This year's seasonal flu forecast predicted a moderate season, based largely on the previous three years of data. The actual season arrived earlier and peaked higher than the model anticipated, catching several regional health systems short-staffed for the surge.
Researchers reviewing the miss found the model had weighted recent, unusually mild seasons too heavily, treating a run of below-average years as the new baseline rather than as statistical variation around a longer-term average.
Where the forecast broke down
The fix isn't dramatic: researchers are widening the historical window the model draws from, so a handful of mild years can't skew the baseline as strongly as they did this time.
None of this guarantees next year's forecast will be accurate — flu seasons are inherently variable, and every model carries irreducible uncertainty. But researchers say the specific failure mode this year was identifiable, and correctable, which is more than can be said for every miss.