This tool deids HL7 messages for troubleshooting purposes.
Don't lose any information beyond what is required by law/common sense.
Let's say a batch contains 100 ELR and exactly 32 of them are about the same person while the rest are about different people. Naive de-id tools would 'lose' that information--the fact that there were 32 messages about the same person. The information would be lost when every name in the file gets set to JOHN DOE. It is possible to preserve 'number of occurances' information without violating anyone's privacy.
Append something like "[the string] must be replaced such that this information is preserved" to each item in the following list, and you've got a list of requirements for this software.
- recurrance: if a string appears in more than one place in the file
- string length
- whitespace
- "non-ascii" Hm, not certain about this one. Could this be used to identify a patient? If we're leaving zipcodes in there, sometimes it could be!