Variation in prescription use among patients with diabetes in the VA Healthcare System

Gellad WF, Mor MK, Zhao X, Donohue JM, Good CB, Fine MJ. Variation in prescription use among patients with diabetes in the VA Healthcare System. Paper presented at: Society of General Internal Medicine Annual Meeting; 2011 May 5; Phoenix, AZ.

Related HSR&D Project(s)

• CDA 09-207 – Addressing Regional Variation in Drug Prescribing and Spending in the VA

BACKGROUND: Regional variation in healthcare use has become a primary indicator of inefficiency in the healthcare system, yet little is known about variation in medication use. The Veterans Health Administration (VA) offers a unique setting to assess regional variation in prescription use, because of its unified national formulary and pharmacy benefit. We sought to examine regional variation in outpatient medication use for the treatment of diabetes. METHODS: We conducted a cross-sectional analysis of all Veterans with type 2 diabetes managed in the VA in 2009 (n = 1,160,895). The cohort was identified based on inpatient or outpatient ICD-9 codes for diabetes or receipt of a hypoglycemic medication. We used VA Pharmacy Benefit Management data to identify all diabetes medications dispensed for the cohort in 2009, aggregated at the VA facility-level. We examined two outcomes: 1) the percentage of patients on oral hypoglycemic drugs at each VA who filled a prescription for a thiazolidinedione (TZD, i.e., rosiglitazone, pioglitazone), which are a brand-name- only expensive class of medication, and 2) the percentage of patients on insulin at each VA who filled a prescription for a long-acting analogue insulin (i.e., detemir, glargine), which are expensive long-acting insulins. We report descriptive statistics and quantify variation using the coefficient of variation (standard deviation/mean x 100). We used Pearson correlation coefficients to assess whether VA facilities with a high proportion of patients on TZDs also have a high proportion of patients on long-acting insulin analogues. We developed multivariable logistic regression models, with fixed effects for each facility, to model the odds of each outcome adjusting for patient age, gender, race/ethnicity, number of providers, and being prescribed medications by a physician (vs. non-physician). RESULTS: Overall, 908,721 (78.3%) of diabetics received a hypoglycemic medication, totaling 6,194,339 prescriptions in 2009. Across 129 VA facilities, the percentage of patients at each facility on TZDs ranged from 1.5% to 26.3%, with a mean of 8.3% (coefficient of variation, 54.2%). The percentage of patients on insulin who used long-acting insulin analogues ranged from 3.7% to 71.4%, with a mean of 38.4% (coefficient of variation, 40.6%). There was a weak correlation between use of TZDs and long-acting insulin analogues (r = 0.28, p < .001); VA facilities with a high proportion of patients on TZDs tend to be facilities with a high proportion of patients on long-acting insulin analogues. After controlling for patient factors, the odds of receipt of TZDs at each facility ranged from 0.11 to 1.93, and the odds of receipt of an insulin analogue ranged from 0.02 to 1.27 (p < .001). CONCLUSION: Significant practice-pattern variation exists across VA medical centers in the use of higher-cost hypoglycemic medications, despite a uniform national formulary and extensive utilization management. This substantial variation is unlikely to be explained by patient factors alone. The provider, facility, and larger regional factors that lead to this variability should be examined, because of the significant cost and efficiency implications of this variation in prescribing.