Variation in Low-Value Service Use Across Veterans Affairs Facilities

Background: It is unclear whether extensive variation in the use of low-value services exists even within a national integrated delivery system like the Veterans Health Administration (VA). Objective: To quantify variation in the use of low-value services across VA facilities and examine associations between facility characteristics and low-value service use.DesignIn this retrospective cross-sectional study of VA administrative data, we constructed facility-level rates of low-value service use as the mean count of 29 low-value services per 100 Veterans per year. Adjusted rates were calculated via ordinary least squares regression including covariates for Veteran sociodemographic and clinical characteristics. We quantified the association between adjusted facility-level rates and facility geographic/operational characteristics. Participants: 5,242,301 patients across 139 VA facilities. Main Measures: Use of 29 low-value services within six domains: cancer screening, diagnostic/preventive testing, preoperative testing, imaging, cardiovascular testing and procedures, and surgery. Key Results: The mean rate of low-value service use was 20.0 services per 100 patients per year (S.D. 6.1). Rates ranged from 13.9 at the 10(th) percentile to 27.6 at the 90(th) percentile (90(th)/10(th) percentile ratio 2.0, 95% CI 1.8-2.3). With adjustment for patient covariates, variation across facilities narrowed (S.D. 5.2, 90(th)/10(th) percentile ratio 1.8, 95% CI 1.6-1.9). Only one facility characteristic was positively associated with low-value service use percent of patients seeing non-VA clinicians via VA Community Care, p < 0.05); none was associated with total low-value service use after adjustment for other facility characteristics. There was extensive variation in low-value service use within categories of facility operational characteristics. Conclusions: Despite extensive variation in the use of low-value services across VA facilities, we observed substantial use of these services across facility operational characteristics and at facilities with lower rates of low-value service use. Thus, system-wide interventions to address low-value services may be more effective than interventions targeted to specific facilities or facility types.