Saturday, June 13, 2020

Can You Trust COVID-19 Case Counts?

A couple weeks ago, I wrote about a 2-week decline in COVID-19 case counts in Los Angeles, based on data from the Los Angeles County Department of Public Health COVID-19 Dashboard. Here is the same figure from my May 30 blog for your convenience:


It seemed like things were getting better until I looked at data from USA Facts. After downloading the data and generating a 7-day rolling average, I got this result:


The 2-week decline in COVID-19 case counts was nowhere to be seen, and in fact it showed a steady increase in new cases/day. Thinking that this must be due to a discrepancy between the data collection methods of LA County and USA Facts, I went back to the LA County data to update my figures and got this result:


The the 2-week decline in COVID-19 case counts is nowhere to be seen in the LA County data download! What happened to our impression that Los Angeles was on the road to recovery? Did the Sith Lord who erased Kamino from the Jedi archives fudge the LA County COVID-19 data too? To further investigate, I compared the LA County data from today (June 13) against a copy of the LA County data that I downloaded when I wrote my other blog (May 30) 2 weeks ago. I saw differences in the case counts and plotted them over time:


It turns out that new cases/day generally stayed the same or increased, although on some days the numbers went down. However, in the last 2 weeks before May 30, there was a significant increase in cases/day between the May 30 download and the June 13 download, which completely erased the download trend in the 7-day rolling average number of cases/day. I suspect that this is due to a backlog of statistical data that had yet to be entered, and given that numbers went down on some days, maybe some data were re-categorized into other days or errors were corrected. Whatever the case, from these observations, I conclude that you can’t really trust the numbers until a couple weeks later when the data stewards are mostly caught up with data entry and the numbers stabilize.

A couple weeks ago, I thought that Los Angeles was starting to crush the curve, but unfortunately the COVID-19 numbers tell us that we are not, and in fact things are getting worse.

No comments:

Post a Comment