[Rain Man]

I think the way prioritization should work is that you build a model of life expectancy losses by demographics. You build a first-order model to look at life expectancy changes for people who get it, and then you add a second-order effect to account for their impact on others. Whoever has the biggest expected losses gets it first, and then you work down the list.

For example, you look at an 85 year-old in a nursing home. What is the likelihood of that person getting it, and how many years of life do they lose? Who can they spread it to, and how many years would those people lose?

Now you do the same thing for a health care worker. What's their likelihood of getting it, and how many years do they lose? Who can they spread it to?

Just build categories and work down the list. You can keep expanding the categories as you go to get more precision.