Hacker News Reader

> Given the high correlation between COVID# cases (or %Positive) and hospitalizations, why not just use COVID# and "gain" 2 weeks of information?

Because it is not given and - if given - is not reassuringly close to 1. Correlation is positive, alright. But if you calculate hospitalization as percentage of cases, even adjusting for a lag, it is far from constant. Eg in Canada this ratio was 6x time higher in the first wave than in the second. It strongly depends on testing policies and hospital admission criteria.

Sure, over the whole run of the pandemic the case count has to be adjusted for changes in the testing regime. But in many countries the testing has been reasonably constant for many months, so the case count is a good indicator of the state of things.

The most vulnerable people are being vaccinated right now, so we should expect (hope?) that the number of hospitalizations and deaths will decrease significantly even if the number of cases does not.

I'm curious about your data set. The correlation of cases to deaths using data from covidtracking.com for the US over the past year is 0.28-0.3 when I ran it. I slid it by two, three and four weeks.

I'm not sure that a simple sliding correlation really captures how treatments, protocols, and behaviors have changed over time. Leaving aside the winter holidays case peak (which is much more multi-modal than the others), I see two peaks:

* A peak of cases around Apr 11, followed by a peak of hospitalizations on Apr 22, with a peak of deaths also around Apr 22.

* A peak of cases around Jul 22, followed by a peak of hospitalizations around July 26, followed by a peak of deaths around August 4.

If I were going to do a more detailed analysis, I would want to try breaking out individual states/counties (subject to some reasonable population minimum), such that multiple distinct trends nationally don't interfere with each other in the data.

Totally agree. I ran it for New Jersey with the similar results, but it is brute force. Scratching the surface quickly leads to many more variables. For example, more testing would lead to more cases. Then, of course, we'd need to look at how the testing was done (eg random or hospital entry) and what test it was.

I really wish that stochastic testing were discussed more seriously.

> Given the high correlation between COVID# cases (or %Positive) and hospitalizations, why not just use COVID# and "gain" 2 weeks of information?

As we vaccinate the people at the highest risk of hospitalization, the correlation will change: Numbers may stay very similar, but hospitalizations should go way down.

This, and also hospitalizations are a less exploitable metric. Self-selection bias isn't much of a problem, and the number of tests being done doesn't influence the results.

I would look at either hospitalizations or deaths once vaccinations reach a large percentage of the population.

> As we vaccinate the people at the highest risk of hospitalization, the correlation will change: Numbers may stay very similar, but hospitalizations should go way down.

Then we'll know in 2 weeks to change the policy and account for it.

Note that vaccinations will *also* cause the %positive and case# to decline. USA is approaching 15% vaccinated at over 95% efficacy means that you'll have 15% fewer cases (as well as 15% fewer hospitalizations later on). I'm not convinced that cases will become desynchronized with hospitalizations: my expectation is that vaccination will cause a decline in both case# and hospitalization#, roughly in proportion.

But if case# and hospitalization# become less correlated, then it won't take long (~2 weeks to see the first effects, maybe 4-weeks to be sure of the effects) to see such a split in the time-delayed correlation.

-------

EDIT: Why the downvotes? Today, there's a new study being pre-pub'd that shows that Pfizer's mRNA vaccine is ~90% effective at stopping the spread of the virus (https://thehill.com/policy/healthcare/539783-pfizer-vaccine-...).

When you have a vaccine that's both 90% effective at stopping the spread and 95% effective at stopping hospitalizations, then the spread and hospitalization numbers will both go down severely (that is: #cases and #hospitalizations reported both go down).

This assumption that #cases and #hospitalizations will become "desynchronized" isn't necessarily written in stone. Its possible both numbers drop down dramatically in the coming weeks as vaccines are distributed... indeed, its highly likely IMO.

Pretty much all countries are distributing vaccine to the elderly / at-risk population first. We're doing that for the obvious reason that the most-at-risk population is most-at-risk, and thus most at risk of hospitalization.

