For each eligible prescription submitted, we constructed explanatory variables to assess the relationship between physician, patient, treatment, and pharmacy characteristics and dispense as written use and prescription filling. Physician variables included primary specialty, practice type (primary care, specialist, non-physician prescriber), and prescriber age. Patient characteristics included age (in years), gender, and US census region of residence. Treatment variables included the dispense as written assignment, GPI4/GPI2-designated therapeutic class, brand/generic status, and patient out-of-pocket cost (in dollars per 30-day equivalent prescription). Pharmacy characteristics included the type of dispensing pharmacy (retail or mail). Prescriptions were categorized as either acute or maintenance (chronic) using the First Data Bank designation. Maintenance medications were further categorized as either an “initiation” or “continuation” of therapy. Initiation prescriptions were defined on the basis of no paid pharmacy claims for a drug in the same therapeutic class in the 6 months before the index prescription claim. Maintenance continuation prescriptions were preceded by 1 or more paid claims in the previous 6 months, indicating recent use.
Viagra super active online
Analysis Plan
We used descriptive statistics to evaluate patient, physician, pharmacy, and prescription characteristics. We described rates of dispense as written for both single-source brands and multi-source brands, despite the fact that dispense as written for single-source brands may not have any effect on prescription delivery. We also present rates of prescription reversals, prescription claims approved by a payer and then reversed by the pharmacy because they were not purchased by the patient and went unfilled, stratified by dispense as written designation and prescription type.

To assess the relationship between physician, patient, prescription, and pharmacy characteristics with physician and patient dispense as written requests, we used generalized estimating equations to adjust for clustering at the patient level. Our outcomes, at the submitted prescription level, were the presence or absence of physician dispense as written in one model and the presence or absence of patient dispense as written for the other. We studied whether physician, member, treatment, and pharmacy characteristics were associated with the submission of prescriptions with a dispense as written designation. When comparing rates of dispense as written requests by drug class, we selected oral diabetes medications as our referent category because they are essential medications, commonly prescribed, and include both generic and brand-name options.

Multivariate generalized estimating equation models were used to estimate the relationship between patient and physician dispense as written selection and whether the claim was reversed, indicating the medication was not purchased by the patient and went unfilled. In these models, we were interested in the relationship between dispense as written designation and rates of multi-source brand medication filling, because these are the medications for which dispense as written designations most clearly affect the medication received. Thus, in our primary model, we included only multi-source brand and generic medications. We ran a distinct model with single-source brands as a neutral control because we did not expect that dispense as written designation would have any effect on the medication that was delivered and, as a result, the likelihood of actual purchasing. In these models, we controlled for patient, physician, and pharmacy covariates and adjusted for clustering within patients. We included interactions between physician and patient dispense as written designations and prescription characteristics (initiation of a chronic medication, maintenance medication continuation, or acute medication), because we hypothesized that dispense as written designation may have the greatest effect on purchasing rates in new prescriptions or acute prescriptions, when patients first learn about the medication costs. Statistical evaluations were performed using SAS Version 9.1 with SAS/STAT(r) (SAS Institute Inc, Cary, NC) and Stata SE 9.1 for Windows (StataCorp LP, College Station, Tex).