Texas COVID Stats – CDC Weeks 33 & 34

For CDC Weeks 33 & 34 (ending 8/15 & 8/22) the COVID stats for Texas left most of us trying to interpret them shaking our heads.  There are multiple reporting problems and gross numbers simply do not make any sense.  Before expounding on that, let’s look at those gross numbers.

The number of tests reported for the period fell to about 45,000 per day from about 50,000 for the previous two weeks.  The number of positive results fell from about 7,500 per day to about 6,500 per day.  The positivity rate inched down from 14.9% to 14.6%.

The State reported 3,066 fatalities for the period, down from 3,574 for the previous two weeks.

At the end of the period, 5,274 individuals were hospitalized for COVID in Texas, down from 8,065 two weeks ago.  COVID hospitalizations have dropped by half since July 22, when it hit a high of 10,893.  COVID ICU beds dropped during the two-week period from 2,742 to 1,878.

Testing

The problems with the reporting on Texas’ testing have been widely reported.   I raised the problems with testing in my previous post and the COVID Tracking Project voiced similar concerns.  This story in the Texas Tribune outlines some of the problems that have surfaced.

The fundamental problem with the test results is that we do not know when the tests were conducted.  There were several days when the State reported over 100,000 results in a day, which clearly included some backlogged results.  Any test results that are more than about a week old are nearly useless for gauging the extent of community spread.

Also, the positivity number is, frankly, unbelievable.  Last week the positivity rate for Texas was 13.9% compared to 6.2% for the U.S. as a whole.  As discussed below, Texas’ hospitalizations are falling faster than the national numbers.  It is just hard to believe that fewer people would be going to the hospital if the current positivity rate was really still deep into the double digits.  Also, the Texas Medical Center, whose testing I would bet my life on, is only showing a 7.2% positivity rate.

Texas is still lagging in testing compared to other large states.  The test results in New York, California and Florida equal 39%, 28% and 20% of their populations, respectively.  Texas is just over 15%.  This probably means that a larger percentage of people going for testing are symptomatic in Texas, which might account for some of the difference. 

But it seems likely there is still some glitch in Texas’ testing.  For example, one source I talked to at the State said they are having difficulty identifying and adjusting for the situation where a single person gets multiple positive tests.  So, I think we are going to have to continue to take the Texas testing data with a huge grain of salt.

Fatalities

Starting in mid-July, the number of COVID fatalities reported by the State began to rapidly escalate from about to 200 per week, hitting a high of just under 2,000 in the week ending August 1.  That was the same week that the State began reporting fatalities based on the actual date of death from death certificates filed with the State.  That change exposed that some local health agencies were woefully behind in reporting fatalities.  Harris County has been particularly tardy.  Currently, Texas is reporting 2,011 fatalities for Harris County, while the Harris County Dashboard is only reporting 1,197.

For about two weeks after the State made the change in its methodology, many of the fatalities were from previous periods.  But it appears that the State is mostly caught up at this point.  This chart shows how the daily fatalities changed with subsequent reports.  It appears that fatalities topped out around the end of July at around 250 per day and has been slowly declining since.

Keep in mind that the fatalities for the far right side of this chart are likely a small portion of those that will ultimately be reported, so it does not imply that fatalities are falling as rapidly as the shape of the curves suggests.

The surge in Texas fatalities has moved it well up in the ranking with other states.  It currently ranks 20^th overall and 7^th among the ten largest states.  At 39 deaths per 100,000 population, it is still well below the national average of 51.  One caveat is that I have not been able to determine 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

As I have previously mentioned, I have come to the conclusion that hospital censuses are the best indication of the extent of community spread of the virus.  It is a real-time indicator of the number of people getting seriously ill and it does not have the potential  inaccuracy that testing does.  It is somewhat of a lagging indicator of new infections because of the virus’ incubation period.

The course of COVID hospitalizations in Texas is truly remarkable.  They shot up from 2,000 in early June to over 10,000 in mid-July and immediately began to retreat.

Every region in Texas has seen a similar decline but locally the decline has been even steeper.  According to SETRAC, general and ICU bed usage have already returned to their June levels.

The fatality rate for those entering the hospital has been running at about 10%.  And it appears that most people dying from COVID die in the hospital.  For example, as I previously mentioned the state shows that 2,011 individuals have died from COVID in Harris County and the TMC website shows that 1,975 have died in their hospitals.  So, the decline in hospitalizations almost certainly predicts a corresponding decline in the State’s fatalities over the next couple of weeks.

Conclusion

It is impossible to reconcile this decline in hospitalizations with a 13% positivity rate.  One indicates that the rate of community spread is declining and the other that it is increasing.  The fatality data leans slightly toward supporting that the rate is declining.  On balance, it seems more likely at this point that the rate of community spread in Texas is falling.  But, of course, only time will tell.