Internship project: Evaluating quality of synthetic data for training a deterioration prediction AI

Opportunity

Electronic medical records (EMR) could support digital health applications that improve patient outcomes, if innovators can access the data to develop algorithms. Synthetic EMR data could improve privacy and stand in for real data when building proof of concepts, assuming similar performance can be achieved. During this project the intern will replicate a published machine learning model for predicting patient deterioration. They will then benchmark model accuracy when trained on synthetic versus real data. If successful, this will project will lay the groundwork to scale teaching and innovation using synthetic health data.

Project Objectives

• Replicate a published machine learning model for predicting patient deterioration.
• Benchmark the replicated machine learning model against synthetic and real clinical data.
• Develop recommendations for applying a machine learning model to a synthetic data set.