Test random samples of people in EVERY congressional district to extrapolate numbers

[pat_k]

Last edited Mon Mar 16, 2020, 12:17 AM - Edit history (3)

Let's get some meaningful data to better estimate ACTUAL incidence rates!!

We have 435 congressional districts, each one representing about 750,000.

There are pollsters able to identify random samples of people for polling in every single one of them. These companies are equipped to rapidly recruit random samples for any type of poll. For example, identifying people willing to be tested for COVID 19, regardless of their status.

There are different ways to determine an ideal sample size, but one "rule of thumb" is 10% of a population up to a maximum of 1000.

Recruit a random sample of 1000 willing to be tested in EVERY district.

Coordinate with labs and collection sites (ideally drive thru) to get these people tested.

Follow up for outcomes. How many asymptomatic develop symptoms. How may positives develop mild, moderate, or severe symptoms by age.

This doesn't just give you data on infection rates in a population -- symptomatic/asymptomatic -- it will identify some number of asymptomatic and mildly ill people who need to self-isolate. Even if the number identified is relatively small, every "spreader" taken out of circulation reduces the R zero, even if only by a very small amount.

Doing this by congressional district could give us a VERY good idea of infection rates and actual distribution -- where to focus resources for "worst case scenarios" and where to focus aggressive containment efforts to keep low numbers low.

Repeat at intervals to track growth (or leveling off) over time.

It has the political advantage of giving every single member of congress something to "crow" about -- how amazing they were in getting this organized and getting vital data to county and state public health authorities.

If you think this makes sense WRITE TO YOUR MEMBER OF CONGRESS and Senators as a constituent. Write to the Congressional Leadership in the House and Senate in your capacity as a citizen.

[blueinredohio]

1. My weasely little congressman speak up?

He's afraid of his own shadow.

[pat_k]

4. Well, doesn't need to be pushed by congress critters. Congression districts just make...

... geographic breakdown of similar populations. It is also a good way to break up metropolitan areas (may identify district with possible "hotspot" relative to others -- something that would merit further investigation to narrow down).

And even if a given congress critter didn't involve themselves in promoting, if public health does, and there is a benefit, they can take credit for it after the fact.

[gibraltar72]

2. Love your idea.

[pat_k]

8. Thanks for the vote of confidence!

[pat_k]

11. Thx (nt)

[NotASurfer]

3. Interesting idea statistically

Some level of random testing would give us something to work with in assessing the actual number of infections versus symptomatic individuals, and help focus the response to best effect.

Now that the opportunity to ramp up testing early is gone, an accurate estimate would be a good tool. If we have enough time before being overwhelmed and if we can get enough tests evaluated

[pat_k]

5. Well, there are 10 districts in WA

University of Washington virology and infectious disease labs are reportedly ramping up test processing capacity to about 4000 per day. If even a quarter of that were devoted processing tests from people sampled, they would have results for all districts in 10 days. Perhaps space out collection, starting with most densely populated districts.

Whatever the start date for a district, if they retest a new sample every 15 days or so, that would at least give us a curve for the district, even if time frames are different.

On edit:
P.S. Doing one district at a time could involve five (or more) "mobile" drive through sets ups. If each were large enough to collect 100 samples in a day, you would have the 1000 samples collected in 2 days.

[Steelrolled]

6. I would like to see this as well.

But I think the question is how the results would affect our response. It might influence public perception, for better or worse. We do know cases are increasing rapidly everywhere. We are quickly headed towards a major lockdown and I doubt we see that changing.

[pat_k]

7. Response depends on results.

Last edited Mon Mar 16, 2020, 12:25 AM - Edit history (1)

It's hard to predict response in advance of results. However, whatever the estimated incidence rates based on sampling, you'd have some concrete numbers from which you could more accurately extrapolate probable needs and compare to locally available resources -- hospital beds, ventilators, and so on. This type of analysis would identify districts with the widest gaps. Those districts could be targeted for FEMA response to put up temporary hospitals, deploy military physicians and other trained medical personnel, or whatever.

Any district with unexplained high rates could be targeted for investigation to narrow down to "hot spots" within the district that could require more aggressive measures.

If the estimated incidence in a district that is "resource rich" is relative low, for whatever reason, more aggressive testing, tracking, and quarantine efforts may be effective in keeping numbers low. As far as practical, non-covid-19 patients from hospitals likely to be overburdened could perhaps be transferred to hospitals in low incidence areas to free up beds.

Of course, perhaps "turf" wars would prevent such "redistribution" (with every county or city hell bent on hanging on to every bed, every ventilator, etc., regardless of sampling research findings). But it seems to me that our mood of "pulling together" in this would promote better cooperation to save lives.

In any case, repeating sampling tests at intervals of every two weeks or so could identify "leveling off" in some regions and not in others. There could be lessons to be learned in comparing response efforts or attributes of the differing districts.

Those are few ways results could perhaps be used in planning. Although the effort might prove to be of limited use, I think the potential would make it a worthwhile endeavor. I don't think it would interfere with other measures.

