Extracting Diagnosis Pathways from Electronic Health Records Using Deep Reinforcement Learning

TL;DR

This paper explores using deep reinforcement learning to derive diagnostic pathways from electronic health records, demonstrating competitive accuracy and interpretability in diagnosing anemia and its subtypes.

Contribution

It introduces a novel application of deep reinforcement learning for sequential diagnosis pathway extraction, emphasizing robustness to noise and missing data.

Findings

Deep reinforcement learning achieves competitive diagnostic accuracy.

The approach provides interpretable diagnostic pathways.

Robustness to noise and missing data is demonstrated.

Abstract

Clinical diagnosis guidelines aim at specifying the steps that may lead to a diagnosis. Inspired by guidelines, we aim to learn the optimal sequence of actions to perform in order to obtain a correct diagnosis from electronic health records. We apply various deep reinforcement learning algorithms to this task and experiment on a synthetic but realistic dataset to differentially diagnose anemia and its subtypes and particularly evaluate the robustness of various approaches to noise and missing data. Experimental results show that the deep reinforcement learning algorithms show competitive performance compared to the state-of-the-art methods with the added advantage that they enable the progressive generation of a pathway to the suggested diagnosis, which can both guide and explain the decision process.