Top news, reports and insights for today:
- Curated headline summaries for Saturday:
- CDC does the right thing, reverses course on testing after widespread criticism of its tweaking a month ago. Those who have been exposed to an infected person (again) urged to get tested (NBC News)
- U.S. Postal Service planned to deliver face masks directly to households. The Trump administration blocked it (Complex)
- Poorly protected postal workers are catching COVID-19. More than 50,000 workers have taken time off for virus-related reasons, slowing mail delivery and threatening the integrity of the election (ProPublica)
- Two new studies show that coronavirus can spread during airplane travel. The extent of in-flight transmission is not yet known. One of the studies, published in the CDC journal Emerging Infectious Diseases, used genetic analysis to confirm in-flight transmission involving 4 people (CNN Health)
- New study shows rates of COVID-19 deaths are higher in counties with increased level of hazardous air pollutants, even controlling on socioeconomic status, and average ozone and air-borne particles (Environmental Research Letters)
- Daily U.S. COVID-19 cases rebound last week. Numbers rising in 45 of 51 states, possibly foreshadowing arrival of cold/flu season
U.S. daily cases are ping-ponging heading into the last week of September making it hard to know where the U.S. epidemic is headed (See Figure B). After dipping to just 22,000 on September 8, daily numbers have been resurgent, now above 40,000 for three days. I am trying hard not to over-interpret the surge in cases over the past week. The data have been growing less consistent in recent months. However, I see one feature of the data presently that has me pretty nervous. Figure C shows the 7-day growth factors in cases by state. That, as usual, is the ratio of total new cases in the last 7 days compared to the prior week. Values higher than 1 mean cases are growing. Last time I showed you this figure on September 8 (included as Figure D for comparison), new cases were on the decline in all U.S. regions and in 33 of 51 states (plus DC). However, in Figure C, the numbers are >1.0 in 45 of 51 places. That is a worrisome turnaround. New cases are rising in all regions, but especially in the West where we see some impressive jumps last week. Arizona, which had been on the decline recently, leads the way with more than 5,400 new cases last week, which more than doubled from the 2,600 the week before. California now has over 774,000 total cases, which is more than 10% of the U.S. total and puts that state 5# in the world ahead of Columbia (765,000) and behind Russia (1.1 million). Thankfully, new cases grew just 3% last week, but that still was 23,800 new infections. New weekly cases grew by more than 50% in Colorado (+80%), Montana (+63%), Utah (+88%), Wyoming (+62%) and Wisconsin (+64%), the only Midwest state growing at that pace.
The bottom line: the U.S. outbreak remains inscrutable, cases are on the upswing broadly in many states. I worry it is possible we are starting to see the leading edge of cold and flu season as we brace for overlapping outbreaks of respiratory illness.
- Deaths holding steady, approach 200,000. Big picture, there is good news and bad depending on your perspective
Thursday, Friday and Saturday saw deaths fall in the U.S. below 1,000 a day. However, the number of American lives lost continues to mount as we near the 200,000 threshold. If I had told you in March that more than 200,000 American’s will die before October 1, would you have believed it? Actually, I did but that is another story. This week I found myself stepping back from the day-to-day numbers for a look at the epidemic curve of COVID-19 mortality since the beginning. I prepared two additional figures for this purpose to illustrate a key issue that arises in looking at epidemiologic data. First, let’s look at Figure F, showing a plot of cumulative deaths in the U.S. since the very first recorded death on February 29. [Parenthetically, it is most certainly true that COVID-19 deaths occurred before the first recognized case in Washington State on that date, but that is not our focus here.] The figure shows the overall shape of the mortality profile over time on the log scale. Notice how the vertical axis goes up non-linearly. In a typical epidemic, the cumulative event count curves up at first, shifts to a straight diagonal incline during a period of exponential growth, then slows down to a gentle rate of rise before flattening as the rate of new events slows and eventually stops. On the log scale, these shifts look more pronounced. It is just a scaling adjustment, the underlying data are not affected. I have superimposed on the figure some markers for the doubling times, when cumulative numbers increased by a factor of two. The doubling times are a useful tool for gauging the speed of the epidemic. In Figure F, it’s very hard to see first phase when deaths are ramping up. However, the period of exponential growth in deaths is clearly visible as a straight diagonal line (on the log scale) from early March (25 deaths) to early April (8,300 deaths), when the doubling time was less than a week. The slow down phase is evident from then to May 12 when doubling times fell first to 10 days and then 22 days; meanwhile we hit the 75,000 death marker. The next doubling took 91 days (August 11, 152,000 deaths). The good news is that exponential growth in deaths was clearly over and past us. The bad news is that the outbreak didn’t die out (so to speak). The curve never fully flattens. Importantly, on the log scale, it is very hard to see the details of what actually happened over the summer. It looks like things were better but not ideal because the curve keeps rising. Figure F doesn’t help us explain the summer hump clearly visible in Figure E. For that we need to change perspectives.
Shift now to the bottom graph (Figure G), which shows cumulative deaths on the linear scale. It’s the same data, but expressed in a different metric – kind of like looking at the same external world through lenses of a different color. Our attention and our eyes are drawn to different features based on the shape of the data in this perspective. Instead of doubling times, I show you the interval (in days) between each successive 25,000 deaths. The start of the epidemic is in sharper relief here, as we can more clearly see the gradual ramp up. Even though doubling rates were rapid due to exponential growth, it actually took us 47 days to accumulate the first 25,000 deaths. In the next 50 days (from April 14 to June 3) we added 75,000 more. The interval from 100 to 125K was slower at 41 days. But since mid-July, the interval dropped to 25 and then 26 days. That is where the summer hump happened.
What does it mean: The two ways of looking at deaths are neither right nor wrong, but they are different. On the log scale, we can take some comfort in the fact that we are clearly no longer in the explosive exponential growth in deaths we saw in March and April. On the other hand, the epidemic curve never flattened this summer. And the log scale doesn’t do a good job of illuminating the backsliding we actually did over the summer, when deaths began to accelerate again. All of this of course is a prelude for the immanent arrival of cold weather and the unfolding of additional waves of infectious respiratory disease outbreaks that will no doubt complicate the picture moving forward.