Comparative effectiveness for oral anti-diabetic treatments among newly diagnosed type 2 diabetics: data-driven predictive analytics in healthcare

Abstract

A difficult problem in healthcare is predicting who will become very sick in the near future. In our case, we find that the top 10% of newly diagnosed type 2 diabetes patients account for 68% of healthcare utilization. In this paper, we demonstrate how the U.S. healthcare system can provide improved healthcare quality per unit of spend through better predictive data-based analytics applied to the increasingly available troves of healthcare claims data. Specifically, we demonstrate the effectiveness of data mining by applying machine learning methods to large-scale medical and pharmacy claims data for over 65,000 patients newly diagnosed with type 2 diabetes, a common and costly disease globally. This analysis reveals some important heretofore unknown patterns in the cost and quality among of the disease's common treatments and demonstrates the potential for using large-scale data mining for efficiently focusing further inquiry.