Data-driven Automatic Treatment Regimen Development and Recommendation

• DOI: 10.1145/2939672.2939866
• Publication Date: 2016-08-08T18:33:46Z
• Authors: Leilei Sun; Chuanren Liu; Chonghui Guo; Hui Xiong; Yanming Xie
• Keywords: Regimen; Medical record; Code (set theory)

The analysis of large-scale Electrical Medical Records (EMRs) has the potential to develop and optimize clinical treatment regimens. A regimen usually includes a series doctor orders containing rich temporal heterogeneous information. However, in many existing studies, order is simplified as an event code record sequence. Thus, information inherent not fully used for in-depth analysis. In this paper, we aim at exploiting developing data-driven approaches improving treatments. To end, first propose novel method measure similarities between records with consideration sequential multifaceted orders. Then, efficient density-based clustering algorithm summarize records, extract semantic representation each cluster. Finally, unified framework evaluate discovered regimens, find most effective new patients. empirical study, validate our methods EMRs 27,678 patients from 14 hospitals. results show that: 1) Our can successfully typical regimens records. extracted are intuitive provide managerial implications design optimization. 2) By recommending total cure rate data improves 19.89% 21.28%, increases up 98.29%.