Using policy simulation to predict drug plan expenditure when planning reimbursement changes.

BACKGROUND: Drug plan decision makers need accurate financial impact projections before the implementation of new drug policy initiatives. Tools for such projections need to have small margins of error and be based on methodology that is easy to communicate to stakeholders. Ad hoc methods typically used for financial impact projections by health plans are inadequate.
OBJECTIVE: To present a flexible tool for projecting the financial impact of drug policy changes based on historical dispensing data and simulation, and explore its validity using a recent example of a complex drug policy change in British Columbia, Canada.
METHODS: A policy simulator (SAS) program using a Web browser interface) was used to produce 3-year forecasts of expenditure (for the drug plan and for individual families) along with the number of patients who would pay more or less for their drugs (stratified by age and income level) for various proposed drug policies starting in 2003. Drug expenditure under each policy was simulated based on projections from prescription claim records of the British Columbia PharmaNet database of community pharmacy prescriptions from 1 January 2001 to 31 December 2001. The simulator selected a random 1% sample of British Columbia families (175,000 families) in the database. Once the new drug policy was selected and implemented, the accuracy of the predictions were tested by comparing the actual PharmaCare expenditure for the period 1 May 2003 to 31 March 2004 after implementation of the new drug policy with the final simulation made for this policy in February 2003, 2 months before the policy was implemented.
RESULTS: The policy simulation tool produced hundreds of variations for decision makers in the months before the final policy rules were decided upon. When compared with actual drug expenditure after policy implementation, it was found that the tool had predicted spending with <1% error for the first 11 months after introduction of the policy. As well as producing accurate expenditure forecasts for the insurer, the tool was able to predict how family out-of-pocket expenditure would be affected.
CONCLUSIONS: The simulator aided drug policy planning and communication. The tool provided rapid and accurate results that were communicated easily to decision makers. Such policy simulation can be applied to a wide range of health plans and policy changes.