There is a spike in COVID-19 infections! So we are told in stark headlines on national and local media.
There has been a rising drumbeat of this warning ever since the mortality rate in the major US hotspots, particularly New York, abated.
According to many, we are either experiencing a spike or are about to have one, and it’s all because State Governors have relaxed the lockdown order.
Has there been a spike? And if so, how serious is this spike? And if not, is one coming?
From what we have seen in the data there is isn’t clear evidence of a spike. If anything, there is evidence of a continuing downward trend in the US mortality rate, despite the widening relaxation of the lockdowns.
The big picture
General observations on the Graph
The graph above is based on the weekly graph we publish showing the daily mortality rate in the United States – the solid blue line, together with a seven day trailing average. This trailing average smooths the daily reporting cycle offsetting the lack of reporting on weekends from many counties, which cause low weekend counts, and consequent weekday spikes.
This week’s graph adds a solid orange line which depicts 4% of the daily infections. We choose the arbitrary ratio of 4% because it enables us to align the graphs of the two metrics in a useful way. We understand that the actual mortality rate is significantly less than 4% of those infected, and we will revisit this metric a little later.
Further, the daily infections are shifted forward, reflecting in the graph, the count occurring 12 days before the date on the X-axis. This is done because that is roughly the average elapsed time between the COVID-19 infection being diagnosed and death occurring. (In the non-peer reviewed literature this elapsed time is said to range between 11 and 13 days. – see: Wilson, N., Kvalsvig, A., Barnard, L., & Baker, M. G. (2020). Case-Fatality Risk Estimates for COVID-19 Calculated by Using a Lag Time for Fatality. Emerging Infectious Diseases, 26(6), 1339-1441. https://dx.doi.org/10.3201/eid2606.200320.)
We have chosen 12 days because it is clearly the best fit for the data, establishing this metric for the US experience with COVID-19.
In the graph we have annotated 3 dates:
- April 7: On this date, a little over 5 weeks after the first US death, and two weeks before the historic peak, we can see from the chart that the US reported 2,233 deaths. We can also see that the US reported, 12 days prior, a total of 748×25 infections (remember, our chart only shows 4% of reported infections), that is 18,691 infections, an actual mortality rate of 11.95%. (We know this is far in excess of the mortality rate for COVID-19, even in New York City.) The notation in the blue panel on the chart tells us that we had completed 2.2 million tests by that date, and were testing at the rate of 7,000 tests per million population per day.
- May 6: Now 9 weeks into the pandemic, and about two weeks after the peak, the chart shows that we reported 2,531 deaths. We also reported 1558×25, or 25,459 infections, yielding a mortality rate of 6.5%. The infection rate appears to be closer (but not very close) to reality. We see from the blue panel that by this date we had quadrupled the total number of tests to 8.7 million, while we were testing at the rate of 26,000 tests per million population per day.
- June 13: this brings us up to date. At this point, the moving average for the 4% infection line has intersected the mortality rate line, meaning that we are at a mortality rate of about 4% of reported infections. At this point we have performed 23.8 million tests and are testing at the rate of about 72,000 tests per million population per day.
Conclusions to be made from the graph
There is a correlation between number of tests being conducted and the ratio of infection rate to mortality rate.
Simply put, the more testing we do, the higher the number of reported infections in proportion to deaths. This may be the cause of the “spike”.
It is also clear from the graph that not only is there a secular downward trend in the mortality rate, there is ALSO a gradual downward trend in the infection rate, despite the increased testing.
We are confident that the infection rate is still significantly higher than is being reported, and that we will see the gap between the infection rate and the mortality rate widen substantially in the immediate future .
Simply put, we will have a continuing apparent increase in reported infections (with attendant press and political coverage), with a simultaneous decline in the mortality rate.
But, what about the States
We are told that several states “opened too early” and they are suffering from a spike in infections, and that the decline in the national mortality is hiding the spikes in the states. Let’s examine the thesis by looking at states frequently mentioned in these reports and which, because of their large populations, could have a real impact on the national numbers. Let’s look at the three major states frequently cited in these spike reports: California, Florida and Texas, since together comprise 25% of the country’s population.
California lags many states in relaxing their lockdown, and they have cited “worrying statistics”.
Let’s start by looking at a graph of California’s reported daily new cases, with our standard 7 day trailing average to smooth the reporting anomalies:
The “Daily New Cases” graph above would be very worrying to state officials, indicating as it does a constantly rising number of infections. But, we lay the concerns to rest by looking at the actual mortalities arising from those infections:
We see from the graph of California Daily Deaths a flat, if a slow secular decline in mortality. Daily new cases have reached 3,000 per day, but the mortality rate has fallen from a constant 75 or so deaths per day, down to about 60 deaths per day. The mortality rate of 2% is still, in our opinion, high and so we confidently expect the infection rate to increase without a concomitant increase in mortality.
Another state cited for its unfortunate rise in infections is Texas. Here is the evidence:
And here, as would now expect, is the refutation:
What is interesting in these graphs is that Texas has a mortality rate approaching 1%, compared to California’s rate of 2%. It is possible that Texas may have more effectively detected infected populations than California.
Frequently cited as vulnerable, and likely to suffer from spikes because the state was very late to lockdown, and very early to loosen restrictions. Here are the numbers:
Again, we see rising infections, and indeed an alarming “spike” on January 13. Here is the mortality chart:
The jury is still out on this state. It can be seen that the state enjoyed a decline in mortality from May the 8th up to present time despite the lockdown being lifted on may 3rd followed by a small rise in the infection rate from that time.
However, the infection rate start to rise very significantly on June 4th: whether this increased rate of rise will reverse the state’s good mortality experience to date, only time can tell.
When considering all the reports of “Spikes”, color us sceptical. We do not yet see the convincing evidence. We continue to see a steady decline in the mortality, while we see steady – but too slow – progress in testing.