Remember when I said that PHMDC's COVID Case Percentage graph amounts to a giant, steaming pile of poo? OK, so I didn't get quite that colorful; but I'm saying it today. Just one more of those cheap ruses the agency has leveraged from the very first iteration of its COVID dashboard in March 2020--though it's gotten a lot worse since PHMDC overhauled the dashboard in January 2021.
I promised to connect additional dots regarding this graph. Today's the day.
As a quick reminder, here's a screenshot of screen 4, which houses the primary Case Percentages graph, along with a hemi-circular 7- or 14-day Average graph (depending on the setting). I've made some annotations in order to highlight my top three objections and will briefly explain those criticisms before moving on to dot-connection.
Objection #1: Screen title misleads viewers
Viewing a dashboard that provides all of Dane County with COVID-related data, most people will read the phrase "Percent Positivity" and assume screen 4 conveys how many of Dane County's approximately 561,504 residents ( per April 2020 U.S. Census) were COVID-positive on any given day (or, for the hemicircular graph at upper right, how many were positive on average across the prior 7 or 14 days).
However, this page reports only percentages of those who took a COVID test on any given day AND received a positive result. As discussed here: the number of Dane County residents who took a test on any given day will always be SIGNIFICANTLY less than the total Dane County population; and the number of people who tested positive on any given day will remain a fraction of the number tested. So, many or most viewers will unfortunately believe that the case percentages shared on this screen represent case levels several orders of magnitude higher than reality.
Objection #2: No context or transparency
As noted in my last post, unless one knows the numerators (daily cases) and denominators (daily tests) that produced the percentages on screen 4, those percentages have zero discernible meaning.
True, you can obtain daily cases and daily tests from screens 2 and 3 respectively. Based on the ridiculous level of fear in Dane County, though, it's clear that too many people aren't doing so--around 30 county board supervisors being at the top of the list. PHMDC's fear narrative benefits substantially from this lack of immediate context and transparency. What's worse, even fewer dashboard viewers will take the time to put PHMDC's COVID data in the context of total county population, as the agency has staunchly avoided this more accurate perspective on its dashboard for the entire endemic.
Objection #3: Explanatory text should clarify, not further confuse
All of the "may-ing," "might-ing," and "or-ing" in the explanatory text on screen 4 outs PHMDC as having provided a graph from which one can extract almost no meaningful data or clear understanding...ever. The explanatory text amounts to either: 1) an epic fail on the part of a well-meaning but inept data team, or 2) gobbledygook intentionally used to make misleading "data visualization" appear honest.
A quick FYI
The default-view shuffle that I wrote about last week is still occurring. The Daily Tests graph on screen 3 has recently zigged back to an all-time view, while the Case Percentages graph on screen 4 has again zagged to a 14-day default. So, while PHMDC isn't obscuring the historical/contextual test data today, remember that they did from late May 2021 onward. With default views flipping back and forth like light switches, just wait...PHMDC is likely to obscure the historical daily case data again any minute now.
On to the dots...
Why has PHMDC spent so much time obscuring the all-time view of case percentages and, until recently, daily cases and daily tests...? Simple! The longer view of all of these graphs has pretty persistently threatened to expose just how far from reality PHMDC's narrative has consistently been. I'll provide two prime examples to show you what I mean.
Real-world example #1: High Denominators:
In January of 2021, Dane County ramped up testing to unprecedented levels. Loads more testing was required, PHMDC and county leadership told us, in order to find all the unidentified cases lurking within our community. The massive push to test, test, test had the additional benefit of temporarily boosting--pun totally intended--the illusion of overwhelming COVID risk, directing attention away from mostly plummeting case counts, starting in mid-November 2020.
This testing blow-out lasted into late April/early May 2021. You can see it plainly in the all-time view of the Daily Tests graph below:
Problem was...the more testing Dane County did, the fewer cases turned up, the lower daily case percentages dropped, and the more consistent those low percentages became. Compare the daily case graph above to the daily percentages graph below to verify this phenomenon.
So increasingly embarrassing did this situation become that PHMDC shifted to 14-day default views of every single major bar graph on its dashboard in late May 2021. They tucked the truth out of sight. PHMDC only restored all-time views for certain graphs this past December, when rising case numbers again provided a way to scare people with big, scary spikes.
Real World Example #2: Low Denominators
In August 2021, respiratory illness season began about a month earlier than usual. Case percentages predictably began to rise again. However, in comparing the 2020-21 season to 2021-2022, it turns out that when you examine total daily test numbers (DENOMINATORS!) they have has decreased by 8.4% season-over-season, to date. See table below:
RESPIRATORY ILLNESS SEASON | TOTAL TESTS | AVERAGE DAILY TESTS |
---|---|---|
Aug 1, 2020 - Jan 9, 2021 | 637,263 | 3,934 |
Aug 1, 2021 - Jan 9, 2022 | 583,875 | 3,604 |
Percent Decrease | 8.4% | 8.4% |
Note: I've only summed and added through Jan 9th, because daily test numbers can fluctuate wildly in the first three days of reporting. Frankly, they can fluctuate quite a bit for a while. But PHMDC at least acknowledges that the first three days of a daily test number are quite preliminary.
