Today, I’m going to analyze the Covid-19 epidemic in California, using the stats supplied by the California Department of Public Health (CDPH). In July, a series of CDPH errors resulted in the caseload being underreported, and correction of those errors resulted in overreporting for part of August. The system seems to have returned to normal operation now, so that it’s possible to see what’s going on instead of being mislead by the errors — as Gavin Newsom was in July.
The first think I’ll show you is a set of stats that was never subject to the errors, the number of hospitalized patients versus time.
I’m using the CDPH terminology, and I’ll unpack it for you: confirmed hospitalizations are hospitalizations with a positive Covid-19 test, and unconfirmed hospitalizations are hospitalizations with no positive Covid-19 test, but for which Covid-19 is suspected as the primary illness. You can see that there was a large increase in hospitalizations that began in mid-June. The increase reached a peak in late July, and has now fallen to about the levels seen in April and May.
The death statistics were also not affected by the data reporting issues.
The moving average uses a 7-day integration window. The deaths seem to have peaked about a month after the peak in hospitalizations. We are finding that people are spending more time in hospitals than at the beginning of the pandemic, although the CDPH does not report information that would allow the average hospital stay to be calculated.
Now we’ll look at statistics that were affected by the reporting errors. The first is the number of tests performed per day.
The flattening and small dip are the result of underreporting, and the large peak the effect of catching up. Averaging those out, we can see that the number of tests seems to be dropping. This would normally be a bad sign, but the import is lessened by the fact that the positivity rate is not climbing.
The above graph uses a logarithmic scale, with every division vertically being twice the previous one. The WHO recommendation is 5%, or 0.05 on the graph. We are currently approaching that number statewide. The next graph is the number of cases reported daily.
You can see the drop at the end of July. That’s what fooled Newsom, and many other people, into thinking that we were getting on top of things before that actually happened. It does no appear that the number of cases is dropping, but it would be better if that were happening in the face of increased testing rather than the reverse.
We want to avoid exponential growth with an exponent greater than one. If we plot doubling time, periods of rising slope are periods where the growth is slower than exponential. Periods of falling slope are periods where the growth is faster than that.
Ignoring the test reporting glitches, we have been experiencing slower than exponential growth since mid-July. Unfortunately, this has occurred with a large number of cases being reported daily.
Another statistic that is important is the number of deaths divided by the number of cases. The is the case fatality rate (CFR), and can be calculated cumulatively, or as a moving average. In the graph below, I’ve used the cumulative numbers, and included a 10-day delay for the denominator. The CFR is about 2%.
Using the CDCs estimate that 40% of the cases are symptomless, and figuring that symptomless cases have a lower probability of being included in the case statistics (they will still be found by some contact tracing programs), and looking at the positivity rate, it is probable that the infection fatality rate is somewhat below 1%.