I am not enthusiastic about using influenza like illness (ILI) data to guide the public health response to Covid. I think @reichlab and I were actually the first team to use ILI data to look for signs of Covid in the US, so I do see the potential. But.. 1/

We ran our analysis weekly from late Jan until Apr, but a few weeks ago we stopped bc it got too difficult to interpret. Some clinics stopped reporting their results. And in some states, not many clinics participate in the program anyways so the numbers bounce around a lot. 2/

Plus, as Covid worsened, people stopped going to the doctor for regular visits, so the denominator changed. But the numerator was changing too, as real influenza season ended and Covid took off. So with both pieces in flux, it became too hard to make sense of. 3/

And I find it hard to make decisions off of syndromic data generally. It’s not a direct look at an outbreak, it’s more like a helpful hint. But it’s much harder to pull the alarm when the % ILI goes from 4 to 5% than it is for 20 new, diagnosed Covid cases, for example. 4/

Syndromic data can be v helpful and we should incorporate it and even expand our existing systems. But it shouldn’t replace diagnostic testing. We should ensure that all 3 gating criteria are met before leaving Phase I and not rely on ILI alone. 5/