Identifying Risk Factors and Cost Anomalies in Healthcare Spending Using Medicare Claims Data

Abstract

The United States spends more on healthcare per capita and relative to GDP than any other country in the world. Identifying the drivers of costs in healthcare is of great interest to the key stakeholders in the system: policymakers, insurers, providers, and patients. The objective of this work is to develop a methodological framework for identifying provider-level and individual-level factors that contribute to systematically higher costs through machine learning techniques. We propose identification methods for two important types of cost anomalies, super-utilizing patients and low-value areas in the healthcare system, as well as the factors that contribute to these anomalies. Additionally, we conduct a correlational analysis to understand the relationships between health outcomes. To do so, we address the methodological challenges of working with high-dimensional and sparse administrative medical claims data.