Dates of patient death and associated metadata are a known issue for electronic health records, as deaths that occur outside the care system tend to be missing. One approach to this problem is to supplement EHR data by linking to external death records, but this comes with a significant challenge: linking records and managing identity resolution. In this seminar, the TraCS Data Science Lab will review the state death data linkage work that has existed at UNC for many years and introduce the addition of a machine learning classification to optimize match resolution.
Speaker:
JP Powers, PhD
Research Data Scientist
NC TraCS Institute
The NC TraCS Data Science Seminar Series will be held on the third Tuesday of each month. These 1-hour sessions will cover a range of topics broadly applicable to healthcare data science. While some sessions will focus on organizational aspects of starting or getting involved in data science and AI healthcare research at UNC, other sessions will focus on technical aspects of data architecture and modeling, programming, or application of machine learning methods.