@BobKerns CDC addresses this partly by having designated Health Service Areas that constitute contiguous areas of multiple counties--this being part of basis for their (very problematic) "Levels" semiotic.

"Underserved" is interesting idea: implies a baseline of population health, in turn a function of socioeconomic demographics. That said, yeah, some counties appear pretty consistently among top 10 each week, which might read as inadequate "supply".

2/3

@BobKerns Not sure what you would expect "normalized" utilization to achieve but add a layer of abstraction.

If your question is about capacity vs. occupancy (different from disease burden, as occupancy includes elective procedures that may be or have been deferred for extended periods; may not reflect patients sent home early to make room), see the national charts above in this thread.

As for "when I end up in ICU": this assumes any beds in ICU.

3/3

@BobKerns Another note regarding "what will the situation be":

As staffed beds is a measure of staff, not a measure of architecture, the situation stops being terribly clear once we approach or exceed 100%.

A hospital at 100% could be two beds full in every room. Or it could mean patient treatment on gurneys in the cafeteria and parking garage.

What we do know at that point is that patient care is not of the quality one would expect at 80% capacity.

@beadsland That’s a very good point! Is data available on physical vs staff capacity? Even that obscures equipment limitations, etc.

The ultimate reason to ask these questions, is to marshal the needed resources to fix them. Major capital expenditures vs staffing have very different pathways.

@BobKerns Yeah, the only reason we have staff capacity is due to the pandemic, this data only exists in relation to covid-19 beds. Worse, HHS hands over management of this data to the CDC in January, after which...

Statistics are structural.

So no, physical capacity data does not, to my knowledge, exist. Let alone real-time data about triaging equipment and supplies due to materials having been ordered for fewer patients than being provided for.

1/2

@BobKerns That said, once we start getting into long-term questions, it perhaps stops being about datasets and more about data collection.

Major capital expenditures vs major budget increases for hiring and retention are questions answered at the local level. One can always examine floor plans at the local level to gauge physical capacity.

This doesn't address there being fewer employable doctors, nurses, techs, etc. than there were two years ago, however.

2/2

@beadsland While many issues rest in the hands of local decision-makers, my thinking goes toward establishing standards of care. And also, identifying excellence.

Yes, it absolutely becomes about data collection. I’ve been disappointed in our data collection at every turn. I expected we had more to build on from global influenza surveillance, Inconsistent categorization ran rampant and was frequently changed.

The middle of a crisis is no time to be sorting these things out, but people heroically did their best, except for the politicians who should have funded the preparation in the first place!

I don’t think non-data people really appreciate the value of consistent collection practices. Consistent both between reporters and longitudinally.

But I preach to their choir here.

@BobKerns Standards of care are challenging at the best of times, let alone when the reigning zeitgeist is "We don't want to know."

That said, inconsistent categorization is the name of the game regardless. Categorization is a function of expectations, and vice versa.

This is what is meant by the phrase "Statistics are structural." Data collection is always situated, always situational.

Cf. Annemarie Mol, The Body Multiple, p. 133.

@beadsland A lot of the chaos I saw was from there being no shared process at all, which too often opened it to political manipulation and wishful thinking.

Sure, we always have to adjust the data to mitigate inconsistency. And then total laymen will come along and accuse us of “manipulating” the data. Having followed influenza data for year, I just expected more planing and preparation than turned out to be the case. I’m convinced we could have done better—but not by making decisions state-by-state and county-by-county on the fly.

I’m reminded of when ocean temperature data showed an unexpected rise. It was traced to a change in how measurements were collected—manual over-the-side measurements, vs engine cooling water inlet temperatures. Easily adjusted for when identified. Climate deniers screamed about the adjustment!

I once lit into David Sirrota for publishing a number—% of donors with oil ties, but stripped of the context that Beto was running in a state where much of the population had oil ties, and his comparisons were not. He INSISTED he was JUST REPORTING THE FACTS, doing data journalism. But stripped of context, it ceased to inform and instead mislead. A statistic is never just a number (and it’s a stand in for a probability distribution, but let’s not get into that).