Concretely, that means hospitalization rates should decline a LOT faster than community spread. This is going to be less visible in countries that have their shit together and are able to vaccinate very fast / have already moved on to genpop, but in most of the EU (sigh), we've just finished vaccinating care homes and 75+. So now, a couple of weeks from now, we should see hospitalization numbers sharply decline because that share of the population represents the most hospitalizations, and will now be mostly immune.

So despite being at like, 5% total vaccinated, we should see a decline in hospitalizations of up to 75%.

Furthermore, given that most of the spread happens outside the most-at-risk in the first place (since those most at risk were those with the most protective measures before vaccines), 5% vaccinations should not mean 5% less cases total.

The #1 group in the USA was not "at-risk" population, but doctors, nurses, and other front-line staff. The idea is that these groups are seeing many, many COVID19 patients and therefore have a big risk at spreading the virus around.

Once this "Priority 1A" group was vaccinated, then age 75+ individuals were vaccinated in Priority 1B. Even then, Postal Office employees and Grocery Store workers (other "high impact" workers) are in the 1B and 1C prioritization queues.

With efforts being to reopen schools, 1B also includes school-teachers (stop-the-spread focus). So a 21-year-old healthy school teacher is prioritized over a 67-year old obese person (despite the 67-year old's higher risk factors).

---------

So at least in the USA: there's a significant effort being placed on high-impact "stop the spread" kind of vaccination effort. There is an element of "save lives", but stopping the spread also saves lives. So its a difficult calculus. (USA has some risk-factor prioritizations... 1B with 75+ age, and 1C with 65+ age + comorbidities like obesity. But again, Grocery Store workers are in 1C as well).

I realize other countries have different priorities. But hey, I live in the USA so my understanding of things will have a USA-slant. These 1B / 1C things are also CDC recommended. Different states (like Texas) are more aggressively stop-the-spread than CDC guidelines (while other states may lean more towards risk-factor based "save lives / prevent hospitalizations"). 50-different states, 50+ different policies. Welcome to America.

Federal long term care program started early January in most states, long before they had finished medical workers.

Michigan reports having given at least 1 shot to about 40% of over 75. Eligibility overlaps quite a lot rather than dictating the precise order.

See the coverage metrics tab for age group coverage in MI: https://www.michigan.gov/coronavirus/0,9753,7-406-98178_1032...

Group 1C is pretty much "everyone".

To take directly from the CDC [0], "Other essential workers, such as people who work in transportation and logistics, food service, housing construction and finance, information technology, communications, energy, law, media, public safety, and public health."

Doesn't that cover pretty much everyone on HN ?

[0] - https://www.cdc.gov/coronavirus/2019-ncov/vaccines/recommend...

------

Second, shouldn't the focus be #1 - stop deaths; #2 - stop hospitalizations; #3 Stop the disease (which is what "spreading" actually is)

> Group 1C is pretty much "everyone".

No, it's not. It's “essential workers”, which isn't everyone in the listed sectors but people in the listed sectors whose work cannot effectively be done remotely; approximately, the people that were exempted and allowed to work on site during the strongest lockdowns, where they occurred at all.

> Doesn't that cover pretty much everyone on HN ?

Probably not; lots of people on HN are probably in jobs that can be and are being done remotely. Even if it did, “everyone on HN” and “everyone” aren't the same thing.

I disagree, "everyone" isn't too much of a hyperbole. The system is gamed and most of the "essential workers" aren't essential in the sense their work could be done fully remote. I work for a large corporation and we were all deemed essential workers. The employees shifts in office are rotated so that it's not a full capacity at any given time. I know a handful of other corporations doing the same thing.

Outside of big corporate tech, I also know a bunch of people working 100% remote but already got vaccinated because they are an employee of pharmaceutical/medical company and qualify as health care workers.

