A number of friends and visitors to the site have noticed a discrepancy in how we count “new” cases and how the state of Georgia is counting new cases.
For example, on April 21st, Covid Mapping Project reports 934 “new” cases. Covid Tracking, a companion site run by a team from The Atlantic, doesn’t report differentials but they do report a day by day tracking of cases. On April 20, they report Georgia had 18,947 positive cases and on April 21st they reported 19,881 cases. A differential of 934. Same as us.
However, if you got the GA DPH website (where most of us source our data), you’ll see 10 “new” cases. That was later revised to 84. And then 138. It's a moving number that will likely change every day for at least a week. Quite different than the 934 we’re reporting.
Here’s the sitch:
If you look closely at the DPH website, there’s a tiny bit of text that essential says “confirmed cases based on earliest date of symptoms”. I’m not epidemiologist so I consulted with a close friend and colleague that we work with on a number of projects and this is what he said about this definition: “there’s not really a problem with defining it that way.” He encouraged me to read up on how CDC defines cases and counts here: https://www.cdc.gov/csels/dsepd/ss1978/lesson1/section5.html. The TL;DR of this CDC page is this: how you count things can vary and it really matters in how you define things. So if a positive case is counted when symptoms arose? Cool. If it’s when their labs came back as positive? That’s cool too. Consistency is key.
The challenge with this way of defining new cases as it relates to COVID-19 is that the reporting of new case counts to the public will always look bell shaped. That’s because there’s a time lag from the time a person realizes they might be sick, to going to the hospital or a testing location, getting the test, and getting a positive. The state will ALWAYS be backwards adjusting their numbers as new positives come in. For example, there is no way only 10 people in a state of 10M got COVID 19 on April 21st. There were 10 people who saw symptoms and were diagnosed on the 21st (if there’s another way to read that, let me know). This might be fine from an epidemiological perspective, but as far as the public is concerned they will always have to go back in time to see how things are really going.
Our goal with The Covid Mapping Project is to this data on the GPH website (and other states’ websites) and visualize it in a manner that is easy to consume. So in Georgia’s case, we will continue to report number of positive tests and the differentials as they’re reported on the GPH website. We’ll adjust our labeling to say “Number of Positive Tests” which reflects what the state website says.