I felt vindicated when he later went to work for the Sanders campaign!

I always wanted to do a visualization of how COVID case rate curves evolved with time, as reports came in with varying delays. The delays were part of the situation often ignored, leading to premature gloating of “progress”.

I didn’t overcome the challenge of locating the necessary “time series of time series datasets”. I should have bit the bullet and collected it myself, collating the periodic individual time series.

So of course, I love your quote, and the mortality data it gives! Epi model predictions in general are self-invalidating, to the extent anyone pays attention and acts on them, changing the situation.

And then, of course, that becomes “evidence” that the models are “wrong”.

Anyway, i do understand how challenging data collection is. I didn’t mean to suggest that I expected consistency. I just expected that influenza surveillance would have been a template that would have saved a lot of early iterations. A new reportable disease isn’t really a new situation.

If US lawmakers had funded an international effort on emerging disease surveillance, I have little doubt it would have existed.

@BobKerns Ah, but now we're getting back to the question of whether disease surveillance is something we as a society want: if it is something we want to know.

For with such knowledge comes responsibility. Specifically, responsibility to act. Unfortunately, such action is not simply the responsibility of politicians, but the responsibility of the members-in-part of society-as-whole for which those politicians are merely representatives.

1/2

@BobKerns Our #RentierSociety is one that has for generations allowed preventable deaths by influenza year after year, because despite what data surveillance was in place, it wasn't something we, as a society, were going to act upon.

That would have implicated the choices of those who didn't die of influenza, but merely participated in its spread, and we weren't gonna go there. The new situation only cast that #perquisitive silhouette in stark relief.

2/2

@beadsland Right, I should have said “when I need to be in the ICU”. And yes, “occupancy” is the word I was looking for.

I’m just trying to tease apart the different factors. We can’t do that with a single color scale. Specifically, ‘where is ICU load high relative to population” vs “where is ICU occupancy high because of inadequacy + COVID + other Influenza-Like-Illnesses”.

The first is a better indicator of prevalence; the second a better indicator of the impact of health inequity.

As for rural areas, I was assuming we’d apply the same technique of defaulting to the state values if in-county ICU capacity is not there or minimal.

There are, of course, demographic factors such as age, employment (farm vs sweatshop, to pick extremes) in our multivariate space. These might affect appropriate ICU levels, in different ways—accident rates, disease transmission.

But to begin to answer my two questions, starting from ICU capacity, we need to disentangle adequacy.

I could take a stab at some visualizations; I’m sure I can find county population data easily enough. But attending to my medical issues takes most of my time these days, so I can’t promise anything.

But I wanted to hear your thoughts before trying anything.

Thanks.

@BobKerns My contention is that if ICU or hospital capacity, the why is not the point.

So, for instance, Warren Co., NY is almost always in top 10 overcapacity. Are they high relative to population or high due to infectious disease burden or high due to being an accident nexus?

From a public health perspective, the only question that matters is "Do they have the resources to handle yet more stress?" If the answer is "No", then what's our response?

1/2

@BobKerns Whether a county's hospitals are over-capacity due to insufficient resources (which, with more and more leaving the health professions, it's gonna be regardless) or due to overabundance of illness (which, given paucity of NPIs, it now tends to be) is a long term question being applied to a short term crisis.

We're being told "We have the tools," while being presented with CDC semiotics of a map that insinuates hospitals are doing fine.

2/3

@BobKerns Now, this is not to say that long-term question is inappropriate or unnecessary.

Being able to figure out where more resources are needed to respond when disease happens (prevalence) vs. where more resources are needed to prevent disease in first place (inequity), is profoundly important. That said, is perhaps not the sort of thing national choropleth is best suited for.

How else to present such a question? Have been thinking cartogram.

3/3