We can be pedantic on how a 1C essential worker isn't everyone but it is a huge percentage. Maybe my sample of people I know in the Bay Area is too small but at least half of my friends can classify as 1C.

Edit: I found some slides from the CDC which totals the 1C estimate as 129M people. So my small local observation isn't that far off from what the CDC expects. https://www.cdc.gov/vaccines/acip/meetings/downloads/slides-...

1C should exclude students, truck drivers, waitresses, etc which are groups you likely don’t have much contact with during lockdown.

Now people are going to game this stuff, but the distribution is wide enough that it’s not really that import to be completely precise.

> No, it's not. It's “essential workers”, which....

Not to be impolite, but _I am absolutely correct_.

See here [0] for the detail, on which I quote, "in Phase 1c, persons aged 65–74 years, persons aged 16–64 years with high-risk medical conditions, and essential workers not recommended for vaccination in Phase 1b should be offered vaccine."

Further, if you read the double asterisks note at the bottom of that page, you'll see this, "On December 20, 2020, ACIP voted 13 to 1 in favor of the Phase 1b and 1c allocation recommendations."

Finally, if you look at the chart on the lower half of the page, you will see that _group 1c includes 199 million people_ (32 + 110 + 57 ), which is, after the 75 million people in groups 1a & 1b, darn near everyone.

So if you think I'm wrong, offer proof, not an interpretation. With proof, I'll gladly admit error, but the facts are very clear.

[0] - https://www.cdc.gov/mmwr/volumes/69/wr/mm695152e2.htm?s_cid=...

> Second, shouldn't the focus be #1 - stop deaths; #2 - stop hospitalizations; #3 Stop the disease (which is what "spreading" actually is)

Because stopping the disease implicitly stops the deaths and hospitalizations, its not very clear that a focus on deaths-only or hospitalizations-only is optimal.

Especially when you consider that the disease will continue to mutate as it exists (possibly making our vaccines less effective or even obsolete). So stopping the disease first-and-foremost might be the most effective way to stop deaths/hospitalizations (especially when mutations are considered).

--------

Turning the R-value from 1.5 to 1.3 means a 14% decline COMPOUNDED PER GENERATION. After one generation, its 14% fewer cases (and 14% fewer hospitalizations and 14% fewer deaths). After two generations, that's 25% fewer cases (and 25% fewer hospitalizations and 25% fewer deaths). After three generations, its 35% fewer cases (and 35% fewer hospitalizations and 35% fewer deaths). Etc. etc.

As such, "stopping the spread" has a benefit that grows exponentially every week or two (the generational period of this virus). Exponentially growing its results and efficacy.

Keeping our eye on the bigger picture, it seems like stopping the spread is the best way forward to stop deaths and hospitalizations. I realize this is a bit "splitting hairs" (compared to people who would rather "save lives" and focus on hospitalizations and/or deaths). But... it seems like the superior strategy in my opinion.

>> approaching 15% vaccinated at over 95% efficacy means that you'll have 15% fewer cases

That’s not how it works, you’re missing a variable (prevalence).

...meaning that to reduce the number of cases evenly, we'd have to choose people to vaccinate by random, but we're not doing that; we're choosing by risk, and those at higher risk of severe disease are less likely to contract an infection because they move around less and meet fewer people.

However, we could speculate that perhaps we should in fact put more priority on the groups that have most infections, not highest risk? Because the restrictions impact their lives (of young people) most.

However, I'm quite sure that the priorisation of old people will continue, except possibly in places where priorisation is done by money (the rich purchasing vaccinations).

>> those at higher risk of severe disease are less likely to contract an infection

If you were running a business and there was a relatively low incidence of an utterly catastrophic outcome, you’d buy insurance for the eventuality.

If you were running another business with a high occurrence of a mild outcome, you’d price it into the cost of doing business.

Insurance = vaccine. Cost of business = stimulus cheques.