Weekly Overview

Week Ending February 27: Cases, hospitalizations, test positivity, and ER visits for COVID-Like Illness (CLI) continue to drop – these metrics are all down by over 60% from the peak. Cases were artificially high this week due to dumping of old cases for several days straight. Death reports finally showing a significant drop. Deaths by date of death peaked in mid-January, but reporting of deaths from December and January lagged a lot.

Highlights from the Week

  • A lot of old cases were reported this week. Here’s a thread I wrote on Twitter explaining it in more detail. I’m trying to get some media attention to this as it is making it appear that our case declines have leveled off when they haven’t. The 7-day average of confirmed cases is inflated by about 300 cases/day. Without the old cases, we would have seen a 25% decline in confirmed cases from previous week instead of a 5% decline. Antigen cases suffered a similar issue with reporting of old cases this week.
  • Vaccinations continue to make good progress in Georgia. About 60% of seniors have received at least their first shot, and we will soon be expanding to allow teachers and some others to be vaccinated. For more about the vaccine eligibility and how to schedule, go to Georgia DPH COVID-19 Vaccine page. To see current progress of vaccines, visit the Georgia COVID-19 Vaccine Status Dashboard, which shows we have administered ~2 million vaccines as of Feb. 27 – including ~270K reported in the past week.
  • I removed the new Hospitalization and ICU case graphs from this page, because these metrics seem to be suffering from significant reporting delay. The daily numbers DO NOT REFLECT the day the person was admitted, and they fluctuate wildly from weekdays to weekends due to reporting inconsistencies. The GA DPH arcgis site graphs hospitalization cases by date of onset, and based that graph, you can see that new hospitalization cases had a peak onset around New Year’s Day. ICU cases peaked a little later, per HHS hospital data. I get frustrated with the reporting of how many people were hospitalized “in the past 24 hours” as that is not accurate. It’s just when it was reported.
  • Death reports dropped sharply late in the week. Deaths by date of death peaked in mid-January, and hospitalizations have continued downward, so deaths should continue to decline as well.
  • I am still focusing my reporting and analysis on confirmed cases and deaths, not antigen cases and probable deaths, although we got new data from GA DPH, so I am starting to add them to some graphs. The PCR test (used to identify confirmed cases) is considered the gold standard; however, these tests are extremely sensitive and someone can test positive well after they are no longer infectious (up to 3 months, per the CDC). This NY Times article brought the issue to light months ago, and Florida recently started requiring labs in the state to report Ct values, although they are not reporting that data publicly at this time. Rapid tests (antigen tests) have different accuracy issues – they are less sensitive so may not pick up infections as early as PCR tests, but they are also less specific, so 20% of positives may be false positives (particularly in asymptomatic people).
I’ve had several followers who use my web site ask how they could thank me for the work I’m doing, so I’ve set up a Virtual Food Drive benefiting the Atlanta Community Food Bank to help feed families in need during this difficult year. I already met my initial goal, so I doubled it. Thanks to everyone for your support. Click here to donate!

The visualization below shows multiple COVID metrics on the same graph, based on their weekly values as a percentage from the summer peak. Early in the summer, you can first see a peak in cases (by symptom onset) and ER visits with COVID-Like Illness, followed by a peak in hospital admits and current hospitalizations, and then a peak in deaths (by date of death). As you can see, these metrics have continued to follow similar paths. Everything is dropping in unison, even faster than it did in the summer. (NOTE: I do not show cases or deaths for the previous two weeks as the data is still incomplete.)

Cases & Testing

My preference to focus on overall trends and indicators other than raw case counts, but many thresholds for risk levels are based on reported case numbers, so I’m adding some additional tracking and context around these numbers.

Harvard has risk levels based on daily average case rates. We are currently in their red zone for confirmed cases (daily average >25 cases per 100K). As of Saturday, Feb. 27, we are at 18 daily average confirmed cases per 100K (down from 19 last week). If we add in antigen positive cases, we are at 27.5 daily average total cases per 100K (down from 28 last week).

