Survival Data Mining: Treatment of Chronic Illness

Overview

Patients with chronic illness often have many medications available for treatment. Medications are changed, and added when the patient experiences side effects or the disease progresses so that a medication is no longer effective. It is the purpose of this project to use survival data mining techniques to investigate prescribed medications for patients with Type II diabetes. The paper examines the Medical Expenditure Panel Survey for the years 2000-2004 and validates the results using a second database for a cohort of patients with diabetes. The paper preprocesses the data by filtering to medications listed for the ICD9 code for diabetes. Then the author removes medications not specifically related to glucose control, eliminating misspellings and name variations.