Volume 51 | Number 5 | October 2016

Abstract List

Todd H. Wagner Ph.D., Anjali Upadhyay M.S., Elizabeth Cowgill M.P.H., Theodore Stefos Ph.D., Eileen Moran M.H.A., Steven M. Asch M.D., M.P.H., Peter Almenoff M.D.


To compare risk scores computed by Dx (Verisk) and Centers for Medicare and Medicaid Services () V21.

Research Design

Analysis of administrative data from the Department of Veterans Affairs () for fiscal years 2010 and 2011.

Study Design

We regressed total annual costs on predicted risk scores. Model fit was judged by ‐squared, root mean squared error, mean absolute error, and Hosmer–Lemeshow goodness‐of‐fit tests. Recalibrated models were tested using split samples with pharmacy data.

Data Collection

We created six analytical files: a random sample ( = 2 million), high cost users ( = 261,487), users over age 75 ( = 644,524), mental health and substance use users ( = 830,832), multimorbid users ( = 817,951), and low‐risk users ( = 78,032).

Principal Findings

The Dx Medicaid with pharmacy risk score yielded substantial gains in fit over the V21 model. Recalibrating the V21 model using pharmacy data‐generated risk scores with similar fit statistics to the Dx risk scores.


Although the V21 and Dx prospective risk scores were similar, the Dx model with pharmacy data offered improved fit over V21. However, health care systems, such as the , can recalibrate the V21 model with additional variables to develop a tailored risk score that compares favorably to the Dx models.