Medicare beneficiaries and the impact of gaining prescription drug coverage on inpatient and physician spending.(Medicare Prescription Drug, Improvement, and Modernization Act of 2003)

Far-reaching changes in the Medicare program are expected with the passage of the Medicare Prescription Drug, Improvement, and Modernization Act of 2003 (PL 108-172), not the least of which is planning for the expenditures of the approximately 8 million beneficiaries who lacked any drug benefits beforehand. This group is particularly intriguing as they are likely to see the most significant improvement in their access to medications: if lack of drug coverage posed a barrier to getting necessary medications then beneficiaries without any drug coverage should exhibit considerable increases in prescription drug use once enrolled in the Medicare benefit. For policy makers trying to anticipate the impact of giving drug benefits to people without prior coverage, there are many questions about health and quality of life, but two relate to costs: (1) how much will the new benefit directly impact drug spending and (2) how much will it indirectly affect other medical spending, such as hospitalization and physician visits? The first question is straightforward and has been the subject of some study, although estimates vary depending on the study population and the analytic method. The second question steins from a less studied but widely held view that providing drug coverage may create cost-offsets through savings in other health care expenditures. The clinical literature cites many examples where pharmacological interventions reduce emergency and acute care treatments, although it is unclear how these findings translate to an entire population with diverse medical conditions. Both issues are important for informing expectations about the Medicare drug program as projections predict that the majority of Medicare beneficiaries without drug coverage will participate (Shea, Stuart, and Briesacher 2003-2004). This study addresses the questions framed above through comparisons of before and after spending patterns of Medicare beneficiaries who pick up drug coverage (Gainers), relative to those who never had it (Nevers). The analysis used 6 years of data from the Medicare Current Beneficiary Survey and modeling techniques suitable for studying insurance transitions. We treated gaining coverage as plausibly exogenous for the following reasons. Our study covered a period of large expansions in Medicare drug coverage, from 65 percent in 1995 to 77 percent in 1999 (Briesacher, Stuart, and Shea 2002). Previous studies have found most changes in private health plans are because of external circumstances of different plan offerings or new enrollment opportunities rather than pure consumer choice (Cunningham and Kohn 2000; Stuart, Shea, and Briesacher 2001; Rice et al. 2002). In addition, we tested the exogeneity assumption in this study by reestimating the models with only individuals least likely to select into drug coverage--those who gained employer-sponsored coverage.

We expected from prior research to find increases in medication use after gaining prescription coverage because of decreases in patients' out-of-pocket prices (Leibowitz, Manning, and Newhouse 1985; Gianfrancesco, Baines, and Richards 1994; Gluck 1999; Lillard, Rogowski, and Kingion 1999; Pauly 2004). However, the literature was less clear about how medications influence spending for physician services and hospitalization. Economic theory specifies that the relationships depend on whether the medical services are complements or substitutes. The empirical research is quite meager on this topic, although it seems reasonable to believe physician services are a complement to medications since patients must visit physicians to get prescriptions. In contrast, hospitalizations may act as substitutes if outpatient medications prevent acute events, although this association is also largely untested in diverse patient cohorts like the Medicare population. Thus, we had a cautious expectation of an inverse relationship between inpatient costs and prescription drugs.

Our analytical approach addresses two important weaknesses in the literature on health insurance coverage and medical care utilization. The first is our use of longitudinal data. Most analyses use cross-sectional data, which cannot be used to assess changes in medical care need or deteriorations in health (Kasper, Giovanni, and Hoffman 2000). This makes it difficult, if not impossible, to isolate the impact of insurance on health care spending from other influences because of differences within and across individuals. If, instead, individuals are followed over time, the temporal path of medical expenditures can be identified prior to the decision to acquire insurance coverage. Our longitudinal analysis assessed this observable element of selection as well as used a fixed effects framework to control for time invariant factors.

The second contribution of this study is a narrow focus on gaining drug coverage. Analyses of losing insurance may be less useful for understanding the impact of acquiring coverage if these behaviors are not inversely related (Burstin et al. 1998; Long, Marquis, and Rodgers 1998). For example, individuals who lose their coverage may attempt to maintain the levels of care established while insured, in contrast to people who gain coverage and must learn to use the health care system as insured people (Stuart and Coulson 1993). In a study of adults who gained Medicaid or private health insurance, Kasper, Giovanni, and Hoffman (2000) detected relatively high levels of access problems for Gainers even after becoming insured. Alternatively, research that combines the gaining and losing samples together as the "intermittently uninsured" may be masking the unique issues facing people when acquiring insurance (McWilliams et al. 2003).

The literature specific to drug coverage is growing but much of it suffers from the methodological issues described above (Stuart and Coulson 1993; Lillard, Rogowski, and Kington 1999; Atherly 2002). Studies with pre- and postenrollment data on expansions in Medicaid or state drug assistance programs offer mixed findings. A large decline in hospital spending followed an expansion in the drug formulary of the South Carolina Medicaid program for an under age 65 population (Kozma, Reeder, and Lingle 1990). Similarly, Lingle, Kirk, and Kelly (1987) detected lower inpatient expenditures after the implementation of…