Among multi-source brands, physician dispense as written requests were associated with increased rates of prescription reversals (OR 1.16, P<;.001), indicating that patients did not fill these prescriptions. New prescriptions (ie, the first maintenance prescriptions filled in a chronic medication class) had greater odds of reversal than subsequent maintenance prescriptions (OR 2.07, P<;.001). Rates of reversal also were higher for acute medications compared with maintenance medications (OR 1.37 P<;.001). Strong relationships were seen when we tested the interactions of dispense as written requests and initial maintenance or acute prescriptions. Among initial maintenance prescriptions, physician dispense as written (OR 1.50, P<;.001) and patient dispense as written (OR 1.60, P<;.001) were associated with greater odds of reversal. Similar trends were seen for acute medications; physician dispense as written was associated with 1.42 greater odds and patient dispense as written was associated with 1.61 greater odds of reversal compared with those filling subsequent maintenance medications, indicating higher rates of patient failure to fill these prescriptions. Among single-source brand medications, the neutral control where dispense as written designations should have no effect on which medication is actually filled, we found little effect of dispense as written designation on reversal rates. Overall, physician dispense as written was associated with a small reduction in reversal rates (OR 0.89, P<;.001). Acute medications and new prescriptions for maintenance medications were more likely to be reversed (OR 2.2 and 2.9, respectively; P<;.001 for both), although the odds of reversal was not influenced by dispense as written status. Discussion

In this national sample of prescriptions written for patients receiving drug coverage administered by a large pharmacy benefits manager, approximately 5% of all prescriptions included a dispense as written designation requesting dispensing of a brand product. Dispense as written requests were made by prescribers (2.7% of prescriptions) and patients (2.0% of prescriptions). Prescriptions written with dispense as written designations were more likely to be reversed, indicating that they were less likely to be purchased by patients and went unfilled. In particular, when chronic therapy was initiated, physician and patient dispense as written requests led to more than 50% greater odds of non-filling.

By substituting the generic alternative for each multi-source brand that was filled after both physician and patient dispense as written designation, the patient population in this sample could have reduced their charges by more than $1.7 million and the health plans could have experienced more than a $10.6 million reduction in costs in the 1-month study period. By assuming a similar rate of dispense as written requests in uninsured patients, patients covered by state or federal governments, and other commercially insured beneficiaries, we can estimate the savings potential of a policy that eliminates the dispense as written option. With more than 3.6 billion prescriptions filled in the United States annually,14 patient charges could be reduced by as much as $1.2 billion annually and health system costs could be reduced by as much as $7.7 billion by eliminating dispense as written opportunities. We are unable to estimate the implications of specifying dispense as written for single-source brands because these designations likely have minimal effect on the actual prescriptions delivered.