Two models were published today. One from Imperial College in the UK which is a revision of the model we wrote previously.. They have downwardly revised their projections. The other, which I refer to below, is from Chris Murray at IHME in Seattle. This one is just beginning to reach the media here.
Please find also a piece on the perils of modelling for decision making.
Here is what our team does not like about either model. There is an issue with the key underlying assumptions informing these models and the predictions - particular infection morbidity/mortality rates.
Wuhan denominator is patient rates, not the general population. This means there is selection bias.
CDC data reported in MMR are essentially the same.
In NYC, there is tremendous selection bias on who is being reported as a case including in NY where we have no pop surveillance in place or even a call line to be able to report case definitions.
We have not yet seen a strong estimate populate prev of infection. But some suggest we are further along than we think vis a vis how many of us are infected, in particular in a dense urban center where we are all about mixing!
There is also so little we still know for modelling on the role of underlying conditions or age-specific rates of severe disease. As we get more precise data, then the models become more robust. To that end, it would be great if hospitals in the US and could start pooling patient data. And when places like UK and France start doing mass testing (UK to start next week) we will know more.