This is the codebase for the paper titled "Fairness-Optimized Synthetic Electronic Health Record Generation Pipeline for Arbitrary Downstream Predictive Tasks". The following code is for the MIMIC-III experiments found in the paper. You can create and use similar code for the PIC dataset as well.
There are two main notebooks: one for the MIMIC-III dataset and another for the PIC dataset. All the instructions are given in the notebooks. Additionally, you will need to apply, complete training, and download the MIMIC-III requisite files from https://physionet.org.