ya this is why stats are easy to manipulate.
the JDRF states 14,000 new cases T1D, in Children, diagnosed per year.
Childstats/childrensdatabank, etc has the approximate number of children in the US at 75MM.
so, if you want to know the incedence of undiagnosed children with T1D in the US, right now, it's over 1 in 5000. (and that's a totally different number than the number of children who know they have diabetes in the US, right now)
That isn't including adults, or anyone who "knows" they have diabetes, or anyone who got diagnosed last year, or their 4 brothers, etc. So if you had to guess how many undiagnosed children are in a school with 4,600 children attending, the answer is probably zero.
The 1 in 333 is the prevalence of scoliosis, or how many people suffer. It's wrong to compare the incedence with the prevalence because they are not the same. JAMA [vol 282, no 15, 10/1999] estimated the incedence of scoliosis at 1 in 448 for in-school testing. JAMA does a real good job talking about the effectiveness for in-school screening. So then for scoliosis, the best guess (for the same made-up school) is probably 10 kids who need treatment...
Anyway, it's not so easy to screen for diabetes. a kid with a juce box can have a random sugar of 180, in some cases higher, and it could mean nothing. Screening is a good idea for a lot of things and I am a proponent of preventive medicine - I guess it just doesn't add up in this case, imo.