We have a pretty good idea of what the numerator is (hospitalizations and deaths) but we have precisely no clue what the denominator is (infections) for calculating either mortality or hospitalizations. The testing is subject to selection bias (asymptomatic or mildly symptomatic infected people never get tested) and even people who opt-in to testing in most places are not permitted to be tested. We're only testing people who we know are very sick. This means we know that infections are much higher than "confirmed cases," but we are speculating as to how much higher it is. That said, even people who used to be infected and recovered (likely immune) will show up as negative when tested. So our "confirmed cases" numbers as a percentage of total tests is biased upward because we aren't testing entire populations and calculating hospitalizations and mortality as a percentage of total tests is a meaningless number that is contributing to hysteria and poor decisions.

We're in one of two places right now (or somewhere in between) even with all the uncertainty. Either infections are way lower than original estimates and the spread is just nowhere near the level initially projected or the mortality and hospitalization rates are way lower than expected as a percentage of infections. His one policy proposal is to test people and allow people to get back to work if they're negative. Be strategic about how to get the economy going again because an economic recession or depression will cause deaths and calculating deaths from the virus should be weighed against deaths from the economic downturn. He thinks we should be focused on increasing testing, creating anti-body tests, and then getting the testing in the hands of data scientists ASAP. 

As a data scientist myself, I completely agree with and endorse the Dr's approach and comments.