Tag Archives: Resurgence

The Slope of Concern

There is something apparent in the raw case data that is likely hard to see without some training in epidemiology and statistics. It’s very important and is worth understanding as a means to grasp what is changing and how it aligns perfectly with what I have been forecasting. I have previously explained why there is a 3-6 week delay in outcomes from changes in social controls.

The columns represent new cases of COVID-19 and are plotted by day on the x-axis and on the left y-axis. The green section represents the time that I had previously indicated would see benefits from the various lockdown orders. The yellow section represents when there would be gradual increases in cases as these restrictions are removed.

The black line represents deaths and is measured on the y-axis on the right. Deaths lag about one week behind cases.

There is a blue and red arrow along the bottom. This is just a best guess the cooling period of cases from social control and the reheating as those restrictions are lifted. Unfortunately, I realized a better way to do this after I had created each frame so it’s not as smooth as a gradient I would like in most frames but the general idea should be clear. I intentionally picked seven days into the yellow section to start looking at the trend data so there would be a relatively decent number of points from which to work.

There is a purple trend line for the downward trend seen in the green section of the graph. The red trend line is really the most important piece. It will continue to flatten and will start to slope upward, and will become steep in what will seem overnight once we hit the return of exponential growth. That’s when the trouble will be just starting.

Most important, within about a week it will be obvious throughout the US that there is trouble ahead. My warnings of a lack of ICU beds have already been reported in Montgomery, Alabama and also now in Arizona. There are likely already other areas around the country starting to come up against this problem.

Another way to look at this is a graph of the slopes of the red trend line. What is interesting, the trend line for this data is already flat. That line will start slowing up as well and will be a pretty good indicator for future trends. Currently, I have it set to linear, but at some point it may fit the data better as an exponential curve. I expect hospitals in much of the country to be overwhelmed worse than what was seen in NYC sometime in July.


US Trend Forecasting

I’ve described that it takes at least three weeks for changes in policy to begin to show up in the data related to COVID-19. Using that rationale, I’ve also indicated that the impacts of stay at home orders would begin having impact between April 12th and May 3rd. I have also projected that relaxing these restriction would start to show up in the data for most states between May 15th and the end of June. I did this so people could not accuse me of data mining retrospectively so I could have a prospective look at my hypothesis. We have reached that date for these states:

 North Dakota
 Rhode Island
 South Carolina
 West Virginia

A graph of the US data since March 15th can provide some insights. Cases are the bar graphs and use the left axis for the scale, deaths are along the red line and uses the right axis for scale. Splitting cases and deaths onto different vertical scales makes it easier to see them in comparison to each other.

First, I think it’s worth noting that deaths clearly follow the case trend, but with a delay of about a week.

I’ve broken the data into three sections each with a representative trend line for that period, the first based on the emergence of the pandemic and exponential growth.

The second section projects start of impact by stay at home orders in most states, which is represented by the green area over the dates. I’ve also included the following 11 days as part of the benefit of that period, although it likely extends even longer than that.

The orange section is the start of when I expected to see impacts from the relaxing of stay at home orders. Each state did this a little bit differently, but it had been my best effort to project forward from whatever language was used by a given state.

The dark straight trend lines for the cases in each section are what I’m most focused on. It’s clear we had rapid growth through the beginning of April and that we succeeded at flattening the curve as indicated by the second trend line.

While a subtle change, the third trend line is concerning. For those with a mathematical interest, the slope of the benefit of the stay at home section is -216.56. The slope of the line for the resurgence section is -128 currently. The important thing is that it has changed that much with only about 1/2 of the states starting to see the impact of those changes, and 11 of them in the last 8 days, so this is just the start of the change that is coming.

I accept that some of the increased case count is due to more testing. However, I am sticking to my initial estimate of the middle of June to start seeing a rapid climb in cases again followed by quickly overwhelming our ICUs and eventually our hospital beds and morgues.


State Check In

It might be worth looking at data from states that have ended stay at home orders to see if it has any impact on new case trends. Only three states have a sufficient number of cases and population at this point for review. My commentary for each is in red text. The rest of narrative about each state is from CNN and was updated on May 20th. The three week post opening point is the divider between the blue and gray portions of the graph. Trend lines are the dotted lines within each.


Colorado also has some interesting results from PCR versus serology testing. Clearly the testing impacts some of the numbers of cases. More interesting is that such a high percentage of serology tests are negative. I wonder if people think that they had been infected really weren’t. I’ve heard a lot of people claiming that they “probably” had this back in January or late last year. I’ve told them it’s just about impossible if they were in the US during that time frame.

The state’s “safer-at-home” order took effect April 27 and is in effect until May 27.

Retail businesses can reopen with curbside delivery and elective medical procedures can resume. Businesses such as personal training and dog grooming can reopen with social distancing.

Retail businesses began to reopen May 1, while people were permitted to return to non-essential office work May 4.

On May 11, Gov. Jared Polis said state park campsites would be available for rental beginning May 12.

A decision on restaurant reopenings will happen May 25, the governor said.


Georgia has a 14 day lag in collecting data. However, in their trend forecast, they are expecting an even steeper rise in the second graph.

Gov. Brian Kemp started to ease restrictions April 24.

Gyms, fitness centers, bowling alleys, body art studios, barbers, hair and nail salons, estheticians and massage therapists were able to reopen April 24, with certain rules. Theaters and restaurants were allowed to reopen April 27, also with caveats.

The caveats include social distancing and screening employees for illness.

Bars, nightclubs and music venues will remain closed, for now.

A shelter-in-place order for “medically fragile and elderly Georgians” is in place through June 12.

The shelter-in-place order for other Georgians ended April 30.


Oklahoma has something very interesting. When looking at the data for the whole state, it looks like easing up on the stay at home orders. However, when limiting it to the data set of the metropolitan counties in the Oklahoma City and Tulsa areas, a different pattern emerges. There are 14 of the 77 counties in the state in this view. This is pretty good anecdotal evidence that population density makes a big difference.

Also, I realize that the graphs are a bit different and the midpoint is two days later in the bottom one. It’s a function of pulling state data in batches to match the change date and I didn’t think of this view until I had changed some of the formulas in my data table. I think it’s safe to assume that the trend lines are still representative of the data.

Gov. Kevin Stitt allowed some businesses to reopen beginning April 24.

Among them are personal care businesses, restaurants, dining rooms, movie theaters, sporting venues and gyms if they maintain “strict social distancing and sanitation protocols.”

Bars, however, will still be closed.


Science is funny about conclusions. I think from these examples, there is a common trend, however, like many analyses, further data is needed. That will come when more states have 4-5 weeks of data after ending stay at home orders.