CF Liu, VA Puget Sound Health Care System and University of Washington; KL Sloan, VA Puget Sound Health Care System and University of Washington; AE Sales, VA Puget Sound Health Care System and University of Washington; J Todd-Stenberg, VA Puget Sound Health Care System; ND Sharp, VA Puget Sound Health Care System; P Fishman, Group Health Cooperative of Puget Sound; AK Rosen, Bedford VAMC and Boston University; S Loveland, Bedford VAMC

Objectives: RxRisk, formerly the Chronic Disease Score, is a pharmacy-based case-mix measure for profiling and predicting health care resource utilization. Pharmacy data provide a consistently coded, direct measure of actively treated comorbidity and disease severity. The RxRisk, a non-proprietary methodology developed in a commercial population, uses outpatient pharmacy data to group patients into 29 disease conditions. VA is a good environment for using a pharmacy-based case-mix adjustment method, because VA enrollees have financial incentives to obtain all their medications from VA and pharmacy data are available in the VA national administrative databases. This study explored how the RxRisk method characterizes the disease burden of the VA population and how it predicts outpatient utilization compared to a demographics-only model (age and gender) and the Diagnostic Cost Groups/Hierarchical Co-existing Conditions (DCG/HCC) model, one of the most widely used diagnosis-based case-mix adjusters.

Methods: The study consisted of veteran users from inpatient and outpatient facilities in VISN 20 during Fiscal Year (FY) 1998 (n=122,840). Outpatient pharmacy information, ICD-9-CM diagnoses, and utilization data at the patient level were obtained from the VISN 20 data warehouse. Number of provider-related face-to-face ambulatory encounters was used to measure outpatient utilization. Prospective risk adjustment models were fitted, using FY 1998 diagnoses, outpatient prescriptions, and demographics (age and gender) to predict FY 1999 outpatient utilization. Age and gender were included in the RxRisk and the DCG/HCC models.

Results: The RxRisk method assigned 31% of users into at least one of 29 disease categories. The 10 most common categories were pain/inflammation (27.7% of users), hypertension (22.9%), gastric acid disorder (21.9%), heart disease/hypertension (22.2%), depression (19.6%), pain (18.1%), cardiac disease (16.3%), hyperlipidemia (13.4%), asthma (12.6%), and diabetes (10.1%). In the RxRisk model, 25 out of 29 disease categories significantly increased outpatient provider encounters in the following year. The disease categories with the most impact on outpatient utilization included psychotic illness (12.1 visits), renal disease (5.8 visits), bipolar disorder (5.1 visits), depression (4.2 visits), epilepsy (4.0 visits), HIV (3.2 visits), anxiety/tension (3.0 visits), and malignancies (2.8 visits). The R-square of the RxRisk model in predicting the outpatient provider encounters was 0.14 compared with 0.007 for the demographic model and 0.17 for the DCG/HCC model.

Conclusions: This study shows that the RxRisk provides an alternative case-mix measure for the VA. As we expected, the RxRisk model performed much better than the age/gender model. The classification of the RxRisk is more parsimonious than that of the DCG/HCC model (29 versus 118 disease categories); however, the R-squares were comparable.

Impact: The results of this study show that RxRisk can provide a low-cost, practical, and clinically relevant pharmacy-based case-mix measure to predict VA health care utilization.