Online Program

County-Level Determinants of Prescription Drug Consumption in Selected U.S. States

Thierry Nianogo, The University of Memphis 
*Demba Fofana, The University of Memphis 
Albert Okunade, The University of Memphis 

Keywords: Prescription drugs, Box-Cox transformation model, US county level data

Prescription drugs expenditure, currently accounting for 10% of total US healthcare spending, is the third largest and a rapidly growing component of healthcare costs. The 2006 Medicare Part D drug benefits and the 2010 Affordable Care Act are catalysts for further increases in drug spending due to greater insurance coverage. Consequently, this research investigating the drivers of prescription drug consumption is important and timely for its cost containment policy implications. Significant geographic variations in population health status, access to care, socio-economics and demographics, which are core drivers of health care costs; and the cross-state disparities convergence in pharmaceutical expenditures motivate the need to construct and estimate separate econometric models for selected low (ID, SD, and WA), average (AR, ND) and high (TN) prescription drug consumption US states. Since healthcare data tend to be skewed, we fitted the variance stabilizing Box-Cox power family of transformations model to 2010 county-level observations to investigate the drivers of prescription drug consumption separately in low and high drug consumption regions. This innovative study is the first to separately model drug consumption, using the most recent county level data, in these disparate regions. Our study reveals several interesting findings. First, the optimal model ?-power transformation parameter estimates for the dependent variable in high (?=0.568) Average (?=0.696), and high (?=0) spending regions differ significantly. Second, the income elasticity estimates also differ in high (0.536) and low (0.481) spending regions. Third, contrasting the many earlier studies modeling drug expenditures (rather than number of filled prescriptions used in our study) we detect prescription drugs to be a normal good and a technical necessity (the income elasticity for the pooled data model is 0.461) which generally accords with a priori theory. Fourth, while the numerical estimates of the extent to which similar factors drive prescription drug consumption differs across regions, access to primary care physicians in the high consumption region is highly statistically significant. Policy implications are explored.