A study led by Johns Hopkins describes a method for automating slow, inaccurate manual chart review when searching for signs of misdiagnosis.
