There were dramatic reductions in influenza worldwide. I am hoping to repeat this analysis using international data. Here's what happened in Japan. https://www.mdpi.com/1999-4915/14/7/1417
Thanks! It's been disappointing to see how many people address this question without treating it in a straightforward counterfactual way. It's great to see that fixed.
Good point. I will add a post on this. It’s a complicated issue that actually has nothing to do with flu mortality. I think you are referring to the question of COVID mortality in the elderly and the concept of focused protection. The GBD is maddeningly non specific in what that means. They don’t define the elderly and they don’t explain how to come them separate from the NPI free masses. But even if we assume that we could have totally protected everyone over 65 my model suggests the lifting COVID NPIs would have cost 200,000 lives. In reality, it is absurd to think we could have segregated older population groups in terms of NPI’s. More to come.
Flu by and large most seriously affects the old and the young. COVID by and large only seriously affects the old. You used the reduced flu mortality combining old and young in the model, but the young who did not die of flu because of NPIs *would not have died of COVID* because they're young.
I suspect what would happen if you used the reduced flu mortality only among the old is that the predictive power of the model would improve. Also, the counterfactual estimate of COVID deaths without NPIs would be less overstated, probably by at least 100K if the back of my napkin is right.
I am looking at the change in flu mortality within a specific population, which appears to be a good indicator of effectiveness of NPIs. A huge difference in age stratified mortality could affect results, but the age distribution of deaths is similar to COVID with predominance of elderly. To some extent, the fit of the model validates the variable choices. https://archive.cdc.gov/www_cdc_gov/flu/about/burden/2018-2019.html
The Japanese have masked to what most considered extreme for years before Covid. I wonder who could extend the IRM with that data?
There were dramatic reductions in influenza worldwide. I am hoping to repeat this analysis using international data. Here's what happened in Japan. https://www.mdpi.com/1999-4915/14/7/1417
Thanks! It's been disappointing to see how many people address this question without treating it in a straightforward counterfactual way. It's great to see that fixed.
Thanks. It took me a while to realize that influenza mortality data could be used like this.
Why didn't you look at reduced flu mortality among the elderly in the model? The answer is going to come out a bit differently...
Good point. I will add a post on this. It’s a complicated issue that actually has nothing to do with flu mortality. I think you are referring to the question of COVID mortality in the elderly and the concept of focused protection. The GBD is maddeningly non specific in what that means. They don’t define the elderly and they don’t explain how to come them separate from the NPI free masses. But even if we assume that we could have totally protected everyone over 65 my model suggests the lifting COVID NPIs would have cost 200,000 lives. In reality, it is absurd to think we could have segregated older population groups in terms of NPI’s. More to come.
No, I wasn't talking about the GBD.
Flu by and large most seriously affects the old and the young. COVID by and large only seriously affects the old. You used the reduced flu mortality combining old and young in the model, but the young who did not die of flu because of NPIs *would not have died of COVID* because they're young.
I suspect what would happen if you used the reduced flu mortality only among the old is that the predictive power of the model would improve. Also, the counterfactual estimate of COVID deaths without NPIs would be less overstated, probably by at least 100K if the back of my napkin is right.
I am looking at the change in flu mortality within a specific population, which appears to be a good indicator of effectiveness of NPIs. A huge difference in age stratified mortality could affect results, but the age distribution of deaths is similar to COVID with predominance of elderly. To some extent, the fit of the model validates the variable choices. https://archive.cdc.gov/www_cdc_gov/flu/about/burden/2018-2019.html