Attachment

one would wear masks or shelter in place, for instance--but of course that wouldn't be the case in reality. He wanted to show how high the stakes were, not drive everyone into a panic. The next time you hear about some prediction made by a disease modeler, keep a couple of things in mind. First, all models have limitations, and the report you're hearing might have left out some important caveats. The level of uncertainty, for example, can be quite high. Remember Mike Famulare's estimate of 570 cases in Washington state, with a 90 percent certainty that it was between 80 and 1,500? Any report that omitted the range of possibilities left out some pretty important context. Second, every variant is different, and it's hard to predict the impact of each one until there are a few weeks of data available. Finally, modeling needs to be done with a healthy dose of modesty, especially if the forecast stretches beyond four weeks or so. Modelers are on the most solid ground with predictions about the impact of certain steps: "How many cases could be prevented if a certain percentage of the population gets vaccinated? What if they mask up instead?" *** What everything in this chapter adds up to, I believe, is a clear agenda for the kind of disease surveillance that we need in order to prevent pandemics. One step is to invest in all the elements of a robust health system that make it possible to detect and report diseases, as well as treat them. That's especially true in low- and middleincome countries, whose health systems are often underfunded. If the doctors and epidemiologists don't have the tools and training they need, or their national health agency is weak or nonexistent, we're going to keep seeing outbreak after outbreak.