Daw CM (Houston Center for Quality of Care and Utilization Studies, Michael E. DeBakey VA Medical Center), Yu HJ
(Houston Center for Quality of Care and Utilization Studies, Michael E. DeBakey VA Medical Center), Kuebeler M
(Houston Center for Quality of Care and Utilization Studies, Michael E. DeBakey VA Medical Center), Alsarraj A
(Houston Center for Quality of Care and Utilization Studies, Michael E. DeBakey VA Medical Center), Petersen LA
(Houston Center for Quality of Care and Utilization Studies, Michael E. DeBakey VA Medical Center, and Baylor College of Medicine)
Assessment of DSS cost data is ongoing among VA researchers. We explored variation in patient-level Decision Support System National Data Extracts (DSS NDE) data and impact of this variation on a research question.
: Using FY05 DSS NDE records, we calculated facility-level median, mean, standard deviation (SD), and range of per-visit DSS cost by facility for 65 outpatient clinic stops. We examined correlation of visit volume and facility median per-visit cost. We compared two cohorts of primarily-outpatient-pharmacy (POP) users (non-pharmacy outpatient costs < $800; pharmacy costs > $100 and at least 33% of annual costs; < 4 outpatients visits)--one using DSS NDE and another using cost data from the VA Health Economics Resource Center (HERC) Average Cost Datasets (ACD).
We observed dramatic variation in the DSS NDE per-visit cost within and across facilities. The widest ranges in facility-level median per-visit cost were observed with Telephone Triage (median of facility-level medians was $52; lowest facility-level median was $-316, highest was $59,428); Cardiac Catheterization (median: $2,668; range of medians: $146-$19,260); and Thoracic Surgery (median: $369; range of medians: $5-$7,166). The ratio of SD/Mean of facility-level medians was > 1 for 28 of the 65 clinic stops. Outliers were not clustered at any particular site. Facility-level per-visit cost was somewhat, but not highly, correlated with number of visits: correlation coefficients ranged from -0.4 to 0.15. In the POP-user cohort comparison, the HERC-based cohort numbered 777,756 patients and the DSS-based cohort totaled 708,060; 648,406 met cohort criteria using both cost methods. Among patients who qualified using HERC ACD but not with DSS NDE (n=117,272), mean DSS cost was $1,245, (SD $2,565), range $0-$241,843. Among patients qualifying using DSS but not HERC cost (n=58,647), mean HERC cost was $795 (SD $466), range $0-$21,451.
There are a notable number of outliers and wide variability in DSS costs. The POP-user cohort comparison illustrated that the composition of a subgroup is affected by whether one uses DSS NDE or HERC ACD data.
Variation in DSS NDE data may impact health services research. For patient-level research involving specific types of services, researchers may wish to explore other sources of data such as the HERC ADC.