Data privacy concerns and laws surrounding patient confidentiality continue to pose a large barrier to researchers accessing electronic health records. The process of deidentifying data is often tedious and costly, and can distort important features from the original dataset. More recent attempts at overcoming the issues of sharing data have revolved around the generation of synthetic data. Prior work has used approaches such as sampling from distributions, random forests, generative adversarial networks, and Bayesian networks. However, many of these methods have either not been applied to the area of healthcare data or have failed to capture its complexity and dimensionality. This project seeks to investigate the use of Bayesian networks, which have been previously used to create synthetic data based on census datasets, to simulate electronic health care records that closely mimic the features of the original data.

“SuperUROP offers me the chance to do advanced research and connect with those who will help me make that research more impactful. My background in interning at the Fred Hutch, SMART, and Philips, as well as participating in UROP in CSAIL, have equipped me with the experience to make the most of this opportunity. As a double major in Courses 6-7 and 18, I am excited to apply my skills in computer science and mathematics toward my interest in the medical field!”