Ok... but how does that figure into testing sick people? Don't symptoms determine who gets tested?
Absolutely. Unless you test everyone in a population, you have a selection bias to some degree.
For example, we don't do heart caths on 25 y/o healthy folks. Of course, in a world of perfect medical access, we could do non-invasive 100% accurate testing at a low cost on 100% of the population for everything.
Let's put it this way - if we did colonoscopies on every adult American every year, our incidence of colon cancer would be nearly zero - because colonoscopies can prevent colon cancer. But we only screen people over 50 (45 now in some instances) or those who have a family history or have some kind of symptoms that make us suspicious. So we have a selection bias.
Testing everyone in a given population eliminates that and gives you 'ideal' numbers - but it's usually at a horrendous expense and effort to do so. You have to use your resources in the smartest way possible. Therefore, when you're dealing with testing large numbers of people in different geographic areas there are lots of variables that will skew the numbers one way or another and you're not truly comparing apples to apples. Or at best, maybe Granny Smiths to Red Delicious to Fujis.