Table 5:

Summary of lessons learned

•	Lack of clear regulatory guidance for de-identification of decedent records for research use (in contrast to the Safe Harbor Method for general de-identification within a whole population IDR)
•	Need for metadata on document structure (empty EHR forms; pre-populated fields, secondary person tokens as part of the form question; plain free-text vs. structured provider e-signatures)
•	Phone number as the most sensitive secondary ID and possible disambiguation methods for phone numbers (local hotel or pharmacy vs. secondary person private number)
•	Provision of prior information to scrubber about local provider names or permitted alphanumeric IDs, such as experimental drug names.
•	Importance of accurate deceased status using local institutional data or external data from State Vital Statistics Administration (research context only) or federal death index data