Just one more week with an alarming rise in the COVID-19 infection rate, but again, no sympathetic rise in the mortality rate.
Erroneous reporting (again)
Yesterday, 7/4/2020, saw the USA daily mortality rate reported from COVID-19 decline to a level not seen since March 23rd, and come within a hairsbreadth of penetrating the benchmark of 250 deaths for the day. For the week, total deaths were 3,648, a weekly number lower than any since week ending March 28th, and an increase in the rate of decline week over week from the prior week.
The dramatic rise in infections appear to have plateaued at about 50,000 per day (although Dr. Fauci is talking about a potential rise to 100,000 per day), and this just as the rate of testing is reaching upward toward the 1 million a day mark.
When we began this series of reports on the pandemic, we pointed out that the number of infections were undoubtedly being under-reported by a very significant margin, as the IFR (Infection Fatality Rate) was then (mid-March) being reported as well below 3%, and possibly as low as 1%. We now know that this rate is significantly less than 1% for the population.
It stood to reason in the early days that total infections were in fact many times the number of deaths, rather than numbers being reported. For example, on March 15, 73 new cases were reported, bringing the cumulative number of cases to that date to 310, compared to the 121 deaths to date. We now know with fair certainty that the actual number of infections at that point were more than 12,000.
Just as the statistics pointed to substantial under-reporting of the number of infections in those early days, they almost certainly point to a likely over-statement of concern today. We were bound to report a very significant increase in the number of infections once our testing reached current levels.
The urgent question is how much of the increase is a result of wider testing, and how much is real growth in number of infections; and if it is, will we once again face a rapid growth in the mortality rate?
The Key States
To answer this question, we are tracking California, Texas and Florida. These are the key states experiencing very substantial rises in infections, and they represent a very large fraction of the US population.
The following graph summarizes the historic record for these three states since mid-May. As all three states suffered modest rates of infection, the graph shows a small decline in the 7 day moving average of the mortality rate to the end of May, and an also modest rise in number of deaths back to previous highs by the end of June – again still at modest levels – while the apparent infection rates rise precipitously.
In the forthcoming days, the combined death toll looks set to rise in sympathy with the soaring infection rate, but at a lower proportionate rate if the rise of the last two weeks is a good indicator.
However, the graph is based upon the number of days established in the literature – twelve – between infection and death. What if this is not correct, and the average time between infection and death is longer – say three, or even four weeks? This would mean that the proportionate rise is much larger than indicated in the graph, which is calibrated to the twelve-day delay. We raised this concern last week, but the evidence this week appears to show that the twelve-day lag may be correct.
Hospitalizations
Last week we began tracking the health systems’ ability to deal with growing infection rates, particularly in the states we are following, California, Texas and Florida.
Using the IHME (Institute for Health Metrics and Evaluation, University of Washington) numbers, we summarized the hospitalization statistics for California, Texas and Florida. Our review of the metrics this week show that the IHME have revised downward their estimates of the facilities that may be necessary based on their projections. In all cases, from an analysis of the total resources available, it appears that the states have adequate care for the IHME projected number of patients (percentage utilization required is shown in the table.)
Testing
Last week we featured the Johns Hopkins test tracker. This week we use the tracker to illustrate the continuing improvement in testing, by comparing this week (the image on the left) with last week (the image on the right). Each diagram is easy to understand: states in red are testing far too little, while those in yellow are at optimal level, with varying degrees of orange in between. The colors denote the fraction of those tested that are NOT positive. In other words, it isn’t helpful to control the disease if you are only testing those that are already ill. By testing a wider sample, you are able to track the extent of the spread, and so control it.
California achieved a significant improvement (as did several other states, including North Carolina), while Texas and Florida remain pink and red (poor and bad) respectively. Improvement in testing in these states would seem to be a way for them to solve potentially serious problems, and to continue to open their economies.
So where are we?
By this time next week we will have much greater clarity of mortality trends against the burgeoning infection rates, as it will have been more than 3 weeks since the climb in the infection rates. Our guess now is that there will be a modest pause in the decline of the mortality rate, but the trend will remain downward. Ultimately, it will be dependent on the states, local leaders, and the general public to ensure that the situation remains in hand and that the pandemic remains under control.
do you think the number of times people are being tested should be accounted for in the testing statistics? anecdotally, it seems doctors are routinely advising multiple tests and three people I know tested negative, with symptoms, then positive on a second test. there are also many people subject to routine testing (medical professionals, athletes, etc).
