No worries. I would strongly suggest reading this piece by the way: https://medium.com/@noahhaber/flatten-the-curve-of-armchair-epidemiology-9aa8cf92d652
Anyway, I don’t necessarily have a better estimate of ‘critical’ cases, because it is extremely situation dependent. The rate in South Korea is markedly lower, in Iran a bit lower, and in Italy higher. It appears to depend a lot on who is infected and how, although we won’t know this for sure for some time.
The ICU days will obviously invalidate your original prediction entirely, but I imagine using an exponential curve you’d still see Australia overwhelmed fairly soon. That’s the thing, however — you can’t just extrapolate exponentials indefinitely during an epidemic. People were doing that in early February for China and predicted hundreds of millions of cases by now — have their predictions come to pass?
With regards to the R², the point is that you’re fitting a curve made from your data source on to the same data. This means by definition it will be nearly 100%, because it’s a curve plotted on data that it was used to generate. Unless I’ve misread the model entirely — possible — you are using the data from Australian cases to construct an exponential function that you then fit to Australian cases, thus guaranteeing a high R² but saying nothing whatsoever about the reliability of the model.
The real issue in all this is that everything is uncertain. You can’t just make predictions about things, because none of your assumptions is firm at all. The number of critical cases will fluctuate, the rate of infection will move, the number of ICU beds will move up and down. If you actually want to make any predictions, you should produce graphs for ~every~ scenario, that is what my colleagues in pandemic modeling are doing right now. Otherwise, I’d suggest that the piece is pretty unhelpful and almost certain not to be an accurate reflection of what’s to come.