G. Decision Support System (DSS)
5. How valid are DSS data?
Researchers are encouraged to use DSS data for healthcare studies because its data has been shown to be valid with respect to several important categories or analyses. For example, encounter-level cost estimates sum to total VA expenditures. In addition, the DSS national data extracts have been shown (for FY2004 data) to link almost perfectly with the VHA discharge and outpatient data sets after adjustments in database design are made. Moreover, annual person level costs according to DSS are similar to HERC annual person-level estimates (see HERC technical report no. 13). There is similar concordance of DSS hospital and outpatient costs (see HERC technical reports nos. 10 and 14).
On the other hand, although DSS data have been audited internally, the database may not be accurate and complete. Researchers should carefully examine the data before using it. Some care may not be recorded as data that can be assigned a cost by DSS. The cost of medical procedures provided to inpatients, including cardiac catheterization, endoscopy, and other procedures, may not be assigned to the patient who had the procedure, but distributed to all patients in proportion to their length of stay. This problem is diminishing as addiction centers improve the way in which they record these procedures.
A second issue is occasional records with outlier cost amounts. If a direct service department has very little or no workload (volume) and costs are not adjusted accordingly, the unit costs for services in that department can be extremely high. For outpatient pharmacy costs, researchers should pay attention to outliers. HERC examined any clinic encounter that cost $100,000 or more as an outlier. In FY2004, there were 47 outliers (86 in FY2003) in the pharmacy records and 121 outliers (123 in FY2003) in the clinic file. Most (37%) of the non-pharmacy outliers were attributed to prosthetics service.
The above issues do not suggest that the DSS data are not good for healthcare studies. In most cases, the issues are concentrated in a number of VA stations. In addition, the quality of DSS data has substantially improved over time and DSS staff are working continuously to improve data quality.