My team at UKHSA’s paper on nowcasting norovirus methods has now been published here:
What we didn’t put in the paper but going through this process highlighted to me:
- nowcasting of routine surveillance has related but quite different challenges to in an outbreak / breakthrough epidemic situation
- working out the right baseline model for a nowcast was tricky and I wasn’t super satisfied with where we ended up
- nowcast > truncate and forecast. I think this is quite an obvious result, but I hadn’t seen it explicitly - this is partly so we have an evidence base within our agency.
- leading indicators are hard
- is there an agreed / sensible way for setting the max delay D?
The motivation for this work which we touch on is to improve norovirus surveillance, particularly in the event of a strain replacement event. This new strain actually arrived during the peer review process, so we have been getting lots of use out of this modelling over winter 2024/25:
NHS England » Record number of norovirus patients in hospital