Modeling drivers of specialty drug prescribing decisions in claims data

Many health care payers pay for physician services at different rates depending on whether the service is delivered in a physician office of hospital outpatient department (HOPD) setting.  Payments for drugs administered to patients by physicians to treat cancer and a range of other conditions are usually an exception to this rule and are reimbursed at the same rate in both settings.  Despite identical payment for drugs themselves, providers may benefit financially by steering patients to the HOPD setting due to higher rates for ancillary services and lower drug costs under the 340B drug discount program.

Our study: (1) examines changes in prescriber decisions regarding the choice of self-administered versus physician-administered drugs, the site of service for physician-administered drugs, and total drug cost over time and associated with key policy changes; and (2) models treatment decisions for individual patients.  We use 2013 through June 2014 multi-payer medical and pharmacy claims data from Symphony Health Solutions.  Our study sample includes all of the medical claims submitted by a panel of nearly 10,000 physicians that we track longitudinally.  We also observe prescriptions written by this sample of providers that are filled in a very large convenience sample of retail pharmacies. 

We are beginning analyses now and will have results to present in early 2016. Our specific approach is to first test for differences within physician, over time, and across payers in the share of drugs by volume and by charges that are (a) administered by the physician versus dispensed at retail pharmacies, and (b) among physician-administered drugs, the proportion administered in the HOPD setting.  Our study sample crosses a range of interesting exogenous shocks – including the ACA coverage expansion, a reduction in Medicare payments for physician-administered drugs, an expansion of eligibility for the 340B drug discount program, and several new drug approvals.  It also covers a period of general consolidation in health care delivery.  We will therefore separately test for an overall trend and responses to specific policy changes.  As a second step, we will fit models for two prescribing decisions – retail versus physician-administered and, for physician-administered drugs, office versus HOPD setting – within provider, at the patient level, and controlling for diagnoses, payer, patient demographics, and a time trend specified to reflect our findings from the first step.  The key independent variables of interest are the market reference price (ASP) and a post-ACA flag and time trend interaction to reflect 340B eligibility changes and the coverage expansion. 

A deeper understanding of how providers make decisions regarding prescribing – including whether or not to initiate treatment with a physician-administered drug and, if so, in what site of service – will help payers update payment policy and respond to consolidation in the health care delivery system without jeopardizing access for patients. This study adds insight into specialty drug prescribing patterns payers other than Medicare – where much of the investigation into physician-administered drugs to date has focused.