Using a different metric, we are in the White House Coronavirus Task Force red zone for new cases. The WHCTF red zone is over 100 cases per 100K in a week or more. I track our weekly cases per 100K metric daily on my Today in Georgia page, which is updated daily. The rates are labeled every Friday, to reflect when the WHCTF pulls their data. As of Saturday, Feb. 27, we are at 128 weekly confirmed cases per 100K (down from 134.5 last week), or if you add antigen positive cases, we’re at 192.5 weekly total cases per 100K (down from 197 last week). Starting with the WHCTF report dated Dec. 13, antigen tests are included in their case rates, and I have updated the graph on my home page to reflect those numbers, now that GA DPH started providing historical antigen case data.

In the graph below, I show confirmed (PCR) cases and testing. The percent of positive tests is shown in a second graph below. (Keep in mind testing was under reported by an estimated 50K tests the week ending Jan. 9.)

If you want to see a table of the raw numbers with additional explanation of the numbers, I created a separate page for the chart of Weekly Testing and Cases.

Reported Cases vs. Actual

In the graph below, the blue bars represent the number of reported cases in a given week, and the orange squares represent the number of COVID cases where symptom onset or lab test occurred that week. All of the actual case total (in orange) are subject to change some over time. However, the majority of changes will occur in the final week of the graph, which is likely less than half of the final amount it will be, and the previous weeks should see smaller changes. In the fall/winter wave, test results were processed and reported fairly quickly, so the peak of cases by onset date and report date were both the week ending January 9th.

In mid-December, we exceeded our weekly case peak from the summer based on report date as well as onset date, and cases continued to climb for several weeks. Remember that the final week of this graph is still quite incomplete. Free COVID tests are readily available all over the state for anyone who wants one.

Reported Deaths vs. Actual

Georgia records deaths by actual date of death on their graph on the DPH web site, and I track this, so that I can report on how many of the reported deaths each week are recent deaths vs. older deaths. Deaths reported in a given week often did not occur in the past week, or even the previous week.

The following table shows when the deaths reported over the past week occurred. There was a net change of 637 deaths since last Saturday.

MonthTotal DeathsChange
March236
April1190
May953
June635+1
July1466+1
August2016+1
September1123
October799
November921+3
December1718+25
January3147+150
February863+257

You can see on the graph below that there have been a few weeks where a lot of older deaths are reported. This happened the week ending June 13, the week ending Nov. 14, and and somewhat again the week ending Jan. 9, as well as the past three weeks (the weeks ending Feb. 13, Feb. 20, and Feb. 27). We should death reports drop significantly in the coming weeks now that they have hopefully caught up again on older deaths.

You can see in the second graph below that initial death reporting during December was very low, presumably due to the holidays, but death reporting has caught up in January, and early reporting has been climbing. Deaths seem to have peaked the week ending January 16th. Keep in mind that deaths are always highest in January and February, due to the normal pattern of respiratory viruses, so it’s not surprising to see an increase at this time, especially given the high case rate we saw throughout December. However, deaths seem to be dropping quickly so far in February.

There is often lower death reporting on Sundays and Mondays, and fluctuations from day to day, with higher days often on Tuesdays, so it remains important not to focus on the number of reported deaths in a single day. It does not reflect the number of actual deaths that actually occurred that day. Read more about how deaths are reported in Georgia.

In the graph below, the blue lines represent the number of reported deaths in a given week, and the orange squares represent the number of known COVID deaths that actually occurred that week. All of the actual death numbers (orange squares) are subject to change some over time. However, the largest changes typically occur within the two final weeks.

Remember that deaths for recent weeks, especially this past week (which is still preliminary) will increase further over the next few weeks. Final death totals for the week are typically around 3-4 times the total at the end of the first week during normal weeks. For holiday weeks, expect the multiplier to be even more. (The week of Thanksgiving originally only reported 30 deaths that occurred that week, but it is now over 200.)

Hospitalization Data

I have links on my Hospitalizations page to external dashboards showing more detailed hospitalization data by region. Current hospitalizations for COVID positive patients fell significantly for the sixth week in a row. I removed the other hospital data from this report, as it appears to be significantly affected by reporting delays and does not seem to reflect the current situation based on other sources of hospital data.

Links to Graphs