There's been a lot of shifts in thinking recently, including by myself, as some things that were explicable via either the typically bad data from an emerging o

There's been a lot of shifts in thinking recently, including by myself, as some things that were explicable via either the typically bad data from an emerging outbreak or "just so" stories about stochasticity, like the cruise ship cases, or the bulk of the cases in South Korea being due to a single massive spreading event in a fairly insular community, turned out to be...well...more common than would be desirable. The field, as a whole, has started talking much more about mitigation, and less about containment, because attempts to intensively trace and isolate cases are only partially holding the line. That being said, important caveats to all of this is: - As Marc Lipsitch notes, those estimates don't tell us how many of those cases will be asymptomatic or mild, because we don't have good numbers for that yet. The difference between "Under the weather for a few days" and "Needs respiratory support in an ICU" is obviously massive. - Epidemic models inherently assume exponential growth (because that's how epidemics work). But that does mean that small perturbations in how we understand how the disease "works" can dramatically chance projections.