Gates Package — page 960

* For me, 10 That What each one tries to do, in essence, is to analyze all the available data--the sources I've described in this chapter, and many others, such as mobile phone data and Google searches--for two purposes. One is to determine why something happened in the past, and the other is to make an educated guess about what might happen in the future. It was computer modeling that showed, early on, that even if only 0.2 percent of the population got infected with COVID, hospitals would be overflowing with patients in no time. My chief experience with disease modeling comes from working with IHME. * For me, one of COVID's chief lessons about modeling is the extent to which every model relies on good data, and just how hard it can be to get that data. How many tests are being done? How many are positive? IHME had all sorts of trouble finding out. Some U.S. states didn't break down their cases by location or gender. Sometimes, reporting would pause over a holiday weekend, and then all the cases would get submitted on the first day people were back in the office, and the modelers would be left to estimate what had really happened. I also couldn't help noticing how often news reports about some modeler's latest findings would leave out important nuances and caveats. In March 2020, Neil Ferguson, a highly respected epidemiologist at Imperial College, predicted more than 500,000 COVID deaths in the U.K. and more than 2 million in the United States over the course of the pandemic.10 That caused quite a stir in the press, but few reporters mentioned a key point that Ferguson had been very clear about: His predictions assumed that people wouldn't change their behavior--that no * Like people in any profession, modelers have their disputes and disagreements. IHME was criticized early in the pandemic for making forecasts that were too optimistic. But they listened to the feedback and improved their model, as any good scientific organization would.