will keep publishing it for the foreseeable future.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
Our daily update is published. States reported 1.2 million tests, 41k cases, 40,212 hospitalized COVID-19 patients, and 839 deaths. This is our final day of data collection after a very long year.
The project was initially created to track testing. The first few days, states reported just a few thousand total tests. Today, states reported 1.2 million tests. The single-day high for the year was December 5 at 2.3 million. Cumulatively, we've tracked 363 million tests.
We ended up tracking other metrics. Cases reached heights we never could have imagined in the early days. The 7-day average got to 250 thousand cases per day in early January. Today, states reported the fewest number of cases since October 6, before the winter surge.
Our daily update is published. States reported 1.7M tests, 69k cases, 42,541 currently hospitalized, and 2,221 deaths.
This is our final weekday update. We'll tweet our last daily data this Sunday, though we will periodically post deeper analysis beyond that date.
Currently hospitalized is under 50 per million people in 8 states, up from only 2 states in early February.
7-day average cases are down over 10% week over week in all US regions save the Northeast. However, testing in the Northeast is up much more than cases in the same period.
The COVID Tracking Project is ending data collection this Sunday. As part of our wind down process, we are for the first time publishing our full data annotations, a set of structured metadata on how states define their COVID-19 metrics. covidtracking.com/analysis-updat…
Health data pipelines in the US are siloed, with each state running its own pipelines to collect data on COVID-19’s spread. In the absence of federal guidelines, and constrained by technical limitations of their systems, many of them use different data definitions.
Definitions are often relegated to small footnotes on state dashboards. But they make a big difference in the data: They can tip the scale between labeling a COVID-19 case as active versus recovered, or between counting or not counting some COVID-19 cases and deaths at all.
Since The COVID Tracking Project is winding down on March 7, we’re packaging up what we’ve learned about federal COVID-19 data in a 101 series. Up today: federal testing data. covidtracking.com/analysis-updat…
Unlike case, death, and hospitalization data, federal testing data doesn’t match well to the state data we collect. The discrepancies point to problems with both state and federal data sources. Our deep dive on that: covidtracking.com/analysis-updat…
Thankfully, federal testing data is standardized, unlike our patchwork testing dataset. But the data can be incomplete—especially in 5 states that still can’t submit data to the federal government:
For our API users: Yesterday, we switched totalTestResults to use values from totalTestsViral instead of being calculated from positive+negative in 4 states: IL, ME, MI, and SD.
These switches caused the totalTestResults field to increase by ~325k (cumulatively). The increases were driven by MI and SD, because our old totalTestResults counted MI positives and SD positives and negatives in units of unique people instead of specimens.
To get more testing history, we changed ME’s testing data source to one that counts residents only instead of residents and nonresidents. ME’s other COVID-19 data only counts residents, so the new source better matches the rest.