Disease trajectories in the absence of the prospect of hospitalisation

[bnlawrence]

The disease trajectory for any individual is currently set at infection. That trajectory includes the prospect of hospitalisation with or without an ICU visit and an endpoint of death or recovery.

In the event that hospitals are full, anyone whose trajectory would have included hospitalisation, but who cannot be hospitalised, is likely to die. This may be unlikely (full hospitals) in the UK in a second wave, but it could be likely in refugee camps etc.

An easy modifications to address this would be to check at the point where we would want to move them to hospital. If that is not possible, then whatever trajectory they would have been on (recovery or death) could be immediately set to result in a non-hospital, but long trajectory to recovery, or death at the next timestep, with a very high probability of the latter. More complex modifications may be possible at some other future time.

It might be argued that this would break the IFR stats, but I think the situation where we have full hospitals may not be one that was properly included in the IFR used. As I understand it we have used IFR probabilities and policies associated with the NHS approach to hospitalisation, in which case anyone we would have wanted to hospitalise would almost certainly have needed (at least) oxygen support at or near the point of admission. If it was not available, it would likely have resulted in death.

[florpi]

This is something we are definitely considering, the problem is how do you determine the probability of dying given that the hospital was full? We try to estimate it using mortality data on a disease we have more data from different countries with varying quality of healthcare. But you can imagine that this probability can be correlated with many personal attributes, such as age. All this is still work in progress, but we'd be very happy to hear about suggestions :)

[bnlawrence]

Well, isn't that related to how you decide on your initial disease trajectory? If you have a trajectory that requires ICU, but when there is no ICU, you've more or less explicitly decided that particular person has a case that would result in death without an ICU. The situation is obviously more complicated if they "only" require hospitalisation without ICU support. However, I think using even something as unfounded as 50% of such patients would die would be more consistent with the way the model is set up than assuming they all survive. One could of course improve on that with time.

[bnlawrence]

(to be explicit, I think it is consistent with the way the model is set up with IFR rates for a hospital system which is not overloaded imposed by the way trajectories are instantiated at infection time, to then effectively truncate that trajectory when the hospitalisation is not available.)