The Chronic Disease Score in a VA Population
Anne E. Sales PhD MSN RN
VA Health Services Research and Development
Seattle, WA
Funding Period: January 2000 - September 2001

BACKGROUND/RATIONALE:
This study created and tested a low-cost, practical and clinically relevant pharmacy-based risk adjuster to predict VA health care costs and utilization. The purpose of this study was to create a modified pharmacy-based risk adjustment measure, RxRisk-V, for use in the VA, and to compare its performance to that of other leading risk adjustment methods. RxRisk-V predicts health care costs using a regression equation with the following independent variables: age, sex and the chronic condition classes in which drug fills are observed. Previous versions of RxRisk-V have proven valid and reliable predictors of future health care costs in the commercial sector and appear to perform similarly to more widely used risk adjustment methods.

OBJECTIVE(S):
Our primary research question asked whether the RxRisk is better able to predict costs than three other leading risk adjustment methods: (1) the commercial version of RxRisk, (2) Adjusted Diagnostic Groups (ADGs, a component of Adjusted Clinical Groups), and (3) Hierarchical Coexisting Conditions (HCCs, a refinement of Diagnostic Cost Groups).

METHODS:
This study compared the predictive power of RxRisk-V, a VA-adapted pharmacy-based risk adjustment method, to that of ADGs, HCCs, and the original Group Health Cooperative version of RxRisk (CDS), using the VISN 20 veteran user population. We examined both concurrent and predictive model fit using two sequential 12-month periods (Fiscal Year [FY] 98 and FY 99) with the patient-year as the unit of analysis, using split-half validation. Results are reported here for the validation sample consisting of 80,640 veteran users between FY 96 and FY 98 in VISN 20. Data were extracted from the VISN 20 Data Warehouse, a relational data warehouse mirroring most of the key elements of VistA for all facilities in VISN 20. We constructed an estimation data set of three-year VA users in each of the eight facilities. Data elements include pharmacy (all drugs prescribed and filled), diagnoses from inpatient and outpatient data, and patient demographic variables. We constructed the RxRisk-V using VA drug classes. Cost data were obtained from the Cost Distribution Reports and the Decision Support System in seven of the eight sites.

FINDINGS/RESULTS:
From our results, RxRisk-V does not perform better than the most sophisticated diagnosis-based measure available on the market, the DCG/HCC system. The DCG/HCC system appears to predict costs most accurately both in the current year and prospectively for this VA population. However, as with the ACG/ADG system, DCG/HCC grouping is proprietary, and further development and tailoring for VA are not possible without access to the proprietary code. ADGs appear to predict costs in the current year and prospectively less well than HCCs. RxRisk-V’s predictive power is less than that of the ADGs. A mixed-model that included HCC summary scores and the RxRisk-V categories performed better than either the HCCs or RxRisk-V alone.

IMPACT:
The results have direct applicability to VA administrators who want to use risk adjustment as an aid in allocating resources among regional networks or facilities, setting capitation rates for private contractors, or comparing outcomes across providers.

PUBLICATIONS:

Journal Articles

1. Liu CF, Sales AE, Sharp ND, Fishman P, Sloan KL, Todd-Stenberg J, Nichol WP, Rosen AK, Loveland S. Case-mix adjusting performance measures in a veteran population: pharmacy- and diagnosis-based approaches. Health Services Research. 2003; 38(5): 1319-37.
2. Sales AE, Liu CF, Sloan KL, Malkin J, Fishman PA, Rosen AK, Loveland S, Paul Nichol W, Suzuki NT, Perrin E, Sharp ND, Todd-Stenberg J. Predicting costs of care using a pharmacy-based measure risk adjustment in a veteran population. Medical Care. 2003; 41(6): 753-60.
3. Sloan KL, Sales AE, Liu CF, Fishman P, Nichol P, Suzuki NT, Sharp ND. Construction and characteristics of the RxRisk-V: a VA-adapted pharmacy-based case-mix instrument. Medical Care. 2003; 41(6): 761-74.

DRA: Chronic Diseases, Health Services and Systems
DRE: Technology Development and Assessment
Keywords: Research measure
MeSH Terms: none