It sounds like all that's going on is the CDC is using good methodology when conducting their survey. First they target a population they want information about (children in a certain age range in a certain area of the country for example). Then they randomly select a subset of this population to actually talk to. They can't talk to all x number of people in their target population (probably their target population is too large), so they select a representative sample - usually a computer program would randomly select the subsample from the larger sample. Hence, you have been "randomly selected".
Now, once they have their subsample selected their job is to get their data from as many of these people as possible, without skipping over any significant portion of their subset. If this stage of the survey is done wrong, then something called response bias creeps into the data. It works like this - they call up a series of parents, and the parents of kids who are vaccinated cheerfully answer their questions, but the parents of unvaccinated kids get squirrely and refuse to answer any questions. The CDC then gets data only from parents of vaccinated kids. The data would then reflect this, perhaps making it look like more people choose to vaccinate their kids than actually do, or perhaps changing the nature of the data they are getting concerning attitudes about vaccines (these are examples, I have no idea what the survey is actually about).
My intro stats professor told the class that if he got a survey in the mail, he would throw it out and see if they send the survey again. If they sent the survey again, he would throw it out. If they sent the survey a third time, he would consider participating because he saw their persistence in trying to eliminate response bias as an indication the survey was being done correctly and would be worth participating in.