I did a quick search to see how many children there are in the United State. The first number I found was 74 million (total of all children under 18, as of 2013).
5% of 74 million is 3.7 million. Since I doubt they are giving Ritalin to toddlers (yet), this estimated number of children with ADHD is amazing close to the number who are on a prescription.
In other words, those are probably not two independently-derived numbers. They're one. There is no independent estimate of what percentage of kids have ADHD: there's only a count of how many are on the meds. This is a classic trick from _How to Lie with Statistics_: when you don't have the number you want (how many kids actually have ADHD), use the number you have (how many are on the meds) and pretend there is no distinction.
This has the side effect of "showing" (with numbers!) that the diagnostic methods for ADHD are nearly perfect. By circular logic, QED.