That is another thing that can be overlooked by some but unfortunately we can't know the exact number. However, human error is always possible and it's a safe assumption that yes some patients will not follow thru.
20% of patient data becoming invalid would not surprise me, especially given the strict parameters as you said, but hoping it's less! Either way I entrust the clinicians to accept only the most good and valid criteria.
News 
Market Data 
Discover 
