Nber Working Paper Series the Impact of Comparative Effectiveness Research on Health and Health Care Spending

Public subsidization of technology assessments in general, and Comparative Effectiveness Research (CER) in particular, has received considerable attention as a tool to simultaneously improve patient health and lower the cost of health care. However, little conceptual and empirical understanding exists concerning the quantitative impact of public technology assessments such as CER. This paper analyses the impact of CER on health and medical care spending interpreting CER to shift the demand for some treatments at the expense of others. We trace out the spending and health implications of such demand shifts in privateas well as subsidized health care markets. In contrast to current wisdom, our analysis implies that CER may well increase spending and adversely affect patient health, particularly when treatment effects are heterogeneous across patients. We simulate these economic effects for antipsychotics that are among the largest drug classes of the US Medicaid program and for which CER has been conducted by means of the CATIE trial in 1999. Using conservative estimates, we find that if Medicaid would have eliminated coverage for the least cost-effective treatments of the CATIE trial then under homogeneous effects, it would save about 90% of the $1.3B Medicaid class sales annually in non-elderly adult patient with schizophrenia. However, taking into account the observed heterogeneity in treatment effects, it would incur a loss of health valued annually at about 98% of class spending and thus a net loss of about 8% of annual class spending. Anirban Basu Section of of Hospital Medicine Department of Medicine University of Chicago 5841 S. Maryland Avenue MC-2007, AMD B226C Chicago, IL 60637 and NBER [email protected] Tomas J. Philipson Irving B. Harris Graduate School of Public Policy Studies The University of Chicago 1155 E 60th Street Chicago, IL 60637 and NBER [email protected] Section 1: Introduction The debate about the appropriate role of public technology assessments has a long history in the US. More recently, as both private and public payers attempt to improve the efficiency of health care spending, comparative effectiveness research (CER) has been offered as a potential solution. The rationale for CER is to generate better evidence about what works and does not work in health care and to thereby improve the productivity of health care spending through improved patient outcomes at lower spending levels. (FCCCER Report 2009; IOM 2009) Moreover, recent public subsidization of CER through the 2009 American Recovery and Reinvestment Act (ARRA) has raised awareness and funding for CER. Although CER has been positioned as a means to improve health and potentially lower costs, little is understood about how exactly this will take place and how CER as it is practiced currently, will translate into different health care decisions. Indeed, despite the importance of comparative effectiveness research in the policy debate, there has been little explicit and quantitative analysis of the potential impact of CER on health or medical care spending. Given this lack of understanding of the consequences of CER, the purpose of this paper is to attempt to provide a framework to quantitatively evaluate the effects of CER. Such a framework is needed to identify relevant designs and studies of CER and also to estimate whether the costs of investments in CER are outweighed by their benefits. The paper may be outlined as follows. Section 2 specifies the economic context in which we analyze the impact of public technology assessments such as CER. We interpret the evidence generated by public subsidies for CER to raise the demand of some treatments at the expense of others. Section 3 considers the healthand costs implications of such demand shifts induced by CER in a private market. Section 4 analyzes these impacts in a subsidized market where the treatments that fare better under CER are the ones that receive better coverage, e.g. through less formulary restrictions, changes in prior authorization requirements, or lower co-pays. They date back at least to the 1970s with the National Center for Health Services Research and the US Congress Office of Technology Assessment. Section 5 discusses the impacts of such CER-responsive coverage decisions when there is heterogeneity in treatment effects. An important issue here is that CER may favor one treatment, deemed “the best” by some summary statistic, even though the best treatment varies across patients. Heterogeneity is necessary but not sufficient for welfare losses from product-specific reimbursement responses to CER. Rather, what’s central is negative dependence across treatments; a patient may not respond to a reimbursed “winner” of CER study but may respond to a non-reimbursed “loser”. Overall, our main conclusions from the conceptual analysis is that CER has indeterminate effects on spending and patient health and, under natural assumptions on how markets respond to new quality information, may even adversely affect both. Among the factors that govern the health and spending impact of CER are the price-elasticity of supply of treatments as well as the evidence-elasticity of demand. Section 6 provides an illustrative empirical simulation for antipsychotic drugs. This drug class is among the top spending classes covered by Medicaid and there has been considerable interest in the comparative effectiveness issues surrounding them. This is partly due to the 1999 CATIE trial that found that second generation therapies were equally effective as first-generation therapies. Following the release of this study, a policy debate emerged around how Medicaid coverage policy might respond. Many argued that Medicaid should preferentially subsidize first generation over second generation treatments. Subsequent to the release of CATIE, approximately 40% of the state-run Medicaid programs have instituted prior authorization restrictions on some second generation drugs. Little is known about the quantitative health and spending effects of such coverage proposals and policies. To better understand such effects, we simulate the impact on health and spending for the counter-factual case of providing no coverage for second generation antipsychotics. We find that if Medicaid coverage policy had responded to the CER results of CATIE to by limiting second generation use then under a standard assumption of homogeneous treatment effects, it would save about 90% of class spending. However, taking into account the observed heterogeneity in treatment effects, limiting access to second generation drugs would incur a net loss valued at about 98% of class spending. Thus, for This reimbursement response is inconsistent with current law, but our analysis is aimed at understanding the desirability of future policies and laws that may potentially those currently in place. one the largest drug classes in one of the major health care subsidy programs, heterogeneity alters not only the magnitude but also the sign of the effect reimbursement responses to CER