It's important to recognize that we'll always have more sick people in fall and winter. And to be fair, Dane County has recently claimed a shortage of test kits. However, the fact remains: Dane County has, overall, tested less this season than last. That testing decrease has contributed significantly to case percentages looking worse this season.
Moreover, when testing decreases during a respiratory illness season, it's a reasonable probability that the test pool will include a higher percentage of folks who genuinely don't feel well. As a general rule, we can assume that symptomatic people have a greater likelihood of receiving a positive COVID test result. That's called SELECTION BIAS. If Dane County did random testing, the pool would likely include significantly more well people; and we'd almost certainly see with greater ease that COVID is far less prevalent in our community than PHMDC so clearly wants (needs) us to believe. In point of fact, we actually approximated random testing through much of the first half of 2021 by testing loads of asymptomatic people. And we saw in Example #1 above that it didn't work out well for PHMDC's narrative.
So...what do we learn from these two examples? Neither case counts nor case percentages necessarily indicate that COVID poses a dire threat to Dane County. We have to look at numerators AND denominators...together.
True markers of reality
If Dane County were in serious danger, we'd see not just spikes in cases and case percentages; we'd observe proportionally steep rises in:
Hospitalizations
ICU occupancy
Deaths
However, we simply don't see such proportional increases--not even when we include the record-high 202 COVID-related hospitalizations recorded on Jan 11th. Average hospitalizations are up only slightly season over season, from 89 to 92--an average of just three additional daily hospitalizations spread across multiple facilities. Not exactly a terrifying increase...unless Dane County hospitals have manufactured staffing shortages for themselves. Also, remember that Dane County hospitalization numbers include patients from other counties. Oh...and then there's this narrative-shredding acknowledgment about hospital admissions.
Dane County ICU occupancy isn't skyrocketing with the rising case count. either. We're up about 21% over last year, but that percentage has dropped from 27% earlier in the season, and average occupancy has been stable for weeks.
If we compare Dane County COVID-related deaths for Aug through Dec 2020 versus the same span in 2021 (even generously including "probable deaths"), we discover a whopping 61.2% in 2021. With just two deaths so far in 2022, we're not at all on track to outpace the total death count for the 2020-21 season. Our circumstances have vastly improved, in fact. And that's without considering mis-categorized deaths--a phenomenon for which Dane Undivided has accumulated a growing amount of evidence.
Better context and perspective
Again, PHMDC has assiduously avoided any opportunity to contextualize its dashboard data in relationship to Dane County's total population. Yet, looking at cases as a percentage of the total county population tells us far more about COVID risk than knowing only the number of tests taken. The only excuse for refusing to provide this more balanced perspective can only be because PHMDC wants to propagate a high level of fear and panic.
To demonstrate how NON-frightening Dane County's COVID situation really is, let's look at Jan 6, 2022, as an example, and see how profoundly figures differ when calculating cases as:
percentage of tests taken; versus
percentage of total county population
Daily cases as percentage of daily tests
Looking at the 14-day view of the Daily Tests graph below, we can easily obtain the number of tests PHMDC currently claims were performed on that date: 9,377.
Next, let's head to screen 2 to grab the number of daily cases for Jan 6th: 1,829
Screen 4 tells us that the daily case percentage for that day was 19.5.
We can easily verify this daily case percentage with a little division:
1,829 cases / 9,377 tests = 0.1951 or 19.5%
But again, this is a very narrow view that likely wraps in a much higher selection bias.
Cases as percentage of county population
Look what happens, though, when we use the better denominator (pool or lens) of total county population. Let's first consider the Jan 6th daily test number from this angle:
9,377 tests / 561,504 Dane County residents = 0.0167 or 1.67%
Now, 1.67 is a very small percentage of Dane County's population. But...the number of people among those tested who actually received a positive test was actually far, far lower. Let's do that math:
1,829 cases / 561,504 Dane County residents = 0.00325 or 0.33%
That's right...just 0.33% of Dane County's total population received a positive COVID test result on Jan 6th, 2022.
And that percentage is based on dashboard numbers from Jan 10th, the day I took screen shots for this post. As of TODAY, the dashboard says daily cases for Jan 6th have DECREASED by 114 (1,715 cases), reducing the above percentage even further to 0.31% of Dane County's total population that tested positive that day.
I leave it to you to connect your own dots on why PHMDC is so determined to promote such a skewed and frightening but ultimately false perspective. The bottom line, though? The spikes PHMDC seems so determined to show us provide zero proof of COVID peril.
コメント