Tracking early ACA coverage expansion impacts in pharmacy transaction data

The key coverage expansion provisions of the Affordable Care Act (ACA) are expected to shift millions of uninsured Americans into health insurance plans.  These provisions, including the expansion of Medicaid programs in some states, the creation of new health insurance marketplaces, and subsidies for low-income individuals to purchase plans through marketplaces, take effect on January 1, 2014.  Since the passage of the ACA in 2010, analysts have applied qualitative and quantitative methods to gauge the likely impact of the coverage expansion provisions.

While government administrative and survey data may ultimately be used to evaluate the impacts of these core ACA provisions, these data are not immediately available.  To get a timely assessment of the early effects of the ACA coverage expansion, we are using prescription drug transaction data, and particularly transitions in the primary payment source for claims in a stable panel of prescription drug users spanning the January 1, 2014 ACA coverage expansion implementation date.  Our sample includes all January 2012 to mid-2014 prescription transactions from a major pharmacy claims aggregator (IMS Health) for a cohort of 15 million individuals.  The transaction data are longitudinal and linked using an anonymous patient ID.

Because pharmacy claims are adjudicated in near-real time, they are available within days rather than months after the transaction occurs.  In addition, patients regularly fill prescriptions for certain common, chronic conditions, resulting in a useful longitudinal, panel data set in which transitions from one payer category to another can be easily observed.  Additional advantages of the data include that they are from a large, nationally-representative sample of pharmacies, and that they include cash-only transactions.

Our main descriptive analyses will track transitions from one payer category to another over time, focusing on the January 1, 2014 date as an inflection point.  In addition, we will model transition outcomes as a function of individual demographic characteristics, state of residence, and a time trend.  We expect considerable state variation in the impact of the expansion. Most directly, whether states opt to participate in the Medicaid expansion will determine the extent of increases in Medicaid coverage in those states. As a result, we will analyze trends separately for expansion and non-expansion states. 

A limitation of our approach is that, although we observe cash transactions, uninsured individuals may have been less likely to obtain prescription drugs in the pre-2014 period.  To address this concern, we will conduct sensitivity analyses limiting the sample to individuals with chronic conditions, since utilization among this group may be relatively insensitive to insurance coverage.  We will separately analyze transition trends for several diagnosis-defined subpopulations, including individuals with one of five chronic conditions and individuals with a prescription for one of five therapeutic classes associated with a chronic condition. 

The study described above is funded and in-process.  We will report results for the calendar months of data that are delivered and analyzed prior to the ASHE conference (likely through Q1 2014).