• Budgetary Impact of TRICARE Antidiabetic Drug Formulary Changes

      Hung, Anna; Mullins, C. Daniel (2018)
      Background: There is guidance on how to conduct budget impact analyses (BIA) from the International Society of Pharmacoeconomics and Outcomes Research (ISPOR). However, there is a growing need to ensure that budget impact models are valid, accurate, and usable to payers. In 2016, the Defense Health Agency implemented antidiabetic formulary changes to the TRICARE pharmacy benefit. The goal of this dissertation is to predict and validate the budget impact of these formulary changes. Objective: The specific objectives of this study were to: i) estimate the annual financial consequences of antidiabetic formulary changes from the TRICARE payer perspective using TRICARE claims data over three years; ii) assess the face validity, internal validity, and predictive validity of the model; and iii) identify and compare cost drivers of both the budget and the budget impact identified through the model versus the empirical analysis. Methods: Following the ISPOR BIA guidance, a budget impact model was created in Microsoft Excel®. The counterfactual was predicted using autoregressive integrated moving average models. One year after the formulary changes, actual utilization was used to determine the realized budget and compare this to what was predicted in the budget impact model. Cost drivers were determined through one-way sensitivity analyses, subgroup analyses, and the comparison of utilization versus price in the model versus the empirical analysis. Results: In the year after the formulary changes, the model predicted a lower budget impact than what was realized ($24 million in savings versus $49 million in savings). Meanwhile, the model predicted a higher annual budget than what was realized ($686 million versus $609 million). The higher-than-predicted savings was largely due to lower utilization seen in the empirical analysis compared to the model. The antidiabetic drug classes that contributed most to these savings were the dipeptidyl peptidase-4 inhibitors, glucagon-like receptor-1 agonists, and sodium-glucose cotransporter 2 inhibitors. Conclusion: Future budget impact models should be validated by waiting at least one year and comparing model predictions to what is realized. The end user of the model should also be involved in the process of creating the model.