Customize your JAMA Network experience by selecting one or more topics from the list below.
Gellad W, Mor M, Zhao X, Donohue J, Good C. Variation in Use of High-Cost Diabetes Mellitus Medications in the VA Healthcare System. Arch Intern Med. 2012;172(20):1608–1611. doi:10.1001/archinternmed.2012.4482
Author Affiliations: Center for Health Equity Research and Promotion, VA Pittsburgh Healthcare System, Pittsburgh, Pennsylvania (Drs Gellad, Mor, Zhao, and Good); Department of Medicine, Division of General Medicine (Drs Gellad and Good), and Departments of Biostatistics (Dr Mor) and Health Policy and Management, Graduate School of Public Health (Dr Donohue), University of Pittsburgh, Pittsburgh; RAND Corp, Pittsburgh (Dr Gellad); and Pharmacy Benefits Management Services, US Department of Veterans Affairs, Hines, Illinois (Dr Good)
The Department of Veterans Affairs (VA), the largest integrated health care system in the United States, may serve as a model of efficient use of prescription drugs. It consistently ranks among the top of all US health care systems in objective ratings of quality of care for chronic diseases,1 and it does so with low medication costs. The VA negotiates steep price discounts with pharmaceutical manufacturers and engages in robust formulary management using a national formulary. This centralized approach to pharmacy benefit management stands in stark contrast to Medicare Part D, which contracts with over 1000 private plans, each with its own formulary, and which has substantial regional variation in per capita drug spending.2 Even within a tightly managed system such as the VA, however, there may also be significant variation across facilities in medication use and spending. We examined national VA data for over 1 million outpatients with diabetes mellitus (DM) to understand how prescribing of high-cost medications varies across facilities.
We identified a national cohort of veterans with type 2 DM using a previously validated approach.3 We focused on the facility-level use, in fiscal year (FY) 2009, of 2 classes of high-cost DM medications: thiazolidinediones (rosiglitazone, pioglitazone) and long-acting insulin analogues (detemir, glargine). We measured the proportion of patients on oral medications receiving thiazolidinediones, and the proportion of patients treated with insulin receiving long-acting analogues. We chose these 2 classes because of their relatively high cost and lack of clear evidence for improved clinical outcomes relative to other DM medications.4-7 Thiazolidinediones were available for use with prior authorization at the time of the study. There were no restrictions on long-acting analogues.
To calculate an adjusted rate of thiazolidinedione and analogue use at each facility, adjusting for differences in patient-level characteristics, we used random effects logistic regression. We adjusted for age, race/ethnicity, sex, modified Charlson score (removing DM and DM complications),8 number of diabetic complications (diabetic retinopathy, neuropathy, nephropathy, or peripheral vascular disease), whether individuals had any visits to an endocrine/DM specialty clinic, and whether they had a medication copay. We centered each of these covariates and used the facility random intercept to compute an adjusted rate of thiazolidinedione/analogue use for patients at that facility with all covariates equal to the population mean. All analyses were performed using SAS statistical software (version 9.2; SAS Institute Inc) and STATA 11 software (StataCorp Inc).
In FY 2009, there were 1 158 809 patients with type 2 DM. Their mean age was 66.5 years, and 97.4% were male. Almost 1 in 7 (13.8%) visited an endocrine/DM clinic, and 30.8% had at least 1 DM complication. Overall, 906 720 patients (78.3%) received 6 182 859 prescriptions for DM medications; 66.7% received an oral medication, and 27.7% received insulin (16.1% received both).
Across the 139 facilities, the adjusted percentage of patients receiving oral DM medications who used a thiazolidinedione ranged from 1.4% at the lowest-using facility to 25.4% at the highest, with a median of 8.2% (interquartile range [IQR], 4.9%-10.5%) (Table). The adjusted percentage of patients receiving insulin who used long-acting analogues ranged from 4.0% to 71.2%, with a median percentage of 40.6% (IQR, 28.1%-52.1%). The adjusted facility-level rates of use were almost identical with the unadjusted rates across facilities (correlation r = 0.99).
In this cohort of over 1 million patients with type 2 DM, we find substantial variation in use of 2 classes of high-cost DM medications—thiazolidinediones and long-acting insulin analogues. This variation exists in an integrated VA system with a uniform national formulary with established criteria for use of drugs, such as the thiazolidinediones, and clinical practice guidelines supporting conservative use of medications. While some variation is expected given reasonable differences in prescribing practices, the observed 18-fold variation across facilities was unexpected.
Adjusting for observable patient characteristics across facilities explained virtually none of the facility-level variation in use of high-cost drugs, suggesting that there are important facility factors at play. Even if some unmeasured patient characteristics are driving some of the variation, the magnitude of the variation is large enough that clinical need alone cannot explain it. Despite the national formulary, decisions about approving requests for high-cost medications are not made centrally, but are often made at each VA facility. In addition to such differences in administration, facility-level differences are likely also driven by local physician norms or preferences about the use of newer drugs, which we were not able to measure.
Our findings suggest that while they may exert powerful effects on medication choice,9 formularies and utilization management tools can only go so far in standardizing health care delivery. While use of more costly agents, such as the thiazolidinediones and long-acting insulin analogues, clearly have a place in the care of patients with DM, more work is needed to understand the mechanisms underlying this variation so that health systems can optimize their use to promote safe, high-value pharmaceutical practice.
Correspondence: Dr Gellad, Center for Health Equity Research and Promotion, VA Pittsburgh Healthcare System, 7180 Highland Dr, Pittsburgh, PA 15206 (email@example.com).
Published Online: October 8, 2012. doi:10.1001/archinternmed.2012.4482
Author Contributions: Dr Gellad had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Gellad, Mor, Donohue, and Good. Acquisition of data: Gellad, Mor, and Zhao. Analysis and interpretation of data: Gellad, Mor, Zhao, Donohue, and Good. Drafting of the manuscript: Gellad and Good. Critical revision of the manuscript for important intellectual content: Gellad, Mor, Zhao, Donohue, and Good. Statistical analysis: Gellad, Mor, and Zhao. Obtained funding: Gellad. Administrative, technical, and material support: Gellad, Donohue, and Good. Study supervision: Gellad, Mor, and Good.
Conflict of Interest Disclosures: Drs Gellad, Mor, Zhao, and Good are all employees of the VA.
Funding/Support: Dr Gellad was supported by a VA Health Service Research and Development Career Development Award (CDA 09-207) and by the VA VISN 4 Competitive Pilot Project Fund (XVA 72-156).
Previous Presentations: Portions of this work were presented at the Society of General Internal Medicine Annual Meeting; May 6, 2011; Phoenix, Arizona.
Disclaimer: The contents represent the views of the authors only and not necessarily those of the VA or the US government. These funding sources had no role in the design or conduct of the study; the collection, analysis, or interpretation of data; or the preparation, review, or approval of the manuscript.
Additional Contributions: Michael Fine, MD, MSc, Joseph Hanlon, PharmD, MS, and John Lowe, RPh, made important contributions to this work.
Create a personal account or sign in to: