The argument being made by you and the "data scientist" are that the actual cases are way too low. The death cases are accurate however so that just means that the reported death percent is an order of magnitude too high.
Ultimately we are probably looking at a death rate giving the entire US population of 1/3 of 1 percent or something like that.
It will be interesting to look at this years flu/virus death rate when it's all over and see if the numbers are out of the norm.
That's not to say that we shouldn't be doing everything (and more) that we are currently doing but the predicted additional deaths are likely to be overblown.
I am not sure that it is as simple as that...
For example she mentions as OF NOW there may be as many as 1.43 million COVID cases in the USA and even with that blind spot the death toll is a conservative 3148
the 1.43 M estimate is an evolving figure and only relevant to the day it refers to.
extrapolate it this way:
1.43 million cases ( some confirmed) 3148 deaths
Assume total infection cases possible = 300 million
300/1.43= 209.7
3148 *209.7 = 660,135 fatalities
Reverse factor: for every 1 death there are 45,425 total cases ( hidden and confirmed)
You get a fatality rate of just 0.22% which is similar to the flu ..(?) assuming an unchanging ratio scenario.
However this assumes that the ratio of community transmission does not increase (unlikely) and that hospitals are not over whelmed ( which is most unlikely)
Community transmission increasing daily and health system failure appears inevitable at this stage
With poor containment ( social distancing) and an over whelmed health system the fatality rates will be considerably more than 0.22%
So..
If the USA can not slow the hidden transmission rates the actual death toll will be much more than 660K
Not good at math so I may have it screwed...( too quick for me)
Remember the Data scientist is using only ball park figures and generalized factoring.