We have little visibility into the testing data, so I agree with you, we need better data. We do get, and I do show, from Johns Hopkins, details on testing done on symptomatic v asymptomatic patients. What we don’t know is how many unique people are tested, as you point out. This would be important, particularly if we could also get statistics on antibody test results. I will do some research to see how much detail is available.
Tom – I just realized that the reply came from you! I hope you are well. Great to hear from you.
Larry,
Excellent work and interesting reading. My big question is how sure can we be that the mortality data is reported properly. Here in Florida, there has been a great deal of doubt about the counts. There has been continued doubt that the virus deaths were not properly recorded or worse yet that the numbers were altered by direction of the Governor to improve results. The data scientist handling the Florida data contends she was fired for refusing to alter counts. . Do you think this happened elsewhere? Can we rely on the mortality rate as accurate if the data is flawed?
Hi Arne – How nice to hear from you! Hope you are keeping well in these sad times. I have a LOT of concern about the figures being reported, and not just in Florida, and not just with political concerns. The standards for diagnosis and cause of death are very, very squishy. I am certain that the arguments were around these issues. At the end of the day the most authoritative source we have is the CDC’s “Expect Death” statistics, which referenced historic death rates for specific states and localities over time to develop expected mortalities, and compares that to actual reports. So I use that to check the numbers that are reported before I use them. Of course the CDC numbers are also subject to adjustment because many localities fail to report on a timely basis, so we do the statistical adjustments as well as we can. So it’s VERY difficult to pin down a timely estimate for a particular state to get a read on mortality trends. I will publish the Florida Excess Deaths numbers this forthcoming week to show you why I feel that the number we have for Florida is about correct. Stay well!
Excellent reporting. What is the biggest issue you’ve seen in the compilation of the data? In your, opinion do you think there’s a good chance of the US falling backward for the next six months? I’m curious that most people reading the data (that is available) begin by saying mortality rates are down, and somehow better, but what about the potential long-term health issues that we are starting to see? What do you surmise? It’s okay to say, I am not sure because I don’t think anyone has a grasp (or there’s an appearance of) a good idea, maybe. Is it still to soon to know? Thanks in advance, I appreciate your insight.
Kent – Great to hear from you, and hope you are keeping well through these trying times. On the data, the problems are immense, and the most serious is that there is not a uniform standard applied to cause of death, and date of reporting of the death. Not only does it vary, but it varies by county. And there are a LOT of counties. So the numbers vary by very significant amounts from time to time and from reporting standards to reporting standard. On the mortality rates, I am very comfortable with the rates I reported from published (but as yet not peer reviewed) material. It varies greatly by age, from sub-.1% in children, to sub-1% in adults under 44 and all the way up to 7.5% in the over 74. These numbers have really remained pretty consistent over time and across continents. What we have had poor insight to is the number of asymptomatic vs symptomatic that has been tested over time. Johns Hopkins is tracking that now, but we can’t compare it to the results we were getting at the height of the deaths in mid-April, so its limited in its usefulness as a data point. It will prove more useful as we get into the last quarter of the year. And yes, we are definitely falling backward…or, should I say, we are not progressing at the rate we should be progressing. So the jury is still out whether this huge growth in detected infections is going to result in a rapid growth in mortality: I think it will result in a rise, but not as great as the rate of infections. Our country is just not geared for a centralized, controlled response to the threat: we are not a centralized state, like Japan or Germany with a compliant, uniform population! We are in for a long, and very trying haul, and must wait for (1) effective treatment (which I think is still at least 3-6 months away), and (2) vaccines, which may be 6-18 months away. I say this based upon the research guys I talk to here in the Triangle, who are quite close to the action. Finally, on the potential of long term medical issues: I am not qualified to say, my expertise is limited to the data. However, the clinicians and epidemiologists I have asked about this seem to all think these are relatively rare occurrences that – given the very large number of cases where recovery has been complete – are more likely related to a limited number of people having particular vulnerabilities to the virus. But I dont think all is yet knowable, and that’s my non-professional opinion!