Intervention Recommendations to Minimize the 30-Day Risk-of-Readmission for Heart Failure
In this thesis, we investigate the problem of designing personalized intervention strategies to minimize 30-day readmission risk for heart failure (HF) patients. In particular, we propose a novel framework that recommends personalized intervention to the patients by analyzing the complex interplay among a multitude of factors, such as, demographic factors, medi- cal diagnoses, clinical factors, and how they contribute to readmission risk. Our proposed framework is flexible enough to include or exclude additional factors, as well as layers, or can even obey constraints provided by the domain experts (i.e., doctors) in the design of this hierarchical network. First, we propose to learn the structure and parameters of a hi- erarchical Bayesian network from the available patient data and use that to design rules to recommend personalized interventions. We propose scalable implementation of our proposed solution on Windows Azure, present comprehensive experimental results based on the pro- posed approach to demonstrate the effectiveness of our proposed framework using large scale high dimensional real patient dataset. Our work is validated with State Inpatient Data for Washington State and MultiCare Health System Data. Finally, we introduce a web-based interactive service Pathway-Finder to support decision making. While our primary effort is to design intervention strategies for HF, the proposed framework could be adapted in designing intervention strategies for other diseases as well.