Subscriber Benefit
As a subscriber you can listen to articles at work, in the car, or while you work out. Subscribe NowIndiana health officials have erred in reporting the state’s COVID-19 positivity rate since the beginning of the pandemic due to a problem with the way it was computed, resulting in a lower rate than would be accurate.
Dr. Kristina Box, Indiana health commissioner, said Tuesday that she believed that the state’s positivity rate would be two to three percentage points higher once it is corrected.
“The error we discovered is in the software logic coding used to calculate our positivity rate. The error has existed since we began reporting the data,” Box said during Gov. Eric Holcomb’s weekly press conference on the pandemic. “We’ve been working with outside data scientists to identify the issue and to develop a fix.”
However, other statistics related to the pandemic reported by the state—including the number of cases per 100,000 individuals, the number of deaths and the overall test counts—have not been affected by the error, Box said.
Holcomb and Box emphasized that the movement trends in the state’s reported positivity rates throughout the pandemic closely mirrored the trends for the accurate rates. Thus, state officials wouldn’t have made any different decisions in combatting the pandemic.
On Tuesday, state health officials reported that the state’s 7-day positivity rate was 12.2% for all tests and 24.2% for unique individuals.
In general, the positivity rate for all tests is calculated by taking the total number of COVID-19 tests with positive results and dividing by the total number of tests performed. Using the “individuals” methodology, a patient is only counted once, even if that person is tested multiple times. The 7-day rate tracks the positivity rate’s average over a 7-day period.
The coding error affected computation of the state’s positivity rate, as well as rates for individual counties. The state plans to resolve the error beginning with its Dec. 30 statistics.
Also on Dec. 30, the state plans to alter the methodology of computing the 7-day positivity rates for the state and counties. Instead of adding each day’s positivity rate in a week-long period and then dividing by 7, the state will add up all of the positive tests for the week and divide by all of the tests done that week to determine the week’s positivity rate.
“This will help to minimize the effect that a high variability in the number of tests done each day can have on the week’s overall positivity, especially for our smaller counties,” Box said.
Please enable JavaScript to view this content.
And to think that many people want a larger government. This is what you get.
Yes, as if large corporations don’t have similar issues. Non sequitur.
it actually does follow, with the difference being that in a large corporation, the employee who made this kind of mistake would be fired for carelessness and/or incompetence. Unfortunately, we’re stuck with the politicians and their appointees until another election rolls around (if we can even trust the elections anymore).
When Indiana changed the reporting methodology to a 1 to 5 scale, different from the rest of the country, it masked the significance of our data and made it difficult to do comparisons. Plus, the dashboard is slow and combines county data, and/or obscured it.
I’ve been using the NYT maps, which tracks the whole nation, down to county levels. Also, much easier to navigate and compare. Still, I doubt that the data is accurate. How can you teack a virus that is asymptomatic for a high percentage of people?
Right on Bernard just another example of people who’s paychecks aren’t affected but can keep screwing up and still keep their phony baloney job’s.
kind of like a weather reporter. LOL
Yep. It’s pretty hard to COUNT accurately.
Sounds like someone is about to go down for all of the fraud and spilling the beans.
How many times is this has Box addressed the flaws in the statistics and numbers?
There needs to be an serious audit done for each reported and documented case for legitimacy.
Then make sure all the perpetrators of the fraud are prosecuted for treason by the Articles of Treason in the United States Constitution.
Right Darrell. Do you really believe that millions faked being sick and 300,000 deaths were faked too?
Wow – glad to see the nutcases are everywhere and not just on free media.
The perps will all get jobs in the Pennsylvania Board of Elections…
Mistakes, yes. Let’s correct them.
But Treason? Did they sell the data to Russia?!?
Wouldn’t this level of prosecution be akin to shooting the messenger? They could have quietly made the correction and it would look at if we had a 2-3% improvement. Maybe adjust historical figures down the road. Or not. Would this be preferable?
Props for sucking it up and being transparent.
+1
Well said. Bit late now to critique government after the last four years…
Isn’t the positivity rate a 4th grade math function? Why do we need teams of software engineers?
In any complex system, there will be errors.
It’s possible that someone was careless or clueless, but far more likely that this is simply natural human error.
Actually writing code is about a third of the work involved in building a reliable system. Given the very wide space of possible unexpected combinations of inputs to any large software project, the number of permutations of inputs that would need to be checked in order to assure that every possible scenario is handled correctly becomes untenable. There’s always a tradeoff between time and resources spent in testing vs. how much functionality can be built.
For sure it’s a bad look that something was done wrong that is a straightforward calculation, but the likelihood of SOMETHING, SOMEWHERE going wrong is actually near 100%.
So I’m not happy that this was wrong, but I’m satisfied that they admitted and are fixing the error.
Must have used dominion machines…
Malfeasance. And to think all these restaurants have shuttered because of this shamdemic is even more of a political hitjob. Shame on politicians. Shame on the sheep.
Yep, they should have been closed sooner.
Just More liberal FAKE NEWS! More Left wing kids do Heroin and elicit drugs FACT!
The notion that private businesses aren’t filled with incompetent morons is laughable. You obviously haven’t spent much time in any private business. They are often promoted, particularly when related to the boss.
Cases in to the total population.
What not enough fear?
Box is a gynecologist, is this a surprise?
Embarrassing for all involved. Especially since it had been part of the calculation since the very beginning. Dr. Box should take accountability and do the right thing…
The explanation I heard is that the old method, they would calculate the positive rate by day, add them up and then decide by 7 to get the weekly positivity rate.
The new method and the method that most states us is is to add up the positive tests for a week, and add up the total tests for a week and then divide to get the weekly positivity rate.
If things are pretty constant, the two methods give pretty close answers, but when there are wide daily swings in either set of numbers, the answers start to diverge. Being a former programmer and decent with statistics, I would have never thought of trying the first method, but now I have heard the details, I could see how that might make sense to some people.
For all of you looking for an evil government plot, you might have to go some where else.
Averages are interesting, and sometimes troublesome statistics. The article describes an adjustment to the method, not a software error. The original method, as described, calculates a 7-day average that “overrepresents” the impact of rates on days when less tests are taken and “underrates” the impact of rates on days when more tests are taken. This is particularly an issue in smaller counties or when the number of tests taken change dramatically on a day-by-day basis. The issue probably became obvious at Thanksgiving when folks took a “testing break” and analysts were looking to identify the impact of Thanksgiving get togethers.
I’m hoping that duplicate test by individuals are filtered out in the new method as well, although there seem to have been some changes in testing protocol that have resulted in non-essential workers getting duplicate tests less often (making the filtering less important). Independently, it would be interesting to look at testing on populations that get multiples tests (like NBA players!).
The story is unclear as to whether or not there was any actual “software error”, which would indicate that the original methodology was implemented incorrectly. If the original method was correctly coded, then the story should be retitled to reflect the fact that there was an error in the choice of methodology, not a correction in software coding. That would result in a more coherent story. It seems the state is unwilling to admit the choice of their original methodology and chose to blame the programmers. Not surprising, but disappointing (for the programmers, who are mostly invisible here).