event-icon
Description

Bipartite networks, which simultaneously model both patients and their characteristics, have been useful for identifying significant and comprehensible patient subgroups, a critical step in designing targeted interventions. However, because clustering algorithms require uniform data ranges to identify subgroups, bipartite networks have been used mainly to model uniform data types such as only the binary status on comorbidities, or only the continuous range-normalized values of gene expressions. Here we demonstrate the use of predictive probabilities to transform variables of multiple datatypes into a uniform range, which enabled the use of bipartite networks to identify and comprehend significant subgroups of readmitted patients.

Learning Objective: 1. Learn why bipartite networks have been mostly used to model single data types.
2. Understand how to use predictive probabilities to model multiple data types in bipartite networks.

Authors:

Tianlong Chen, UTMB
Yu-Li Lin, University of Texas Medical Branch
I-Chia Liao, Baylor Scott and White Health
Yong-Fang Kuo, University of Texas Medical Branch
Emmanuel Santillana, UTMB
Clark Andersen, University of Texas Medical Branch
Laurel Copeland, Baylor Scott and White Health
Suresh Bhavnani (Presenter)
UTMB

Presentation Materials:

Tags