Mortality data based on medical records is necessarily incomplete.  STARR Tools supplements the Stanford medical records mortality data with external data made available by the US Social Security Administration (SSA).  This data is also incomplete, but it has in the past been very trustworthy—if a patient matched against the SSA death data, they were almost certainly deceased.   

https://starr.stanford.edu/data-types/death-data  has more detailed information the sources of death data in STARR.

According to the news media outlets (eg . https://www.washingtonpost.com/politics/2025/04/12/trump-immigrants-dead-social-security/) in April 2025, ~6,000 immigrants living in the United States were declared dead in the death master file (DMF). Consequently, mortality data sourced from the SSA DMF can no longer be presumed accurate.  We recommend that this data be used with extreme caution.

How to mitigate this issue:

We are investigating ways to identify these patients and mitigate problems it might cause. One possibility is to identify a valid clinical event (procedure, orders etc) after the death date. This may not be accurate as the patients might not have any clinical events after they were declared dead in last month (Apr 2025).