Breaking Down Healthcare: Applications of Clustering and Markov Chains for Medical Claims Data

Abstract

In the United States, medical data has recently begun to adopt a much larger role in healthcare. Made available through digital platforms, these files are spurring new developments for healthcare analytics, especially in terms of efficacy and resource management. One of the emerging movements in this field has been predictive pathway modeling, which maps the potential trajectories of patients. In order to conduct such analyses, we apply high-dimensional modeling techniques to medical claims data, focusing primarily on spectral clustering techniques and Markov modeling. The findings offer insight into the relationships between patient outcomes and treatment patterns. We also highlight key attributes related to high costs and overall quality of care.