What’s the problem with the DPH graphs?

It depends on who you ask. The graphs of cases and deaths over time on the Georgia DPH Status Report have been the subject of much debate and confusion. It’s part of the reason I started tracking the data myself, and then why I made this site. It’s impossible to see the current trends for the past 2 weeks using their graphs, because those numbers are all incomplete and subject to change. (And they do change – significantly – over the two week period.)

What’s the reason for the lag?

If you read the explanation of these graphs on the web site, and pay attention to all their warnings about the data being preliminary, you will know there is a lag in reporting for current dates. This is because Georgia DPH is tracking cases and deaths the way epidemiologists commonly track cases and deaths – that is, they track cases by the date of symptom onset (when available), and in the case of deaths, by actual date of death (not the date of the coroner’s report or the day the DPH found out about the death).

That’s best when you only need to look at trends from a few weeks/months ago, but is much less helpful when you want to know trends from the couple of days or weeks. That’s why I started recording and graphing the number of new reports of cases and deaths from the DPH each day. For example, on May 15th, they reported an increase of 795 cases – from 35,977 on May 14th to 36,772 on May 15th (per 7pm updates). But if you look at the graph of new cases on the DPH site on the morning of May 16th, it only shows 40 cases, which seems confusing. But if you understand the graph, it makes sense. Most of the 795 cases reported yesterday were placed on the graph on earlier dates when the person actually got sick or when they were tested.

For more information about understanding the GA DPH site, read my page Understanding the GA DPH Data.

Are the graphs purposely misleading?

Some people interpret this backdating and lag in reporting for recent dates as suspicious, but it’s just the nature of the recording the data with the most accurate possible date.

The current web site includes ample warnings that the data in this 14-day window is pending, but despite that, many people continue to look at the downward trend in the last 14 days as evidence that cases or deaths are declining. Don’t get me wrong – cases and deaths are generally declining, but not as much as the last 14 days of that graph shows. You can still see the decline to the left of that line, or look at my graphs that are done by date of reporting, and not backdated like these graphs to get a better idea of more recent trends.

Georgia DPH graph of new cases with 14-day window warnings
Screenshot taken from the GA DPH web site on 5/16 at 10am. (Red markup is mine.)

Critics have called for the Georgia DPH to remove these graphs because they are misleading, and complained that you have to read the “fine print” to understand the numbers aren’t complete. Maybe that’s the only way to avoid misinterpretation, but I hope they keep them. There is value in the way they are reporting cases and deaths more accurately (in the long term) than just when the paperwork gets to their office.

And I think they are doing just about everything they can to make it clear that this data isn’t complete. That 14-day window warning is in large print directly below the graph, the numbers themselves say “preliminary” and there is a line marked “14-day window” where everything to the right of it is faded out, and the lines aren’t connected. It’s hard to miss. I think journalists who use this graph as evidence of a sharp downturn are being intentionally misleading.

When did they start graphing cases and deaths this way?

There have been a lot of articles critical of the DPH for “changing” to this method of reporting. However, they’ve shown cases and deaths this way since they first added time-based graphs to their daily status report on April 12th. Then they did have “fine print” about the reporting lag, and didn’t have the 14-day window labeled in any way. All they’ve done since then is make the graphs bigger, more interactive, and even more clear that the recent data is incomplete. They never changed the methodology for the data was being graphed.

The following images were taken from screenshots archived by The COVID Tracking Project.

On April 13th, they added the daily deaths graph.

On April 16th, they added the daily confirmed cases graph.