Why fluctuations in reported numbers may not indicate a true increase in coronavirus infections, hospitalizations or deaths.

In trying to understand the thousands of Twitter voices bemoaning the “spike” in infections, I first thought most of this was simply due to more testing. This is a significant factor, as more tests mean more positive tests regardless of the infection rate. However, it turns out there have been changes in several parameters of Coronavirus infection reporting in different states.

I received a great deal of help on this from several readers of this blog. I want to particularly acknowledge JK in New York (but a UM grad!) who provided the observations for much of the information I will be presenting.

I also decided that rather than focusing on states (Texas again!) that have controversy over potential increases in COVID-19 infections, I would focus on Michigan where the decreases in both hospitalizations and deaths are universally accepted (see graph). 

It turns out that Michigan has changed the way it reports cases and deaths recently. This has significant impact and transmission figures for the virus and overall incidence and death numbers.

As reported in Bridge Magazine (started by friend Phil Power!) as of Friday June 5^th Michigan changed the way it calculates COVID-19 deaths, boosting the total by 240 deaths to more than 5,800. Michigan now includes a category called “probably deaths.” The inclusion of “probable” COVID-related deaths reflects cases in which the virus is strongly considered as the cause of death, but victims didn’t test positive for the virus. 

In addition to deaths, the state said there were nearly 5,000 more “probable” COVID-19 cases as well, an 8.5 percent increase over the confirmed cases of 58,241. Most of the suspected cases were in Detroit, where the most infections and deaths were seen.  

To get listed as a confirmed case, the dead person must have tested positive for the virus and COVID-19 must be listed as a cause of death. Probable cases are where the victim did not have a positive test but was presumed infected based on symptoms and another positive case, like a family member.

Unlike some other states Michigan is keeping these categories separate.

As I pointed out the Rt values, which relate to the number of infections a single individual can cause, have increase in several states like Michigan where hospitalizations and deaths continue to fall. The bump up in Rt last week has now gone back to less than one, so it likely was an anomaly of the change in reporting. (N.B., Texas has not changed in the last month despite some alterations in reporting).

States classifications of COVID-19 deaths have been criticized for more liberal or strict interpretations of COVID-19 deaths. Criteria for reporting deaths has changed in some states and cities. New York City in mid-April added more than 3,700 deaths of people who were presumed to have covid-19 but were never tested. Even now, states continue to modify their procedures, so the numbers may continue to fluctuate. Some states are still counting antibody positive individuals as new infections even though they are by definition past infections, which adds to the confusion.

Bottom line; no “spike” in hospitalizations and deaths as of yet in most states. Importantly, understanding fluctuations in COVID numbers is more complicated than it might seem, especially from Twitter!