[crickets]

9. K&R for the post and the discussion.

[pat_k]

10. Thanks!

Somehow I thought the post might get more attention. Perhaps I'm unreasonably attached to the idea. I would love to see someone with a statistics background weigh in on pros/cons.

In any case, I drafted a letter to Pramila Jayapal to fax to local and DC offices. Want to let it sit and work on a bit tomorrow. Whatever I come up with, I'll fax around to senators and leadership too.

I would think the notion of sampling in some way would be "on the table" but I haven't heard a word about it. I can't find any results of such projects, even from countries that have more comprehensive testing programs going. Perhaps I'm not looking in the right place.

In any case, Thanks for the Kick!

[WhiskeyGrinder]

12. Random testing by congressional district doesn't help to identify hot spots or chain of contact.

[pat_k]

14. What would be a better sampling method? (nt)

[WhiskeyGrinder]

16. What's the outcome you want? In your OP It sounded political, which generally isn't the best

approach to take to public health.

[pat_k]

17. The last thing in my mind was politics -- I am looking for data.

Perhaps it is a stupid idea, but sampling by congressional struck me as a way capitalize on a regional breakdown into equal areas that might yield useful statistics.

[WhiskeyGrinder]

18. Going on congressional areas would artificially inflate some numbers and lower others when it would

make more sense to look at hot spots to better deploy resources. I mean, it's hard not to think there's a political motivation when you introduce congressional boundaries, and then add that congress members could "crow" about how well they're doing.

[pat_k]

19. I am not a statistician. What boundaries would be best to start with?

I picked congressional districts because they struck me as a pre-existing break down of roughly equal populations/regions that might be statistically useful.

Of course they are gerrymandered in strange ways, but that could actually be useful in statistical analysis.

Statisticians undoubtedly have a better ways to divide up the nation by geography/population that would yield more useful sampling in the US.

Perhaps census MSA's to start with? However, then we are just looking a the enormous statistical areas.

The bottom line is that I would like to see sampling of some kind to get data that would help answer some basic questions;

[WhiskeyGrinder]

20. Geographic boundaries are a terrible way to go about deciding who gets tested.

At this moment, testing depends on who has the tests. There is a shortage, so best practices until we have more tests is to act like you have and can spread the virus.

[pat_k]

21. There is really no way to collect data through sampling in that would be useful?

You don't explain why any sort of geographic sampling would be so horrific. You just state it as a given.

Why would sampling in some way be so asinine? You don't give any specifics.

[WhiskeyGrinder]

24. Not the way you say it in the OP.

Those who are "willing" to get tested? That skews your data and takes away your "random" testing. It's also still not clear to me what kind of data you're trying to get.

Again, it's a moot point because of the lack of tests.

[pat_k]

27. How would you do it?

Every poll is skewed by some bias as far as I can tell.

Statisticians seem to to a decent job of accounting for in analysis

[WhiskeyGrinder]

30. What is it you want to do?

[DeminPennswoods]

13. Any idea that generates actual data is a good one since

there is such a paucity of that right now.

[pat_k]

15. Thx! If there is a better geographic breakdown, i'd love to see it! (nt)

[Drahthaardogs]

22. Isotope dilution method could work

Not sure if we have enouvh infected yet to ise that method

[pat_k]

26. Please point to reference or explain more -- thx!! (nt)

[Drahthaardogs]

32. Its a chemistry method to calibrate based upon an evological model.

Imagine I want to know how many bass I have in my pond. I take ten bass out and tag them, then I let then go. Now, a day later I throw out a net and grab ten baas, one of which is tagged. I then use that as a ratio to decipher that I caight ten bass, one of which is tagged so there should be 100 bass in the pond.

[Johnny2X2X]

23. We don't have enough tests

Won't be getting any more for a while.

[pat_k]

25. WA can't be the only state beefing up capacity

See
https://www.democraticunderground.com/100213104408#post5

[samnsara]

28. yep..in fact the gvt should offer $$ to ppl who are healthy and get tested..

...just so they can get a data base built up.

[lunasun]

29. + Places like S Korea test results have given the world a lot of info so far . Too bad we are behind

Plus I believe their national medical database connects it all better to analyze in real-time

https://www.businessinsider.com/south-korea-coronavirus-testing-death-rate-2020-3
South Korea has tested more than 140,000 people for the new coronavirus and confirmed more than 6,000 cases. Its fatality rate is around 0.6%.
This suggests that, as many health experts have predicted, the virus' fatality rate seems to decrease as more cases are reported.
That's because more widespread testing leads more mild cases to be included in the count.
The US, by contrast, has tested around 1,500 people. The country has 221 confirmed cases and 12 deaths, suggesting a death rate of 5%.
The US' testing capacity has been limited.

https://www.npr.org/sections/goatsandsoda/2020/03/13/815441078/south-koreas-drive-through-testing-for-coronavirus-is-fast-and-free
Can just pull on up if u have a car it’s fast and free

[defacto7]

31. I think it's a great idea if in large and wide enough proportion

Bravo.

[Steelrolled]

33. This is nothing. Since its inception

the stock market has always went up over time.