Background: All US states have adopted generic substitution laws to reduce medication costs.  However, physicians may override these regulations by prescribing branded drugs and requesting that they are dispensed as written.  Patients also can make these requests.  Little is known about the frequency and correlates of dispense as written requests or their association with medication filling.

Methods: We identified beneficiaries of a large pharmacy benefits manager who submitted a prescription claim from any pharmacy in Jan 2009.  We categorised claims as a physician-assigned dispense as written, patient-assigned dispense as written, or no dispense as written.  We described rates of these requests and used generalised estimating equations to evaluate physician, patient, treatment and pharmacy characteristics associated with dispense as written requests.  We also used generalised estimating equations to assess the relationship between dispense as written designation and rates prescriptions are not filled by patients.

Results: Our sample included 5.6 million prescriptions for more than 2 million patients.  More than 2.7% were designated as dispense as written by physicians, and 2.0% were designated as dispense as written by patients.  Substantial variation in dispense as written requests were seen by medication class, patient and physician age, and geographical region.  The odds of requesting dispense as written was 78.5% greater for specialists than generalists (P less than 0.001).  When chronic prescriptions were initiated, physician dispense as written (odds ratio 1.50, P less than 0.001) and patient dispense as written (odds ratio 1.60, P less than 0.001) were associated with greater odds that patients did not fill the prescription.

Conclusions: Dispense as written requests were common and associated with decreased rates of prescription filling.  Options to reduce rates of dispense as written requests might reduce costs and improve medication adherence.