2008 HSR&D National Meeting Abstract
1077 — Comparing Outpatient Cost Data in the DSS National Pharmacy Extract and the PBM Database
Smith MW (Health Economics Resource Center (HERC)), King SS
The Decision Support System (DSS) and the Pharmacy Benefits Management (PBM) Strategic Healthcare Group both offer prescription-level outpatient pharmacy databases. They cover the same prescriptions but feature different cost variables, leaving researchers unsure of which to choose. We investigated the match between PBM total cost and DSS variable supply cost, the two variables closest in definition. We expected to find similar costs for prescriptions filled through outpatient mail-order pharmacies (CMOPs) but a poorer match for pharmacy window fills. We also explored how observed cost differences vary by drug type and VA station.
We located all FY2004 outpatient pharmacy records for 894 enrollees in a multi-site trial. We grouped records into unique combinations by patient ID, drug name, and quantity dispensed. We then determined the average PBM total cost and the average DSS variable supply cost for each combination. We developed descriptive cost statistics by dispensing location (pharmacy window vs. CMOP) for each dataset separately and for the difference across datasets. ANOVAs provided a broad view of the relation between cost and two variables, dispensing location and station. We then used regression analysis to estimate coefficients and significance values for a subset of those variables.
DSS variable supply cost and PBM total cost variables had very similar means but exhibited substantial variation across combinations within each dataset and across datasets for the same combination. Differences were partly correlated with dispensing location, VA station, and drug type, although considerable variation remained after controlling for these factors. Contrary to expectation, PBM total cost exceeded DSS variable supply cost on average. Finally, facility and drug group were sometimes significant predictors of the cost difference across datasets.
DSS variable supply cost and PBM total cost were very similar. If confirmed in other samples, this result implies that DSS pharmacy data would be sufficient for many cost studies.
PBM data contain many unique prescription-related variables, but DSS data are significantly easier to access than PBM data and are easily linked to other utilization data. Our findings may increase researchers’ confidence in analyzing pharmacy cost using DSS data alone.