Augmenting Date of Death & Cause of Death Ascertainment in Sentinel

Administrative healthcare claims and electronic health records (EHRs) are limited in their capture of information on the fact and cause of death, particularly when a death occurs outside of a hospitalization. This project will develop a set of open-source algorithms to augment the assessment of mortality in EHR data through probabilistic linkages to additional mortality data sources. Outpatient EHR encounter data at two healthcare systems will be augmented with additional sources of mortality data including publicly available obituary and social media data, as well as national and state death certificate registries, to enhance capture of date of death. Machine learning will then be applied to develop probabilistic estimates of cause of death. Development of these and other generalizable approaches to improve death ascertainment can improve the validity of Sentinel activities using mortality as an endpoint. The Innovation Center (IC) will collaborate with experts at Vanderbilt University Medical Center (VUMC), Mass General Brigham (MGB), and Research Triangle Institute (RTI) International.