A hybrid method of recurrent neural network and graph neural network for next-period prescription prediction
- PDF / 1,868,983 Bytes
- 8 Pages / 595.276 x 790.866 pts Page_size
- 13 Downloads / 230 Views
ORIGINAL ARTICLE
A hybrid method of recurrent neural network and graph neural network for next‑period prescription prediction Sicen Liu1 · Tao Li1 · Haoyang Ding2 · Buzhou Tang1,3 · Xiaolong Wang1 · Qingcai Chen1,3 · Jun Yan2 · Yi Zhou4 Received: 30 December 2019 / Accepted: 10 June 2020 © Springer-Verlag GmbH Germany, part of Springer Nature 2020
Abstract Electronic health records (EHRs) have been widely used to help physicians to make decisions by predicting medical events such as diseases, prescriptions, outcomes, and so on. How to represent patient longitudinal medical data is the key to making these predictions. Recurrent neural network (RNN) is a popular model for patient longitudinal medical data representation from the view of patient status sequences, but it cannot represent complex interactions among different types of medical information, i.e., temporal medical event graphs, which can be represented by graph neural network (GNN). In this paper, we propose a hybrid method of RNN and GNN, called RGNN, for next-period prescription prediction from two views, where RNN is used to represent patient status sequences, and GNN is used to represent temporal medical event graphs. Experiments conducted on the public MIMIC-III ICU data show that the proposed method is effective for next-period prescription prediction, and RNN and GNN are mutually complementary. Keywords Medical prediction · Recurrent neural network · Graph neural network · Next-period prescription prediction
1 Background In the last decade, with the rapid growth of electronic health records (EHRs), data-driven studies, such as drug repurposing and medical event prediction, have become more and more popular in the medical domain. An EHR is composed of a sequence of a patient’s visits in chronological order, each of which contains various medical information, such as demographics, vital signs, diagnoses, medications, procedures, laboratory test results, etc. A medical event prediction task is to predict some types of medical events, including diseases, prescriptions, outcomes, etc., using other types of
medical information or medical history. For example, nextperiod prescription prediction is to predict all medications of a patient in the next time using his/her medical history. The critical challenge of medical event prediction driven by EHRs is how to represent patient longitudinal medical data accurately, also known as patient representation. A large number of methods have been proposed to predict medical events, As medical data of each patient is time series data, the typical time series analysis methods such as machine learning methods based on manually-crafted features [1] and Autoregressive Integrated Moving Average (ARIMA) models applied for medical event prediction [2].
* Buzhou Tang [email protected]
Qingcai Chen [email protected]
* Yi Zhou [email protected]
Jun Yan [email protected]
Sicen Liu [email protected]
1
Tao Li [email protected]
Department of Computer Science, Harbin Institute of Technology
Data Loading...
