Keywords: Utilization, Costs, Ordinal Logistic Regression,
The purpose of our work is the identification of persistent high-utilizers, while controlling for unhealthy behaviors, perceived health status and other appropriate covariates. High-utilizers are those whose expenditures exceed $10,000 in 2012 Dollars. We link 2006–2010 National Health Interview Survey (NHIS) to 2-year longitudinal Medical Expenditure Panel Survey (MEPS). A partial proportional ordinal logistic regression predicts the utilization group: low-utilizer, one-time high-utilizer, and persistent high-utilizer. One-time high-utilizers are high-utilizers for only one of two years; persistent high-utilizers are high-utilizers for both years, and low-utilizers are the remainder. The number of unhealthy behaviors are based on physical inactivity, smoking, heavy drinking, and inadequate sleeping. Individuals with 2 or more unhealthy behaviors are more likely of being a one-time high-utilizer or persistent high-utilizer as compared to those with no unhealthy behaviors. One-time and persistent high-utilizers do not differ. Age, education, total number of diagnoses, insurance status, and perceived health status have significant effects on being a persistent high-utilizer.