[swalsh]

I've been wondering, it seems like people have been setting up systems where multiple patients are connected to a single ventilator (with individual regulators?) If that works, maybe it might be more efficient to solve the supply problem, not by building a million cheap individual ventilitors, but rather a few thousand mega ventilators designed for multiple people.

[notatoad]

Putting multiple people on one ventilator requires matching lung capacities, it seems like you can get away with 2-4 people on a ventilator if you really need to and have enough selection of patients to pick matching ones, but more than that wouldn't be feasible.

(I'm not a doctor and my only source of knowledge on this is other hacker news comments)

[enchiridion]

I think that algorithms/ML people will have something to offer here. This essentially changes triage into a hard optimization problem.

IANAD, but it's not just a problem of matching people in the current pool. You also have to plan for incoming patients. Having the parameters of the ventilator not perfectly match the patient likely affects the probability of survival in a smooth way (with in some bounds).

So now the problem becomes minimizing the total death by maximizing the average likelihood of survival. You have to take the patient ventilator parameters into account (tidal volume, lung-compliance, weaning, etc) , as well as information about the distribution of those likely to be sick at the same time (which will change over time based on behavior).

[numpad0]

okay so how long does it take to certify that and how different is it going to be compared to just have General Motors produce under license on existing designs

[enchiridion]

You could have a workable system up in a matter of days.

Supply chains take a long time to spin up.

Why not both?