Title: TPGraph: a hospital readmission prediction method based on temporal phenotype graphs

Authors: Lizhen Cui; Xiangzhen Xu; Shijun Liu; Hui Li; Zhiqi Liu

Addresses: School of Software Engineering, Shandong University, Jinan 250101, Shandong, China ' School of Software Engineering, Shandong University, Jinan 250101, Shandong, China ' School of Computer Science and Technology, Shandong University, Jinan 250101, Shandong, China ' School of Computer Science and Technology, Shandong University, Jinan 250101, Shandong, China ' School of Computer Science and Technology, Shandong University, Jinan 250101, Shandong, China

Abstract: Accurate hospital readmission prediction in a vast amount of healthcare data is important to the reducing healthcare costs and improving treatment patterns. Due to the temporality and sequentiality of the medical records, we propose a method for predicting hospital readmission based on temporal phenotype graphs in this paper, namely the TPGraph. Firstly, we constructed a temporal graph for each patient based on their medical event sequence. Then, we developed an approach to identify the most significant frequent subgraphs as temporal phenotype graphs. After that, an improved greedy algorithm was designed to obtain the optimal expression coefficient of temporal phenotype graphs. Finally, the optimal expression coefficient as a feature, we use random forest algorithm to predict whether the patient will perform hospital readmission. Our experiments demonstrate the effectiveness of our proposed method, and show that our approach to gain better predictive performance compared with the baselines.

Keywords: healthcare; temporal phenotype; TPGraph; temporal phenotype graphs; hospital readmission prediction; frequent subgraph mining; optimal xpression coefficient; temporal graph; medical event sequence; AGM; coronary heart disease.

DOI: 10.1504/IJDMB.2018.094782

International Journal of Data Mining and Bioinformatics, 2018 Vol.20 No.3, pp.247 - 266

Available online: 13 Sep 2018 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article