Generating Synthetic but Plausible Healthcare Record Datasets

TL;DR

This paper introduces a new method for generating synthetic healthcare datasets that are more realistic and interpretable than GAN-based methods, avoiding mode collapse and better preserving data diversity.

Contribution

A novel latent variable-based approach for generating binary healthcare data that outperforms GANs in realism and interpretability.

Findings

Synthetic datasets are harder to distinguish from real data using Random Forests and MMD.

The method avoids mode collapse common in GANs.

Generated data are more interpretable than GAN-based models.

Abstract

Generating datasets that "look like" given real ones is an interesting tasks for healthcare applications of ML and many other fields of science and engineering. In this paper we propose a new method of general application to binary datasets based on a method for learning the parameters of a latent variable moment that we have previously used for clustering patient datasets. We compare our method with a recent proposal (MedGan) based on generative adversarial methods and find that the synthetic datasets we generate are globally more realistic in at least two senses: real and synthetic instances are harder to tell apart by Random Forests, and the MMD statistic. The most likely explanation is that our method does not suffer from the "mode collapse" which is an admitted problem of GANs. Additionally, the generative models we generate are easy to interpret, unlike the rather obscure GANs. Our experiments are performed on two patient datasets containing ICD-9 diagnostic codes: the publicly available MIMIC-III dataset and a dataset containing admissions for congestive heart failure during 7 years at Hospital de Sant Pau in Barcelona.