Day 286: Big Bad Data

The autumn wave has brought new countries into the million case club, most recently #10 Italy. In the US, curfews and lockdowns are spreading: Starting tomorrow, New York City will sleep at 10pm along with the rest of the state, and indoor dining is off the table in San Francisco. Forbes is tracking them all.

In pre-2020 epidemiology, locking down was always the wrong decision. In late 2020, lockdowns are often excused with bad logic or bad data; the Wall Street Journal has gone into some depth over the bad mask compliance data coming out of the IMHE and ending up in journals like Nature:

More than 100 news outlets trumpeted the study’s findings in the days following its publication—“The Price for Not Wearing Masks: Perhaps 130,000 Lives” was the New York Times’s headline—and a few hundred more articles have been written since. Reporting was often paired with calls for a national mask mandate, echoing President-elect Biden and now Anthony Fauci, the top U.S. infectious-disease official. The study was also invoked by Francis Collins, head of the National Institutes of Health, to push for stricter masking requirements. Recognizing the potential importance, Nature Medicine rushed the study into print after an expedited peer-review process that took only seven days to complete.

Unfortunately, the IHME modelers’ findings contained an error that even minimal scrutiny should have caught. The projected number of lives saved, and the implied case for a mask mandate, are based on a faulty statistic. Using a months-old survey, IHME modelers assumed erroneously that the U.S. mask-adoption rate stood at only 49% as of late September, and therefore had plenty of room to increase to “universal adoption,” defined as 95%, or to a more plausible 85%. According to more recent survey findings, however, America’s mask-adoption rate has hovered around 80% since the summer.

P.S. Massachusetts cases were up 1.5% again today.