It's from the CDC. There is a ILI surveillance network where they are getting data from hospitals. You can read all about it here:
For your second point, that could be part of it for sure. There's a few ideas I have to try and estimate that impact, but I don't know when/if I'll get around to it. However, there's no reason to think that that same panic wouldn't result in people taking their children as well. That's why I like looking at the ratio. If you assume that the historical ratio is more or less constant, then you've adjusted for the impact of different overall levels of flu. So a really bad flu year and a really light flu year actually end up remarkably similar historically. For the most recent week we're typically within +/- 7% of 35%. It's amazing how stable this actually is year to year.