After a couple of months of data anomalies, the Texas COVID reports for the last two weeks (ending 9/5) appeared to stabilize.  There were no apparent data dumps of backlogged information and the COVID Tracking Project is now giving Texas an A+ rating for its data.

Testing

Average daily testing fell by about 14% over the last two weeks from about 45,000 tests per day to 39,000.  As fewer people are getting symptoms, it is becoming harder to persuade them to be tested.  I am sure you have heard some of the pleas to public to be tested, including Sylvester Turner’s screed that if you are not going to be tested you are being irresponsible.  The reality is that few people are going to voluntarily go for testing unless they are symptomatic.  If the current downward trends in infections continue, testing will also continue to track down.

The number of daily positive test results averaged a little over 4,400, down 32% from about 6,500 in the previous two weeks.  The positivity rate fell from 14.6% to 11.4%.  For the last week the rate dropped to 10.2%.  I am still a little skeptical about the State’s testing reports.  The Texas Medical Center is only showing about a 5% positivity rate in its testing.  I suspect the State is still having problems eliminating multiple positive test results for the same person.

Fatalities

Reported fatalities also fell by nearly a third over the last two weeks.  If you recall, Texas began reporting fatalities by the date of death in early July.  So, the numbers we see each day are the number of fatalities confirmed by the State for that day, not the number of people that died that day.  Now that we have a couple of months of this data presentation, we are beginning to get a better picture of the timing of the State’s fatalities.  The chart below is the number of fatalities by the actual date of death as shown by the State on various report dates.  This appears to confirm that the fatalities peaked in the third week of July at about 275 per day and has been falling since then.

Exactly how fast it is coming down we don’t know because of the lag time in death certificate filings, but the reported fatalities for the two weeks was down average of about 225 per day to 160. (Reminder – the number of fatalities on the far right side of this chart are understated because of this lag time and will increase over time as is demonstrated by looking at earlier reports.)

The delay in Harris County reporting fatalities continues.  It is currently reporting 1,442 deaths on its Harris County Dashboard while the State is showing 2,300.  I don’t know why the County is even bothering at this point and not just using the State’s data.

Texas fatalities now stand at 46 deaths per 100,000 population.  The State still ranks 20^th overall compared to all the other states and 7^th among the ten largest states.  The State’s rate is still below the national average of 55 but that gap has been closing in recent weeks.  Again, just a reminder that we don’t know whether other states have gone to the actual date of death reporting or not.  So, the comparisons between states may not be apples to apples.

Hospitalizations

All the hospital indicators also continued to improve.  Total hospitalizations and ICU bed usage both fell by 25% (5,274→3,973 and 1,878→1,409, respectively).  However, the decline began to flatten toward the end of the second week.

According to SETRAC, Houston area hospitalizations have fallen by 80% from their July highs and ICU bed usage by two-thirds.

The Rt Estimates

There are a number of groups which are producing a daily Rt estimate.  This is supposedly the estimate of the current reproductive number for the virus, i.e., the number of other people that one infected person will infect.  With all due respect to those making this attempt, is it, in my opinion, complete nonsense and I hope that no policymakers are factoring it into their decision-making process.

These estimates are based on the State’s data and assume that each new positive test result is a new case which occurred on the day it was reported.  Both of those assumptions are wildly inaccurate.  Some the models make adjustments for the obvious problems with this approach, like tests results being delayed.  (See UT Health Science notes on its modeling here.)  Nonetheless, this is a prime example of the old IT adage, “garbage in, garbage out.”  We have almost no visibility into how many people are actually being infected every day, and without that data any estimate of Rt is little more than a WAG.

Conclusion

Fortunately, the State’s data is beginning to make more sense and it is clearly showing a decline, at least, in serious illness from COVID.  Of course, there is much angst around the Labor Day weekend and schools beginning to reopen.  In a couple of more weeks we should begin to see if those factor have begun to reverse this improvement or not.