Why two different numbers for hospitalizations?

A few weeks ago, Governor Kemp celebrated a new milestone of reaching less than 1000 people currently hospitalized with COVID in the state, many have raised questions about how this number is so different than the Total Hospitalized number that appears on the DPH web site. They are two completely different numbers that come from different sources, so I thought it would be helpful to explain what I know about these metrics.

Currently Hospitalized

This tracks the number of people currently in the hospital with an active COVID infection, as reported by hospitals to Georgia DPH. This number is shared by GEMA on Twitter and Facebook, as well as in the Daily Situation Report on the GEMA web site. The number is typically updated only once per day.

This number includes anyone in the hospital who is positive for COVID, whether you were tested at the hospital or not, and whether you are hospitalized for COVID or another reason (labor & delivery, broken bone, etc.)

Hospital Overall Capacity Numbers

Hospitals report exactly how many ER beds, ICU beds, and the general hospital beds are occupied each day by all patients (not just COVID-related patients). These numbers fluctuate, and are reported in the GEMA Daily Situation Report by region.

They also calculate and report state-wide occupancy totals and percentages of these beds daily. ERs remains around 1/3 full, while the ICU and general beds remain ~75% full. From what I have been able to gather, this is a typical percent of occupied beds at a hospital, and gives us plenty of room to handle incoming COVID patients as needed. While a couple of smaller regions sometimes get very low on ICU beds, there is excess capacity in other regions where patients could be transferred if needed.

Complete data on current hospitalizations in Georgia is updated daily on my Current Hospitalizations page.

Hospitalized Total on DPH site

According to the DPH Status Report definitions PDF:

“Hospitalizations: Defines the count of confirmed cases in which the patient was hospitalized at the time of reporting to DPH. Because of how this number is reported to DPH, it may be underreported. This number does not represent the number of confirmed COVID-19 cases currently hospitalized.”

This number is cumulative, so it only ever goes up. Many people track this number and report on the increase each day. This number has often fluctuated wildly, and based on how it is collected and the lag in reporting, I am not certain how valuable it is.

As the definition above states, it is based on the patient’s hospitalization status “at the time of reporting to DPH”. Another epidemiologist clarified to me that it means the person was hospitalized at the time their test was administered. That is not the same as requiring hospitalization for COVID.

  • It DOES NOT INCLUDE anyone who was tested at an outside lab, who then required hospitalization for COVID.
  • But it DOES INCLUDE anyone who tested positive for COVID while in a hospital for a reason other than COVID. Now that elective surgeries have resumed, patients are being tested for COVID by the hospital, and so asymptomatic or mildly symptomatic patients would be included as a new COVID hospitalization, even if they did not require hospitalization for COVID symptoms.

Finally, because this count is based on confirmed case reports sent in by labs, and not on patient counts submitted by hospitals, there is a reporting lag. There are dips every weekend when less lab reports come in (it does not mean less patients are admitted to hospitals on the weekend), and there are random spikes every couple of weeks, presumably when the DPH gets a big batch of test results or makes an extra push to enter cases on a particular day.

UPDATE AS OF JUNE 25: It seems like they may have figured out the reporting lag and these trends may become more usable.

As an example, in a 7 day span in May, we had one day where the total hospitalizations number increased by 12 (5/10) and another where it increased by 249 (5/17). There is absolutely no way that one day had over 20 times more admits than the other. That’s another reason why I stress that trends cannot be seen from one data point. A lot of people saw that 249 number and reported that hospitalizations had spiked in Georgia, when in reality, it was probably a bunch of test results that were just entered late.

UPDATE AS OF JUNE 25: I’ve also been tracking the hospitalization rate from this number (the 7-day average of new hospitalizations divided by the 7-day average of new cases), and I have added a graph for that below. Despite the spikes from prior reporting lags, you can see the overall downward trend, indicating that less cases are being identified in hospitals than before. I’m not sure that means less overall cases are being hospitalized due to the way this data is collected however.