Recently, due to complaints about how data is presented on Georgia DPH COVID Dashboard, they added a section showing new data reported daily, along with the ability to graph cases and deaths by date reported. Previously, they only shared totals for cases/deaths/etc., not daily changes. They also only showed graphs of cases by date of onset (or date of lab result if no symptom onset available), and deaths by date of death. These are welcome changes, but have led to some additional questions.
Graphing cases and deaths by date of onset/death are more accurate ways to view the data and analyze the epidemic curve, but they will always be incomplete in recent days due to incomplete data reporting, which can be confusing to some. As a result, I have always graphed cases and deaths by date of report on my web site, using the method of subtracting the previous day’s total reported cases from the current day’s reported cases, and displaying the difference. This way the numbers never change. For a simplified example, if the total cases was 1,000 on Monday, and 1,050 on Tuesday, the change reported for Tuesday will be 50, and that number will never change.
Understanding the “Reported Today” Numbers
However, Georgia DPH is using a slightly different method to show how many cases and deaths are reported daily. In the “Reported Today” section of their site, they show the number of new additions, without subtracting out any data that was removed. Their explanation for this section states:
These numbers show cases, hospitalizations, and deaths that have been newly reported since the previous day’s posting. These are included in the overall totals above. The cases reported today may not equal the difference in total cases between yesterday and today because previously reported cases may be removed as duplicate reports are corrected or may be reclassified as additional information is collected during case investigation.
In my previous example above, a change of 50 cases from one day to the next doesn’t necessarily mean there were exactly 50 new cases added that day. It could mean there were 53 new cases added, and 3 cases were removed (because they were duplicates or errors or some other reason). So while most sources will list this as 50 new cases reported (the net change), Georgia DPH lists it at 53 new cases on Tuesday.
How does this affect the “Date of Report” graphs?
This understanding of when cases and deaths are reported also affects the “Date of Report” graphs. Continuing with my example above, on the Georgia DPH web site, they would graph the 53 new cases reported on Tuesday. However, if the 3 cases that were removed Tuesday had been previously added on Monday, they would remove 3 cases from the number reported on Monday. So the numbers on the by report date graphs also change from day to day. This is especially noticeable on the county graphs, where things shift more often because they might learn that someone lived in a different county than initially reported. So a new case added on Tuesday could be initially listed in Cobb County, but then if DPH gets more information at a later date, and find that the person actually lives in Douglas County, the number reported for Cobb County would go down by one on that Tuesday and the number reported for Douglas County would go up one on that Tuesday, even if the county correction was made on Friday.
Basically, the “Date of Report” graph is really a graph of all cases by the date the case is reported to DPH, not a graph of the daily changes in DPH reported totals. If DPH finds out about 53 new cases on Tuesday, they will show 53 for Tuesday, but if one of those Tuesday cases is later removed, they will change Tuesday’s total to 52. And if one of those Tuesday cases is moved from one county to another, the county totals for Tuesday will change (up in one county and down in another). This is not how anyone would expect a by report date graph to work, so I wanted to provide some clarity here.
A Real-World Example & Confirmation from DPH
Here’s an example with real data from July 28, the first day of the new changes to the DPH web site:
- On July 28, GA DPH reported 175,052 cases, an increase of 4,209 from the previous day’s total of 170,843. (170,843 + 4,209 = 175,052)
- On the web site for July 28, GA DPH listed 4,293 new cases in the “reported today” section. That means they added 4,293 cases, but removed 84 previously entered cases for various reasons. (170,843 + 4,293 – 84 = 175,052)
- On July 28, the “Cases by Date Reported” graph, showed 4,293 on for July 28.
- On August 10, the “Cases by Date Reported” graph now shows 4,271 for July 28. This means, of the 4,293 new cases added on July 28, 22 have since been removed. (4,293 – 22 = 4,271)
Reporter Nick Wooten with the Ledger-Enquirer contacted GA DPH to get confirmation on the explanation for this seeming discrepancy.
Excerpted from the article linked in the Tweet above:
Cases: 175,052 (+4,293 reported today). The number of new cases reported in a day may not match the difference in total cases over a 24-hour period. This occurs because previously reported cases may be removed as duplicate reports are corrected. An older confirmed case may also be reclassified as additional information is collected during an investigation.
How Data is Graphed on Covid-Georgia.com
Tests, cases, and deaths are graphed on this web site by the date reported, using the net change each day, so numbers from previous dates never change. That is also how the number of new tests, cases, and deaths are calculated and displayed at the top of the Today in Georgia page. I also provide a graph of deaths by date of death on my Today in Georgia page (directly below the deaths by date of report graph). In addition, my Weekly Overview has sections where I graph cases and deaths per week by both the reported date and the actual date (symptom onset for cases, date of death for deaths).