The math: why COVID-19 is declining in hotspots and flat nationwide, but deaths are still rising

The past ten days have brought a mix of good and bad news for the United States and COVID-19. On the good side, the moving average of new COVID-19 cases in the USA has flattened after over a month of rapid growth. New cases have even fallen by about 3.4% in recent days. The bad news: deaths are still rising to a moving average of over 1000 a day.

This extremely welcome news on cases and continuing bad news on deaths has come with a lot of discussion of what it means, including some mistaken notions and even conspiracy theories. It’s important to clear up what’s actually happening, where it’s happening (since different regions of the United States have different dynamics), and what to expect next.

Here’s what you need to know:

1. The case-count-flattening is almost certainly real.
2. Flattening numbers of new cases are driven mainly by turnarounds in the hottest hot spots.
3. Other emerging clusters so far aren’t as lucky.
4. Deaths are likely to keep rising for several weeks in even the most optimistic scenario.

As we’ve been reporting for a while, COVID-19 outbreak trends are mainly the result of behavior with a significant time delay. Once a big hotspot emerges, people tend to react by fixing their behavior to be more careful. However, as in the first phase of the outbreak in the United States, places that haven’t yet seen a level 4 outbreak have a hard time changing behavior to curb the spread. National control still eludes us.

In general, it remains extremely important to base one’s evaluation of the situation on local data. Sentiment remains very detached from the data, and national data obscure local dynamics.

The flattening of new cases is very likely real, and not a coverup

The last few days have seen conspiracy theories about the flattening of cases. People have very good reason to distrust many governments, including this one, on the topic of COVID-19 statistics and messaging. This crisis hasn’t been managed well.

But the theories say that the Trump administration is concealing a flood of new cases and that cases have actually risen during this time. Folks spreading this conspiracy theory have seized on the Trump administration’s attempt to take COVID-19 data collation activity away from the CDC and place it elsewhere in the Department of Health and Human Services (HHS), and the fact that this transition happened right around the time cases started to flatten.

Of course, if the administration really were concealing new cases, it would be a serious crime. However, there’s no clear indication that case statistics have been fabricated.

For one, although the fight over CDC data access is ongoing, the data from which the CDC was cut off was hospitalization data, not case data, and it is the case data that has flattened. Hospitalizations are actually still rising. So, while the taking of data away from the CDC is still a four-alarm outrage, and it would be difficult to rule out a nefarious motive from the government at this point, there’s no current evidence of doctoring and the data in question has not been redirected (that we know of).

For another, if the new case counts were being doctored, we’d expect to see test volumes suspiciously lower than usual. But the current flattening doesn’t seem to be associated with a reduction in test volume or an increase in positivity, either nationwide or in the states seeing the biggest numeric declines.

Finally, it’s tempting to think that we should be seeing rising cases if death counts are rising as much as they are. But the ongoing rise in deaths doesn’t actually contradict the flattening of overall cases. We know at this point in the pandemic that death numbers lag behind case numbers, just as they have in the past.

It’s overwhelmingly likely the flattening is real, and not a conspiracy.

Declines in the biggest hotspots

The flattening is real, but it’s also not uniform. The states — and particularly the counties — with the largest caseloads have seen a decrease in new cases because people there have been taking more precautions. But many — even most — areas of the United States are still showing increases.

The four states which have been showing the largest case numbers recently, Florida, Arizona, California, and Texas, have all recently shown significant drops in new caseloads, and they’ve all shown R drop significantly below 1. Right now, Texas and Arizona have the lowest estimated R of any states in the country. However, dozens of other states are showing rising caseloads and R significantly above 1. These effects are, right now, largely canceling each other out.

On a nationwide basis, The Prepared pulled CSSE data for all US counties up to July 23. We looked in particular at the top 280 most affected counties, which account for about 80% of recent cases. Among these counties, there’s a stark contrast.

The 21 counties with the top declines in case numbers in absolute terms account for 200% (two hundred percent) of the decline in nationwide numbers between the week ending July 23 and the prior week. That is, the top 21 counties went down by twice as much as the aggregate of all 280 counties. The other 259 counties, taken as a group, had a net increase in new cases during this time. These 21 counties with the biggest declines are located in:

• California (Los Angeles and Orange)
• Arizona (Maricopa)
• Texas (Bexar, Harris, Travis, Tarrant, Montgomery, and Nueces)
• Florida (Orange, Lee, Broward, Hillsborough, Duval, Pinellas, and Palm Beach)
• North Carolina (Mecklenburg)
• Tennessee (Shelby and Davidson)
• South Carolina (Charleston)
• Arkansas (Hot Spring).

Of the 18 counties with more than 400 new cases a day, 13 are showing declines in cases.

In the rest of the counties on the top list, taken as a whole, cases are still rising. In over half of them, as well as over thirty states, R is greater than 1.

This includes lots of major metro areas: Fresno, Las Vegas, Dallas, Sacramento, Chicago, Baltimore, San Diego, St. Louis, Philadelphia, Salt Lake City, Seattle, Omaha, Washington DC, Atlanta, Portland, and San Juan, where cases are rising fast.

Dr. Fauci identified Tennessee, Indiana, Ohio, and Kentucky as being at particular risk.  And we don’t always have a great picture of how bad things are; 13 states show test positivity over 10% and a handful, like Mississippi and Alabama and Florida, are still at the levels the USA as a whole attained in March.

Total cases could rise again

There’s absolutely no guarantee that this plateau will be sustained. While top hotspots have seen declines, those declines won’t continue unless residents continue to express caution in their actions. In addition, a little math shows that even if the top hotspots sustain their downward trend, the national trend will revert to case growth within weeks if nothing changes in the emerging hotspots mentioned above.

The United States is sitting on a mountain of COVID-19 cases, and our journey from about 65,000 new COVID-19 cases per day down to more sustainable levels means pushing R below 1 in the vast bulk of states and keeping it there for a long time. If we don’t do this (and we haven’t yet), then growth will resume, and from a higher baseline.

Deaths will keep rising for a while — even if new cases start falling

As we’ve reported before, deaths lag behind new cases. The beginning of the current deaths increase was deferred by over three weeks relative to the increase in cases (about July 7, vs June 15).

So it’s likely that deaths will similarly continue to increase for several weeks after cases plateau or decline. This is worrisome since the rate of new deaths has already doubled from its early July trough to over 1000 new COVID-19 deaths a day.

The bottom line: Keep your eye on the ball

Figuring out exactly what’s going on isn’t very straightforward. If you want to know what’s going on around you, here are some steps you can take:

1. Look at data, not sentiment. My work as a data scientist has taught me that people can’t always be trusted. People are fully capable, individually and in groups, of being blithe when the data looks bad or tearing their hair out when things are improving. But the data doesn’t lie (so far).
2. Look at local data. Around the country, the situation is pretty mixed. Some areas are in great shape, and others are exploding — all at the same time.
3. Look at both the magnitude of cases and the direction of change. Phoenix is trending down and St. Louis is exploding, but Phoenix still has a higher current caseload.  Los Angeles is bad but flat, and New York is low and flat.  What’s happening in your area is more than just one number; it’s at least two numbers. Case counts might decrease and still be enormous, or they could rise and still be quite small.
4. Remember that deaths lag behind cases.