Online Program

Components of Cost Variation in a Veterans Pain Population

*Evan Paul Carey, University of Colorado School of Public Health 
Gary K Grunwald, PhD, University of Colorado School of Public Health 
Colin I O'Donnell, MS, University of Colorado School of Public Health 
Michael Ho, MD, PhD, VA Eastern Colorado Health Care System 
Robert Kerns, Yale 

Keywords: cost, pain, VA, zero-inflated gamma, hurdle

The annual cost of chronic pain in the United States is estimated to exceed $500 billion. Chronic pain disproportionately affects combat veterans. Understanding the variation in pain-related costs over time and hospitals is critical for the Veterans Health Administration, the largest integrated health care system in the U.S. We constructed a data set of monthly costs for 673,676 veterans in primary care reporting pain from June 2010 through 2014. We split costs into outpatient, inpatient, primary care, and emergency department cost buckets. We separately modeled each monthly cost bucket at the patient level using logistic regression for monthly cost occurrence and a log-gamma regression for the distribution of non-zero costs. We used natural splines for calendar time and months since baseline pain to explore nonlinear time trends. We used random hospital effects to explore facility-level variation in cost trends. Using this approach, we dissected overall costs into individual cost buckets, system calendar trends, and patient-level disease time trends. Additionally, health care use is modeled separately from prices. We discuss the policy implications of these findings.