Errors in the COVID-19 Data?

The data remains as puzzling as ever, and in our view no one has yet been able to interpret them with any accuracy. Today we are going to review emerging aspects of data that raise questions.

We start with Figure 1. our own tracking graph that seeks to find the correlation – if any – between testing, reported infections, and mortality.

Last week we noted that the 7-day moving average of the mortality rate showed a distinct upward trend, mirroring the sudden upward turn in the reported infection rate from 26 days earlier. We noted that this could indicate we had a renewal of the crisis looming, given the huge rise in the infection rate since that time.

However, the growth rate in the mortality data declined this week, which indicated at least a degree of detachment from the reported infection rate from 26 days prior.

Readers of this column know that we hold the reported infection rate as unreliable, reflecting only a portion of those infected. Based on the mortality rate, and the known IFR (Infection Fatality Rate) of the disease, we have high confidence that the total infections in the US exceed 14 million, compared to a reported 3.8 million. The number of infections is a complex mix of increase in testing, and an apparent increase in infections.

However, when it comes to mortality in general across the US, we have a paradox in the data, as yet unreported in the media.

The graph in the figure above is maintained by the CDC, and reports excess deaths from COVID-19 and other reported sources. It has been a source of data that we have used in these columns, and has – until recently – reflected the mortality rates being reported on COVID-19.

NOTE: the CDC warns that the data in most recent weeks may not be complete, and that while it makes statistical adjustments to correct for this, the data for those weeks may not be reliable. We have found over the last several months that we may rely upon data up to two weeks prior to the reporting date, and we have marked that point – June 20th, 2020 – on this and the following graphs with red arrows.

It can be seen, however, that the “Excess Mortality” – that is the number of deaths higher than the historic average (adjusted for population) – has fallen dramatically from the peak in late April. Mortality is now (or was, on June 20th) at the average expectation level for a “normal” year.

The story is continued in the following CDC graphs.

These data indicate that the causes of death closely associated with COVID-19 have fallen off dramatically in recent weeks to the point that the deaths are at expected levels. The only the exception is deaths due to Alzheimer’s disease, which shows a mild rise. That rise has not impacted the total weekly count of deaths in the 85-year and older group, which are shown as remaining at the normal expected rates in recent weeks.

These declines in aggregate, absolute and relative terms, indicate a far more dramatic drop in mortality rate than those reported daily and as attributed to COVID-19. Furthermore, the data also indicated that the deaths reported prior to the peak in late April included a significant number of deaths that were so-called “pulled-forward” – that is patients with comorbidities that succumbed to the disease, but who may have died in any event within a few months. This is implied by the decline to below average mortalities in recent weeks.

What about the western and southern states, which are reporting a very significant rise in infections and deaths. Figure 4 is our weekly tracking graph of infections versus deaths in California, Texas and Florida. We aggregated those states as convenient proxies for the southern and western states, their populations being so large that there numbers are decisive for any statistical analysis.

In this graph there is a very strong correlation between infection rate – pulled forward by 26 days – and the mortality rate, and the trend is indeed alarming. It indicates more than a doubling within a week, with no apparent end in sight. However, when we investigated the CDC Excess Death data for these three states, the pattern for all three states remain similar as that of the U.S. shown in Figures 2-4, which is to say that in the weeks leading up to end-April the mortality rate spiked to much higher than normal levels in in the higher age groups and for the causes of deaths associated with COVID-19, and then subsided to a current level of below normal by June 20th. The average number of deaths in all three states shown in the CDC graphs at June 20 is at normal levels.

Conclusion

The actual mortality rate from COVID-19 remains elusive, but not as much as the infection rate. Excess Death rates were cited early in the pandemic as an indicator of under-reporting. It now seems that the opposite may be true.